This document discusses applying machine learning techniques to agricultural data. It describes a software tool called WEKA that allows experimenting with different machine learning algorithms on real-world datasets. As a case study, the document examines using machine learning to infer rules for culling less productive cows from dairy herd data. Several machine learning methods were tested on the data and produced encouraging results for using machine learning to help solve agricultural problems.
Machine Learning in Agriculture Module 1Prasenjit Dey
Discuss the opportunities of incorporation of machine learning in agriculture. Briefly discuss different machine learning strategies. Briefly discuss the ways of machine learning can be used
Application of Machine Learning in AgricultureAman Vasisht
With the growing trend of machine learning, it is needless to say how machine learning can help reap benefits in agriculture. It will be boon for the farmer welfare.
Machine Learning for Disease PredictionMustafa Oğuz
A great application field of machine learning is predicting diseases. This presentation introduces what is preventable diseases and deaths. Then examines three diverse papers to explain what has been done in the field and how the technology works. Finishes with future possibilities and enablers of the disease prediction technology.
prospects of artificial intelligence in agVikash Kumar
This document provides an overview of artificial intelligence (AI) and its applications in agriculture. It discusses how AI is used in agriculture for automated farming activities, pest and disease identification, crop quality management, and environmental monitoring. The document also covers perspectives on AI progression, from narrow to general to super AI. It discusses recent AI developments in India and applications in agriculture like precision farming, yield prediction, and optimized resource use. Limitations of AI include data and infrastructure challenges. The document concludes that AI can boost agriculture through optimized resource use and complement farmer decision making.
Artificial Intelligence is an approach to make a computer, a robot, or a product to think about how smart humans think. AI is a study of how the human brain thinks, learns, decides and work when it tries to solve problems. And finally, this study outputs intelligent software systems. The aim of AI is to improve computer functions that are related to human knowledge, for example, reasoning, learning, and problem-solving.
This Machine Learning Algorithms presentation will help you learn you what machine learning is, and the various ways in which you can use machine learning to solve a problem. At the end, you will see a demo on linear regression, logistic regression, decision tree and random forest. This Machine Learning Algorithms presentation is designed for beginners to make them understand how to implement the different Machine Learning Algorithms.
Below topics are covered in this Machine Learning Algorithms Presentation:
1. Real world applications of Machine Learning
2. What is Machine Learning?
3. Processes involved in Machine Learning
4. Type of Machine Learning Algorithms
5. Popular Algorithms with a hands-on demo
- Linear regression
- Logistic regression
- Decision tree and Random forest
- N Nearest neighbor
What is Machine Learning: Machine Learning is an application of Artificial Intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed.
- - - - - - - -
About Simplilearn Machine Learning course:
A form of artificial intelligence, Machine Learning is revolutionizing the world of computing as well as all people’s digital interactions. Machine Learning powers such innovative automated technologies as recommendation engines, facial recognition, fraud protection and even self-driving cars.This Machine Learning course prepares engineers, data scientists and other professionals with knowledge and hands-on skills required for certification and job competency in Machine Learning.
- - - - - - -
Why learn Machine Learning?
Machine Learning is taking over the world- and with that, there is a growing need among companies for professionals to know the ins and outs of Machine Learning
The Machine Learning market size is expected to grow from USD 1.03 Billion in 2016 to USD 8.81 Billion by 2022, at a Compound Annual Growth Rate (CAGR) of 44.1% during the forecast period.
- - - - - -
What skills will you learn from this Machine Learning course?
By the end of this Machine Learning course, you will be able to:
1. Master the concepts of supervised, unsupervised and reinforcement learning concepts and modeling.
2. Gain practical mastery over principles, algorithms, and applications of Machine Learning through a hands-on approach which includes working on 28 projects and one capstone project.
3. Acquire thorough knowledge of the mathematical and heuristic aspects of Machine Learning.
4. Understand the concepts and operation of support vector machines, kernel SVM, naive Bayes, decision tree classifier, random forest classifier, logistic regression, K-nearest neighbors, K-means clustering and more.
5. Be able to model a wide variety of robust Machine Learning algorithms including deep learning, clustering, and recommendation systems
- - - - - - -
A confluence of factors have converged to afford the opportunity to apply data science at large scale to agricultural production. The demand for agricultural outputs is growing and there is a need to meet this demand by utilizing increasingly mechanized precision agriculture and enormous data volumes collected to intelligently optimize agriculture outputs. We will consider the machine learning challenges related to optimizing global food production.
Artificial intelligence can benefit the agriculture sector by increasing productivity and sustainability. As the global population grows, AI technologies like drones, automated systems, agricultural robots, remote sensing, and decision support systems can help monitor crop conditions, identify issues, automate processes, and support farmers' decisions. While these applications may have initial financial and expertise barriers, their benefits include enhanced crop yields, quality and safety, efficient farm management, and reduced risks. Overall, AI can help modernize agriculture and optimize outputs to better feed the world's growing population.
Machine Learning in Agriculture Module 1Prasenjit Dey
Discuss the opportunities of incorporation of machine learning in agriculture. Briefly discuss different machine learning strategies. Briefly discuss the ways of machine learning can be used
Application of Machine Learning in AgricultureAman Vasisht
With the growing trend of machine learning, it is needless to say how machine learning can help reap benefits in agriculture. It will be boon for the farmer welfare.
Machine Learning for Disease PredictionMustafa Oğuz
A great application field of machine learning is predicting diseases. This presentation introduces what is preventable diseases and deaths. Then examines three diverse papers to explain what has been done in the field and how the technology works. Finishes with future possibilities and enablers of the disease prediction technology.
prospects of artificial intelligence in agVikash Kumar
This document provides an overview of artificial intelligence (AI) and its applications in agriculture. It discusses how AI is used in agriculture for automated farming activities, pest and disease identification, crop quality management, and environmental monitoring. The document also covers perspectives on AI progression, from narrow to general to super AI. It discusses recent AI developments in India and applications in agriculture like precision farming, yield prediction, and optimized resource use. Limitations of AI include data and infrastructure challenges. The document concludes that AI can boost agriculture through optimized resource use and complement farmer decision making.
Artificial Intelligence is an approach to make a computer, a robot, or a product to think about how smart humans think. AI is a study of how the human brain thinks, learns, decides and work when it tries to solve problems. And finally, this study outputs intelligent software systems. The aim of AI is to improve computer functions that are related to human knowledge, for example, reasoning, learning, and problem-solving.
This Machine Learning Algorithms presentation will help you learn you what machine learning is, and the various ways in which you can use machine learning to solve a problem. At the end, you will see a demo on linear regression, logistic regression, decision tree and random forest. This Machine Learning Algorithms presentation is designed for beginners to make them understand how to implement the different Machine Learning Algorithms.
Below topics are covered in this Machine Learning Algorithms Presentation:
1. Real world applications of Machine Learning
2. What is Machine Learning?
3. Processes involved in Machine Learning
4. Type of Machine Learning Algorithms
5. Popular Algorithms with a hands-on demo
- Linear regression
- Logistic regression
- Decision tree and Random forest
- N Nearest neighbor
What is Machine Learning: Machine Learning is an application of Artificial Intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed.
- - - - - - - -
About Simplilearn Machine Learning course:
A form of artificial intelligence, Machine Learning is revolutionizing the world of computing as well as all people’s digital interactions. Machine Learning powers such innovative automated technologies as recommendation engines, facial recognition, fraud protection and even self-driving cars.This Machine Learning course prepares engineers, data scientists and other professionals with knowledge and hands-on skills required for certification and job competency in Machine Learning.
- - - - - - -
Why learn Machine Learning?
Machine Learning is taking over the world- and with that, there is a growing need among companies for professionals to know the ins and outs of Machine Learning
The Machine Learning market size is expected to grow from USD 1.03 Billion in 2016 to USD 8.81 Billion by 2022, at a Compound Annual Growth Rate (CAGR) of 44.1% during the forecast period.
- - - - - -
What skills will you learn from this Machine Learning course?
By the end of this Machine Learning course, you will be able to:
1. Master the concepts of supervised, unsupervised and reinforcement learning concepts and modeling.
2. Gain practical mastery over principles, algorithms, and applications of Machine Learning through a hands-on approach which includes working on 28 projects and one capstone project.
3. Acquire thorough knowledge of the mathematical and heuristic aspects of Machine Learning.
4. Understand the concepts and operation of support vector machines, kernel SVM, naive Bayes, decision tree classifier, random forest classifier, logistic regression, K-nearest neighbors, K-means clustering and more.
5. Be able to model a wide variety of robust Machine Learning algorithms including deep learning, clustering, and recommendation systems
- - - - - - -
A confluence of factors have converged to afford the opportunity to apply data science at large scale to agricultural production. The demand for agricultural outputs is growing and there is a need to meet this demand by utilizing increasingly mechanized precision agriculture and enormous data volumes collected to intelligently optimize agriculture outputs. We will consider the machine learning challenges related to optimizing global food production.
Artificial intelligence can benefit the agriculture sector by increasing productivity and sustainability. As the global population grows, AI technologies like drones, automated systems, agricultural robots, remote sensing, and decision support systems can help monitor crop conditions, identify issues, automate processes, and support farmers' decisions. While these applications may have initial financial and expertise barriers, their benefits include enhanced crop yields, quality and safety, efficient farm management, and reduced risks. Overall, AI can help modernize agriculture and optimize outputs to better feed the world's growing population.
Artificial intelligence has the potential to help address challenges facing the agricultural sector as the global population increases. New technologies like drones, driverless tractors, automated irrigation, and machine learning are helping farmers monitor crops and soils, apply inputs precisely, and increase yields. Startups are developing tools using computer vision, satellites, and deep learning to diagnose plant health, predict weather, and optimize resource use. These AI solutions aim to help farmers "do more with less" and help feed the world's growing population in a sustainable way.
The document discusses the concept of Internet of Things (IoT) and its applications in agriculture. It defines IoT and describes how physical objects can be connected to collect and exchange data. Some key applications of IoT in agriculture mentioned include monitoring soil moisture and temperature for controlled irrigation, livestock monitoring, pest monitoring, and mobile money transfers. However, constraints for implementing IoT in Indian agriculture include small land holdings, connectivity and affordability issues. Some case studies on precision agriculture and reducing water usage through IoT are also summarized.
Internet of Things (IoT) is the internetworking of physical devices. This system has the ability to transfer data over a network. Mostly without requiring human intervention.Internet-connected to the physical world via ubiquitous sensors.
It is connecting each and everything to the internet.
Deep learning is a type of machine learning that uses multiple processing layers to learn representations of data with features that become more complex at each layer. Deep learning has achieved human-level performance in areas like image recognition by learning from large datasets. In healthcare, deep learning has been applied to tasks like detecting pneumonia from chest X-rays and skin cancer from images with accuracy comparable to doctors. However, challenges remain around data variability, uncertainty, class imbalance, and data annotation. Cross-area collaboration and data sharing are seen as key to realizing the potential of deep learning in healthcare.
The document discusses the benefits of exercise for mental health. Regular physical activity can help reduce anxiety and depression and improve mood and cognitive functioning. Exercise boosts blood flow, releases endorphins, and promotes changes in the brain which help regulate emotions and stress levels.
Remote Sensing (RS), UAV/drones, and Machine Learning (ML) as powerful techni...nitinrane33
Precision agriculture utilizes modern technology to optimize agricultural practices, resulting in increased productivity while reducing costs and environmental impact. The use of remote sensing (RS), drones or unmanned aerial vehicles (UAVs), and machine learning (ML) has significantly transformed precision agriculture. These advanced technologies provide farmers with accurate, cost-effective, and timely tools to manage crops and resources effectively. This paper evaluates the use of these techniques in precision agriculture, including their benefits, and effective applications. Remote sensing involves using satellites, aircraft, or drones to collect data on crops and the environment, such as soil moisture, temperature, and vegetation indices. With high-resolution images and three-dimensional maps of crops, UAVs enable farmers to identify and address issues like pest infestations or nutrient deficiencies. Machine learning algorithms analyze large amounts of data to predict crop yields, optimize irrigation and fertilization, and identify areas of the field that need attention. Several case studies highlight the effectiveness of these techniques in different agricultural settings. However, the paper also acknowledges the challenges associated with adopting these technologies, such as cost, data management, and regulatory issues. While the initial investment in drones and sensors may be high, the long-term benefits in terms of increased yields, reduced costs, and environmental sustainability are substantial. Farmers need to be trained in the use of these technologies to make informed decisions, and effective data management and analysis are crucial. Additionally, regulatory frameworks are still evolving, and clear guidelines are required for data privacy, safety, and ethical use. Although challenges remain, the benefits of increased productivity, reduced costs, and environmental sustainability make these technologies an attractive investment for farmers worldwide.
Artificial Intelligence In Agriculture & Its Status in IndiaJanhviTripathi
Worldwide, agriculture is a $5 trillion industry, and with the ever increasing population, the world will need to produce 50% more food by 2050 which cannot be accomplished with the percentage of land under cultivation. Factors such as climate change, population growth and food security concerns have propelled the industry into seeking more innovative approaches to protecting and improving crop yield. As a result, Artificial Intelligence is steadily emerging as part of the industry’s technological evolution which help can help farmers get more from the land while using resources more sustainably, yielding healthier crops, control pests, monitor soil, help with workload, etc
*All the media belongs to the respective owners*
Contact me for further queries & discussions...
This document discusses the use of artificial intelligence in agriculture. It notes that the global population is expected to double by 2050, requiring a 70% increase in food production. AI can help address this challenge through automated farming activities, pest and disease monitoring, crop quality management, and machine vision systems. Examples provided include automated irrigation systems to save water, remote sensing for crop health monitoring, AI-based harvesting of vine crops, and early warning systems for pest outbreaks. Decision support systems using neural networks, genetic algorithms and other techniques can also help with yield prediction. Additional applications mentioned are driverless tractors, targeted weed removal robots, and AI-guided farming decisions. The document concludes that AI can optimize resource use and help solve labor
A PPT which gives a brief introduction on Machine Learning and on the products developed by using Machine Learning Algorithms in them. Gives the introduction by using content and also by using a few images in the slides as part of the explanation. It includes some examples of cool products like Google Cloud Platform, Cozmo (a tiny robot built by using Artificial Intelligence), IBM Watson and many more.
Machine learning and artificial intelligence are explained. Machine learning uses algorithms and past data to allow computers to optimize performance and develop behaviors without being explicitly programmed. It is a branch of artificial intelligence that uses supervised and unsupervised algorithms to apply past information to new data or draw conclusions from datasets. Case studies show how machine learning reveals influences and predicts user preferences. Artificial intelligence aims to simulate human intelligence through computer science, psychology, and other fields. Industries like healthcare and finance will benefit from machine learning and artificial intelligence applications like disease prediction and financial recommendations.
Machine learning and its applications was a gentle introduction to machine learning presented by Dr. Ganesh Neelakanta Iyer. The presentation covered an introduction to machine learning, different types of machine learning problems including classification, regression, and clustering. It also provided examples of applications of machine learning at companies like Facebook, Google, and McDonald's. The presentation concluded with discussing the general machine learning framework and steps involved in working with machine learning problems.
Machine learning involves developing systems that can learn from data and experience. The document discusses several machine learning techniques including decision tree learning, rule induction, case-based reasoning, supervised and unsupervised learning. It also covers representations, learners, critics and applications of machine learning such as improving search engines and developing intelligent tutoring systems.
AI bots in the agriculture field can harvest crops at a higher volume and faster pace than human laborers. By leveraging computer vision helps to monitor the weed and spray them. Thus, Artificial Intelligence is helping farmers find more efficient ways to protect their crops from weeds.
IOT has many applications in agriculture such as crop water management using soil moisture sensors, pest management using motion detecting PIR sensors, precision farming using sensors and drones, and livestock monitoring using sensors on wearables that track temperature, activity, and health indicators. IOT helps optimize resources like water, increase yields, and monitor livestock health remotely. However, challenges include infrastructure and connectivity issues, costs, and difficulties in implementation and data analysis for some farmers. Overall, IOT solutions have potential to increase agricultural sustainability and competitiveness.
Artificial intelligence : Basics and application in AgricultureAditi Chourasia
Agriculture is the mainstay of Indian economy as about 60% of our population depends directly or indirectly on agriculture.Exploration of technology in digital world gave birth to a whole new field of making intelligent machines i.e. Artificial intelligence (AI). AI is making a huge impact in all domains of the industry. Every industry looking to automate certain jobs through the use of intelligent machinery. Factors such as climate change, population growth and food security concerns have propelled the industry into seeking more innovative approaches to protecting and improving crop yield. As a result, AI is steadily emerging as part of the Agricultural industry’s technological evolution. The automation in agriculture is the main concern and the emerging subject across the world. AI in agriculture not only helping farmers to automate their farming but also shifts to precise cultivation for higher crop yield and better quality while using fewer resources.Technological advancement in the future will provide more useful applications to the sector helping the world deal with various farming challenges used to be faced in traditional agricultural practices.
This is a deep learning presentation based on Deep Neural Network. It reviews the deep learning concept, related works and specific application areas.It describes a use case scenario of deep learning and highlights the current trends and research issues of deep learning
We can predict soil moisture level and motion of predators.
Irrigation system can be monitored .
Damage caused by predators is reduced.
Increased productivity.
Water conservation.
Profit to farmers.
The document discusses the differences between machine learning (ML), statistical learning, data mining (DM), and automated learning (AL). It argues that while ML and statistical learning developed similar techniques starting in the 1960s, DM emerged in the 1990s from a merging of database research and automated learning. However, industry was much more enthusiastic about adopting DM techniques compared to AL techniques, even though many DM systems are just friendly interfaces of AL systems. The document aims to explain the key differences between DM and AL that led to DM's greater commercial success.
DALL-E 2 - OpenAI imagery automation first developed by Vishal Coodye in 2021...MITAILibrary
The document provides a review of machine learning interpretability methods. It begins with an introduction to explainable artificial intelligence and a discussion of key concepts like interpretability and explainability. It then presents a taxonomy of interpretability methods that are divided into four main categories: methods for explaining black-box models, creating white-box models, promoting fairness, and analyzing model sensitivity. Specific machine learning interpretability techniques are summarized within each category.
Artificial intelligence has the potential to help address challenges facing the agricultural sector as the global population increases. New technologies like drones, driverless tractors, automated irrigation, and machine learning are helping farmers monitor crops and soils, apply inputs precisely, and increase yields. Startups are developing tools using computer vision, satellites, and deep learning to diagnose plant health, predict weather, and optimize resource use. These AI solutions aim to help farmers "do more with less" and help feed the world's growing population in a sustainable way.
The document discusses the concept of Internet of Things (IoT) and its applications in agriculture. It defines IoT and describes how physical objects can be connected to collect and exchange data. Some key applications of IoT in agriculture mentioned include monitoring soil moisture and temperature for controlled irrigation, livestock monitoring, pest monitoring, and mobile money transfers. However, constraints for implementing IoT in Indian agriculture include small land holdings, connectivity and affordability issues. Some case studies on precision agriculture and reducing water usage through IoT are also summarized.
Internet of Things (IoT) is the internetworking of physical devices. This system has the ability to transfer data over a network. Mostly without requiring human intervention.Internet-connected to the physical world via ubiquitous sensors.
It is connecting each and everything to the internet.
Deep learning is a type of machine learning that uses multiple processing layers to learn representations of data with features that become more complex at each layer. Deep learning has achieved human-level performance in areas like image recognition by learning from large datasets. In healthcare, deep learning has been applied to tasks like detecting pneumonia from chest X-rays and skin cancer from images with accuracy comparable to doctors. However, challenges remain around data variability, uncertainty, class imbalance, and data annotation. Cross-area collaboration and data sharing are seen as key to realizing the potential of deep learning in healthcare.
The document discusses the benefits of exercise for mental health. Regular physical activity can help reduce anxiety and depression and improve mood and cognitive functioning. Exercise boosts blood flow, releases endorphins, and promotes changes in the brain which help regulate emotions and stress levels.
Remote Sensing (RS), UAV/drones, and Machine Learning (ML) as powerful techni...nitinrane33
Precision agriculture utilizes modern technology to optimize agricultural practices, resulting in increased productivity while reducing costs and environmental impact. The use of remote sensing (RS), drones or unmanned aerial vehicles (UAVs), and machine learning (ML) has significantly transformed precision agriculture. These advanced technologies provide farmers with accurate, cost-effective, and timely tools to manage crops and resources effectively. This paper evaluates the use of these techniques in precision agriculture, including their benefits, and effective applications. Remote sensing involves using satellites, aircraft, or drones to collect data on crops and the environment, such as soil moisture, temperature, and vegetation indices. With high-resolution images and three-dimensional maps of crops, UAVs enable farmers to identify and address issues like pest infestations or nutrient deficiencies. Machine learning algorithms analyze large amounts of data to predict crop yields, optimize irrigation and fertilization, and identify areas of the field that need attention. Several case studies highlight the effectiveness of these techniques in different agricultural settings. However, the paper also acknowledges the challenges associated with adopting these technologies, such as cost, data management, and regulatory issues. While the initial investment in drones and sensors may be high, the long-term benefits in terms of increased yields, reduced costs, and environmental sustainability are substantial. Farmers need to be trained in the use of these technologies to make informed decisions, and effective data management and analysis are crucial. Additionally, regulatory frameworks are still evolving, and clear guidelines are required for data privacy, safety, and ethical use. Although challenges remain, the benefits of increased productivity, reduced costs, and environmental sustainability make these technologies an attractive investment for farmers worldwide.
Artificial Intelligence In Agriculture & Its Status in IndiaJanhviTripathi
Worldwide, agriculture is a $5 trillion industry, and with the ever increasing population, the world will need to produce 50% more food by 2050 which cannot be accomplished with the percentage of land under cultivation. Factors such as climate change, population growth and food security concerns have propelled the industry into seeking more innovative approaches to protecting and improving crop yield. As a result, Artificial Intelligence is steadily emerging as part of the industry’s technological evolution which help can help farmers get more from the land while using resources more sustainably, yielding healthier crops, control pests, monitor soil, help with workload, etc
*All the media belongs to the respective owners*
Contact me for further queries & discussions...
This document discusses the use of artificial intelligence in agriculture. It notes that the global population is expected to double by 2050, requiring a 70% increase in food production. AI can help address this challenge through automated farming activities, pest and disease monitoring, crop quality management, and machine vision systems. Examples provided include automated irrigation systems to save water, remote sensing for crop health monitoring, AI-based harvesting of vine crops, and early warning systems for pest outbreaks. Decision support systems using neural networks, genetic algorithms and other techniques can also help with yield prediction. Additional applications mentioned are driverless tractors, targeted weed removal robots, and AI-guided farming decisions. The document concludes that AI can optimize resource use and help solve labor
A PPT which gives a brief introduction on Machine Learning and on the products developed by using Machine Learning Algorithms in them. Gives the introduction by using content and also by using a few images in the slides as part of the explanation. It includes some examples of cool products like Google Cloud Platform, Cozmo (a tiny robot built by using Artificial Intelligence), IBM Watson and many more.
Machine learning and artificial intelligence are explained. Machine learning uses algorithms and past data to allow computers to optimize performance and develop behaviors without being explicitly programmed. It is a branch of artificial intelligence that uses supervised and unsupervised algorithms to apply past information to new data or draw conclusions from datasets. Case studies show how machine learning reveals influences and predicts user preferences. Artificial intelligence aims to simulate human intelligence through computer science, psychology, and other fields. Industries like healthcare and finance will benefit from machine learning and artificial intelligence applications like disease prediction and financial recommendations.
Machine learning and its applications was a gentle introduction to machine learning presented by Dr. Ganesh Neelakanta Iyer. The presentation covered an introduction to machine learning, different types of machine learning problems including classification, regression, and clustering. It also provided examples of applications of machine learning at companies like Facebook, Google, and McDonald's. The presentation concluded with discussing the general machine learning framework and steps involved in working with machine learning problems.
Machine learning involves developing systems that can learn from data and experience. The document discusses several machine learning techniques including decision tree learning, rule induction, case-based reasoning, supervised and unsupervised learning. It also covers representations, learners, critics and applications of machine learning such as improving search engines and developing intelligent tutoring systems.
AI bots in the agriculture field can harvest crops at a higher volume and faster pace than human laborers. By leveraging computer vision helps to monitor the weed and spray them. Thus, Artificial Intelligence is helping farmers find more efficient ways to protect their crops from weeds.
IOT has many applications in agriculture such as crop water management using soil moisture sensors, pest management using motion detecting PIR sensors, precision farming using sensors and drones, and livestock monitoring using sensors on wearables that track temperature, activity, and health indicators. IOT helps optimize resources like water, increase yields, and monitor livestock health remotely. However, challenges include infrastructure and connectivity issues, costs, and difficulties in implementation and data analysis for some farmers. Overall, IOT solutions have potential to increase agricultural sustainability and competitiveness.
Artificial intelligence : Basics and application in AgricultureAditi Chourasia
Agriculture is the mainstay of Indian economy as about 60% of our population depends directly or indirectly on agriculture.Exploration of technology in digital world gave birth to a whole new field of making intelligent machines i.e. Artificial intelligence (AI). AI is making a huge impact in all domains of the industry. Every industry looking to automate certain jobs through the use of intelligent machinery. Factors such as climate change, population growth and food security concerns have propelled the industry into seeking more innovative approaches to protecting and improving crop yield. As a result, AI is steadily emerging as part of the Agricultural industry’s technological evolution. The automation in agriculture is the main concern and the emerging subject across the world. AI in agriculture not only helping farmers to automate their farming but also shifts to precise cultivation for higher crop yield and better quality while using fewer resources.Technological advancement in the future will provide more useful applications to the sector helping the world deal with various farming challenges used to be faced in traditional agricultural practices.
This is a deep learning presentation based on Deep Neural Network. It reviews the deep learning concept, related works and specific application areas.It describes a use case scenario of deep learning and highlights the current trends and research issues of deep learning
We can predict soil moisture level and motion of predators.
Irrigation system can be monitored .
Damage caused by predators is reduced.
Increased productivity.
Water conservation.
Profit to farmers.
The document discusses the differences between machine learning (ML), statistical learning, data mining (DM), and automated learning (AL). It argues that while ML and statistical learning developed similar techniques starting in the 1960s, DM emerged in the 1990s from a merging of database research and automated learning. However, industry was much more enthusiastic about adopting DM techniques compared to AL techniques, even though many DM systems are just friendly interfaces of AL systems. The document aims to explain the key differences between DM and AL that led to DM's greater commercial success.
DALL-E 2 - OpenAI imagery automation first developed by Vishal Coodye in 2021...MITAILibrary
The document provides a review of machine learning interpretability methods. It begins with an introduction to explainable artificial intelligence and a discussion of key concepts like interpretability and explainability. It then presents a taxonomy of interpretability methods that are divided into four main categories: methods for explaining black-box models, creating white-box models, promoting fairness, and analyzing model sensitivity. Specific machine learning interpretability techniques are summarized within each category.
The technology for building knowledge-based systems by inductive inference from examples has
been demonstrated successfully in several practical applications. This paper summarizes an approach to
synthesizing decision trees that has been used in a variety of systems, and it describes one such system,
ID3, in detail. Results from recent studies show ways in which the methodology can be modified to deal
with information that is noisy and/or incomplete. A reported shortcoming of the basic algorithm is
discussed and two means of overcoming it are compared. The paper concludes with illustrations of current
research directions.
International Journal of Engineering Research and Development (IJERD)IJERD Editor
journal publishing, how to publish research paper, Call For research paper, international journal, publishing a paper, IJERD, journal of science and technology, how to get a research paper published, publishing a paper, publishing of journal, publishing of research paper, reserach and review articles, IJERD Journal, How to publish your research paper, publish research paper, open access engineering journal, Engineering journal, Mathemetics journal, Physics journal, Chemistry journal, Computer Engineering, Computer Science journal, how to submit your paper, peer reviw journal, indexed journal, reserach and review articles, engineering journal, www.ijerd.com, research journals,
yahoo journals, bing journals, International Journal of Engineering Research and Development, google journals, hard copy of journal
The document provides an overview of cognitive tools for educational technologies, including their goals, functions, and examples. It discusses how cognitive tools can support learning by reducing cognitive load and helping students externalize and visualize knowledge. Key tools discussed include learning diaries, concept mapping, and simulations. The document emphasizes that cognitive tools should follow constructivist learning theories in providing self-guided, discovery-based learning.
Indexing based Genetic Programming Approach to Record Deduplicationidescitation
In this paper, we present a genetic programming (GP) approach to record
deduplication with indexing techniques.Data de-duplication is a process in which data are
cleaned from duplicate records due to misspelling, field swap or any other mistake or data
inconsistency. This process requires that we identify objects that are included in more than
one list.The problem of detecting and eliminating duplicated data is one of the major
problems in the broad area of data cleaning and data quality in data warehouse. So, we
need to create such a algorithm that can detect and eliminate maximum duplications.GP
with indexing is one of the optimization technique that helps to find maximum duplicates in
the database. We used adeduplication function that is able to identify whether two or more
entries in a repository are replicas or not. As many industries and systems depend on the
accuracy and reliability of databases to carry out operations. Therefore, the quality of the
information stored in the databases, can have significant cost implications to a system that
relies on information to function and conduct business. Moreover, this is fact that clean and
replica-free repositories not only allow the retrieval of higher quality information but also
lead to more concise data and to potential savings in computational time and resources to
process this data.
Index
This document discusses knowledge application, which is the final step in the knowledge management cycle where knowledge that has been captured and shared is put to actual use. It describes how user and task modeling can help promote effective knowledge application at the individual, group, and organizational levels. It also discusses knowledge management systems, knowledge reuse, and the strategic and practical implications of facilitating knowledge application within an organization.
This document describes formative research, a methodology for improving instructional design theories through developmental research or action research. Formative research involves creating an application or instance of a design theory, then formatively evaluating it to identify weaknesses in the application that could reflect weaknesses in the theory and improvements to the application that could reflect improvements to the theory. The key aspects of formative research discussed are: 1) using designed cases, in vivo naturalistic cases, or post facto naturalistic cases; 2) evaluating research based on effectiveness, efficiency, and appeal; and 3) following methodological procedures that involve designing an instance of a theory, collecting formative data, revising the instance, and proposing revisions to the theory.
Review: Semi-Supervised Learning Methods for Word Sense DisambiguationIOSR Journals
This document provides a review of semi-supervised learning methods for word sense disambiguation. It discusses how semi-supervised learning uses both labeled and unlabeled data, requiring only a small amount of labeled data. The document outlines several semi-supervised learning techniques for word sense disambiguation, including bootstrapping algorithms like Yarowsky's algorithm, and graph-based approaches like label propagation. It provides details on Yarowsky's bootstrapping algorithm and how it is able to generalize to label new examples through exploiting properties like one-sense-per-collocation and language redundancy.
1) The document discusses two paradoxes that arise when people learn to use computers: the production paradox and assimilation bias paradox.
2) The production paradox stems from users' strong motivation to be productive and complete tasks, which reduces their motivation to spend time learning new skills. This causes users to rely on familiar methods even if less efficient.
3) The assimilation bias paradox occurs when users apply existing knowledge to new situations, which can lead to incorrect understandings when old and new knowledge do not align well. These paradoxes mutually reinforce each other, exacerbating learning challenges.
Document contains some of the questions from the Domingos Paper. Overall idea is to understand what Machine Learning is all about. This paper helps us to understand the need of Machine Learning in our day to day lives. Well I you will find this document helpful.
This document provides an overview of machine learning. It defines learning and discusses different types of learning including rote, supervised, and unsupervised learning. It explains the need for machine learning to allow systems to learn on their own from data. Machine learning is described as a branch of AI that allows systems to learn from examples without being explicitly programmed. Various machine learning tasks and applications are mentioned like optical character recognition. Different machine learning techniques are then summarized, including learning through examples, explanation based learning, and learning by analogy.
The document discusses data mining and knowledge discovery in databases. It defines data mining as the nontrivial extraction of implicit and potentially useful information from large amounts of data. With huge increases in data collection and storage, data mining aims to analyze data and discover patterns that can provide insights and knowledge about businesses and the real world. The data mining process involves selecting, preprocessing, transforming, and analyzing data to extract hidden patterns and relationships, which are then interpreted and evaluated.
An Essay Concerning Human Understanding Of Genetic ProgrammingJennifer Roman
This document discusses the relationship between theory and practice in genetic programming. It argues that genetic programming practice is increasingly moving toward biology by borrowing mechanisms from biological systems like DNA dynamics, neural learning, and regulatory networks. As a result, genetic programming theory should also increasingly borrow from biology to remain relevant to advances in practice. Specific challenges for future theory are discussed, such as accounting for evolved code structure, architectural evolution, and developmental processes at genetic, morphological, and behavioral levels. The document suggests that genetic programming theory will need to consider mechanisms that regulate genetic stability and evolvability, as these are now understood to play important roles in biological systems and in genetic programming research.
USING ONTOLOGIES TO IMPROVE DOCUMENT CLASSIFICATION WITH TRANSDUCTIVE SUPPORT...IJDKP
Many applications of automatic document classification require learning accurately with little training
data. The semi-supervised classification technique uses labeled and unlabeled data for training. This
technique has shown to be effective in some cases; however, the use of unlabeled data is not always
beneficial.
On the other hand, the emergence of web technologies has originated the collaborative development of
ontologies. In this paper, we propose the use of ontologies in order to improve the accuracy and efficiency
of the semi-supervised document classification.
We used support vector machines, which is one of the most effective algorithms that have been studied for
text. Our algorithm enhances the performance of transductive support vector machines through the use of
ontologies. We report experimental results applying our algorithm to three different datasets. Our
experiments show an increment of accuracy of 4% on average and up to 20%, in comparison with the
traditional semi-supervised model.
McKinsey Global Institute Big data The next frontier for innova.docxandreecapon
McKinsey Global Institute
Big data: The next frontier for innovation, competition, and productivity 27
2. Bigdatatechniquesand technologies
A wide variety of techniques and technologies has been developed and adapted to aggregate, manipulate, analyze, and visualize big data. These techniques and technologies draw from several fields including statistics, computer science, applied mathematics, and economics. This means that an organization that intends to derive value from big data has to adopt a flexible, multidisciplinary approach. Some techniques and technologies were developed in a world with access to far smaller volumes and variety in data, but have been successfully adapted so that they are applicable to very large sets of more diverse data. Others have been developed more recently, specifically to capture value from big data. Some were developed by academics and others by companies, especially those with online business models predicated on analyzing big data.
This report concentrates on documenting the potential value that leveraging big data can create. It is not a detailed instruction manual on how to capture value, a task that requires highly specific customization to an organization’s context, strategy, and capabilities. However, we wanted to note some of the main techniques and technologies that can be applied to harness big data to clarify the way some
of the levers for the use of big data that we describe might work. These are not comprehensive lists—the story of big data is still being written; new methods and tools continue to be developed to solve new problems. To help interested readers find a particular technique or technology easily, we have arranged these lists alphabetically. Where we have used bold typefaces, we are illustrating the multiple interconnections between techniques and technologies. We also provide a brief selection of illustrative examples of visualization, a key tool for understanding very large-scale data and complex analyses in order to make better decisions.
TECHNIQUES FOR ANALYZING BIG DATA
There are many techniques that draw on disciplines such as statistics and computer science (particularly machine learning) that can be used to analyze datasets. In this section, we provide a list of some categories of techniques applicable across a range of industries. This list is by no means exhaustive. Indeed, researchers continue to develop new techniques and improve on existing ones, particularly in response to the need
to analyze new combinations of data. We note that not all of these techniques strictly require the use of big data—some of them can be applied effectively to smaller datasets (e.g., A/B testing, regression analysis). However, all of the techniques we list here can be applied to big data and, in general, larger and more diverse datasets can be used to generate more numerous and insightful results than smaller, less diverse ones.
A/B testing. A technique in which a control group is compa ...
Designing for Change: Mash-Up Personal Learning EnvironmentseLearning Papers
Authors:Fridolin Wild, Felix Mödritscher, Steinn E. Sigurdarson.
Institutions for formal education and most work places are equipped today with at least some kind of tools that bring together people and content artefacts in learning activities to support them in constructing and processing information and knowledge. For almost half a century, science and practice have been discussing models on how to bring personalisation through digital means to these environments.
The document discusses a study on user perceptions of machine learning. Researchers provided datasets to be analyzed by machine learning. Interviews found that machine learning produced results matching statistical analysis with less preparation time. Researchers saw potential for machine learning to help visualize data patterns but noted it may miss some relationships. There was interest in future machine learning use if researchers better understood the technique.
The past two decades has seen a dramatic increase in the amount of information or data being stored in electronic format. This accumulation of data has taken place at an explosive rate. It has been estimated that the amount of information in the world doubles every 20 months and the size and number of databases are increasing even faster. The increase in use of electronic data gathering devices such as point-of-sale or remote sensing devices has contributed to this explosion of available data. Figure 1 from the Red Brick company illustrates the data explosion.
Similar to Applying Machine Learning to Agricultural Data (20)
Este documento analiza el modelo de negocio de YouTube. Explica que YouTube y otros sitios de video online representan un nuevo modelo de negocio para contenidos audiovisuales debido al cambio en los hábitos de consumo causado por las nuevas tecnologías. Describe cómo YouTube aprovecha la participación de los usuarios para mejorar continuamente y atraer una audiencia diferente a la de los medios tradicionales.
The defense was successful in portraying Michael Jackson favorably to the jury in several ways:
1) They dressed Jackson in ornate costumes that conveyed images of purity, innocence, and humility.
2) Jackson was shown entering the courtroom as if on a red carpet, emphasizing his celebrity status.
3) Jackson appeared vulnerable, childlike, and in declining health during the trial, eliciting sympathy from jurors.
4) Defense attorney Tom Mesereau effectively presented a coherent narrative of Jackson as a victim and portrayed Neverland as a place of refuge, undermining the prosecution's arguments.
Michael Jackson was born in 1958 in Gary, Indiana and rose to fame in the 1960s as the lead singer of The Jackson 5, topping music charts in the 1970s. As a solo artist in the 1980s, his album Thriller broke music records. In the 1990s and 2000s, Jackson faced several legal issues related to child abuse allegations while continuing to release music. He married Lisa Marie Presley and Debbie Rowe and had two children before his death in 2009.
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...butest
This document appears to be a list of popular books from various authors. It includes over 150 book titles across many genres such as fiction, non-fiction, memoirs, and novels. The books cover a wide range of topics from politics to cooking to autobiographies.
The prosecution lost the Michael Jackson trial due to several key mistakes and weaknesses in their case:
1) The lead prosecutor, Thomas Sneddon, was too personally invested in the case against Jackson, having pursued him for over a decade without success.
2) Sneddon's opening statement was disorganized and weak, failing to effectively outline the prosecution's case.
3) The accuser's mother was not credible and damaged the prosecution's case through her erratic testimony, history of lies and con artist behavior.
4) Many prosecution witnesses were not credible due to prior lawsuits against Jackson, debts owed to him, or having been fired by him. Several witnesses even took the Fifth Amendment.
Here are three examples of public relations from around the world:
1. The UK government's "Be Clear on Cancer" campaign which aims to raise awareness of cancer symptoms and encourage early diagnosis.
2. Samsung's global brand marketing and sponsorship activities which aim to increase brand awareness and favorability of Samsung products worldwide.
3. The Brazilian government's efforts to improve its international image and relations with other countries through strategic communication and diplomacy.
The three most important functions of public relations are:
1. Media relations because the media is how most organizations reach their key audiences. Strong media relationships are crucial.
2. Writing, because written communication is at the core of public relations and how most information is
Michael Jackson Please Wait... provides biographical information about Michael Jackson including his birthdate, birthplace, parents, height, interests, idols, favorite foods, films, and more. It discusses his background, career highlights including influential albums like Thriller, and films he appeared in such as The Wiz and Moonwalker. The document contains photos and details about Jackson's life and illustrious music career.
The MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazzbutest
The document discusses the process of manufacturing celebrity and its negative byproducts. It argues that celebrities are rarely the best in their individual pursuits like singing, dancing, etc. but become famous due to being products of a system controlled by wealthy elites. This system stifles opportunities for worthy artists and creates feudalism. The document also asserts that manufactured celebrities should not be viewed as role models due to behaviors like drug abuse and narcissism that result from the celebrity-making process.
Michael Jackson was a child star who rose to fame with the Jackson 5 in the late 1960s and early 1970s. As a solo artist in the 1970s and 1980s, he had immense commercial success with albums like Off the Wall, Thriller, and Bad, which featured hit singles and groundbreaking music videos. However, his career and public image were plagued by controversies related to allegations of child sexual abuse in the 1990s and 2000s. He continued recording and performing but faced ongoing media scrutiny into his private life until his death in 2009.
Social Networks: Twitter Facebook SL - Slide 1butest
The document discusses using social networking tools like Twitter and Facebook in K-12 education. Twitter allows students and teachers to share short updates and can be used to give parents a window into classroom activities. Facebook allows targeted advertising that could be used to promote educational activities. Both tools could help facilitate communication between schools and communities if used properly while managing privacy and security concerns.
Facebook has over 300 million active users who log on daily, and allows brands to create public profile pages to interact with users. Pages are for brands and organizations only, while groups can be made by any user about any topic. Pages do not show admin names and have no limits on fans, while groups display admin names and are limited to 5,000 members. Content on pages should aim to provoke action from subscribers and establish a regular posting schedule using a conversational tone.
Executive Summary Hare Chevrolet is a General Motors dealership ...butest
Hare Chevrolet is a car dealership located in Noblesville, Indiana that has successfully used social media platforms like Twitter, Facebook, and YouTube to create a positive brand image. They invest significant time interacting directly with customers online to foster a sense of community rather than overtly advertising. As a result, Hare Chevrolet has built a large, engaged audience on social media and serves as a model for how brands can use online presences strategically.
Welcome to the Dougherty County Public Library's Facebook and ...butest
This document provides instructions for signing up for Facebook and Twitter accounts. It outlines the sign up process for both platforms, including filling out forms with name, email, password and other details. It describes how the platforms will then search for friends and suggest people to connect with. It also explains how to search for and follow the Dougherty County Public Library page on both Facebook and Twitter once signed up. The document concludes by thanking participants and providing a contact for any additional questions.
Paragon Software announces the release of Paragon NTFS for Mac OS X 8.0, which provides full read and write access to NTFS partitions on Macs. It is the fastest NTFS driver on the market, achieving speeds comparable to native Mac file systems. Paragon NTFS for Mac 8.0 fully supports the latest Mac OS X Snow Leopard operating system in 64-bit mode and allows easy transfer of files between Windows and Mac partitions without additional hardware or software.
This document provides compatibility information for Olympus digital products used with Macintosh OS X. It lists various digital cameras, photo printers, voice recorders, and accessories along with their connection type and any notes on compatibility. Some products require booting into OS 9.1 for software compatibility or do not support devices that need a serial port. Drivers and software are available for download from Olympus and other websites for many products to enable use with OS X.
To use printers managed by the university's Information Technology Services (ITS), students and faculty must install the ITS Remote Printing software on their Mac OS X computer. This allows them to add network printers, log in with their ITS account credentials, and print documents while being charged per page to funds in their pre-paid ITS account. The document provides step-by-step instructions for installing the software, adding a network printer, and printing to that printer from any internet connection on or off campus. It also explains the pay-in-advance printing payment system and how to check printing charges.
The document provides an overview of the Mac OS X user interface for beginners, including descriptions of the desktop, login screen, desktop elements like the dock and hard disk, and how to perform common tasks like opening files and folders. It also addresses frequently asked questions for Windows users switching to Mac OS X, such as where documents are stored, how to save or find documents, and what the equivalent of the C: drive is in Mac OS X. The document concludes with sections on file management tasks like creating and deleting folders, organizing files within applications, using Spotlight search, and an overview of the Dashboard feature.
This document provides a checklist for securing Mac OS X version 10.5, focusing on hardening the operating system, securing user accounts and administrator accounts, enabling file encryption and permissions, implementing intrusion detection, and maintaining password security. It describes the Unix infrastructure and security framework that Mac OS X is built on, leveraging open source software and following the Common Data Security Architecture model. The checklist can be used to audit a system or harden it against security threats.
This document summarizes a course on web design that was piloted in the summer of 2003. The course was a 3 credit course that met 4 times a week for lectures and labs. It covered topics such as XHTML, CSS, JavaScript, Photoshop, and building a basic website. 18 students from various majors enrolled. Student and instructor evaluations found the course to be very successful overall, though some improvements were suggested like ensuring proper software and pairing programming/non-programming students. The document also discusses implications of incorporating web design material into existing computer science curriculums.
1. 1
Applying Machine Learning to Agricultural Data
ROBERT J. McQUEEN
Management Systems, University of Waikato, Hamilton, New Zealand
email bmcqueen@waikato.ac.nz
STEPHEN R. GARNER
CRAIG G. NEVILL-MANNING
IAN H. WITTEN
Computer Science, University of Waikato, Hamilton, New Zealand
email srg1@waikato.ac.nz, cgn@waikato.ac.nz, ihw@waikato.ac.nz
Abstract
Many techniques have been developed for learning rules and relationships automatically from
diverse data sets, to simplify the often tedious and error-prone process of acquiring
knowledge from empirical data. While these techniques are plausible, theoretically well-
founded, and perform well on more or less artificial test data sets, they depend on their ability
to make sense of real-world data. This paper describes a project that is applying a range of
machine learning strategies to problems in agriculture and horticulture. We briefly survey
some of the techniques emerging from machine learning research, describe a software
workbench for experimenting with a variety of techniques on real-world data sets, and
describe a case study of dairy herd management in which culling rules were inferred from a
medium-sized database of herd information.
2. 2
Introduction
Machine learning is an emerging technology that can aid in the discovery of rules and patterns
in sets of data. It has frequently been observed that the volume of recorded data is growing at
an astonishing rate that far outstrips our ability to make sense of it, and the phrase “database
mining” is now being used to describe efforts to analyze data sets automatically for significant
structural regularities (Piatetsky–Shapiro & Frawley, 1991). Potential applications of these
techniques in domains such as agriculture and horticulture are legion. There are many
possible ways to capitalize on any patterns that are discovered. For example, their implicit
predictive ability could be embedded in automatic processes such as expert systems, or they
could be used directly for communication between human experts and for educational
purposes.
This paper explores what machine learning can do in the agricultural domain. We begin
with an overview of the technology, concentrating in particular on the more widely-applicable
“similarity-based” techniques. One of the practical problems in applying machine learning is
that it is hard to acquire a variety of learning tools and experiment with them in a uniform
way. We describe a software workbench, called WEKA, that collects together a number of
schemes and allows users to run them on real-world data sets and interpret and compare the
results. Next we show how the workbench can be applied to an agricultural problem: dairy
herd management. The aim is to infer the rules that are implicit in a particular farmer’s
strategy for culling less productive cows. These rules might be used, for example, to
communicate one farmer’s strategy to another, and are likely to be far more acceptable in
practice than a numeric “productivity index” such as is often used for this purpose. Several
unanticipated problems arose in the application of machine learning methods to the recorded
data. Once these problems were overcome, the results were encouraging, and indicate that
machine learning can play a useful role in large-scale agricultural problem solving.
Machine learning
As used in everyday language, “learning” is a very broad term that denotes the gaining of
3. 3
knowledge, skill and understanding from instruction, experience or reflection. For the
purposes of the present work, we take it in a much more specific sense to denote the
acquisition of structural descriptions from examples of what is being described. There are
numerous other words that could be used to mean much the same thing; indeed others have
defined terms such as “generalization” (Schank et al., 1986), “inductive learning” (Michalski,
1983), and “inductive modelling” (Angluin & Smith, 1983) in almost identical ways.
Moreover, what is learned—our “structural description”—is sometimes called a
“generalization,” a “description,” a “concept,” a “model,” an “hypothesis.” For present
purposes we regard these as equivalent, and simply use the term “concept” to denote the
structural description that the machine acquires.
“Learning” in this sense implies the acquisition of descriptions that make the structure of
generalizations explicit. This rules out a number of interesting software paradigms that
parallel the skill acquisition process in humans by learning how to do something without
encoding this knowledge in a form which is easy to interpret. One example is connectionist
models of learning, which embed knowledge in high-dimensional numerically-parameterized
spaces and thereby transform learning into a process of weight adjustment. Another example
is genetic algorithms, which emulate an evolutionary form of adaptation by mutation and
natural selection. A third example is adaptive text compression, which creates a model of
incoming text and uses it to predict upcoming characters. The reason that we are prepared to
rule out such schemes is that we envisage that in the application domain being considered, the
acquired knowledge will frequently be used for purposes of communication, and the implicit
descriptions that are used in these schemes cannot be communicated between people nor
between machines having different architectures.
METHODS OF MACHINE LEARNING
The last decade has seen such an explosion of methods for machine learning that it is difficult
to classify them into a small set of main approaches. It is more useful is to examine several
dimensions along which they can be compared. Although these dimensions tend to be
overlapping rather than orthogonal, they do provide a useful framework for examining
4. 4
machine learning schemes (Witten et al. 1988, MacDonald et al., 1989.)
Similarity-based versus knowledge-based. Many learning methods use only the observed
similarities and differences between examples in order to form generalizations; we refer to
these as similarity based. Similarity-based learning analyses data more or less syntactically,
with little use of semantics. A few examples of such schemes can be found in Winston
(1972), Michalski (1980), and Lebowitz (1986). In contrast, knowledge based methods use
prior knowledge—often called “background” knowledge—in the form of a “domain theory”
that guides the interpretation of new examples. If the domain theory is complete, of course,
there is no new knowledge to learn: the theory already contains a full prescription for
interpreting, or “explaining,” all the examples that will be encountered. However, it may still
be possible to learn new and more efficient ways of employing that theory to interpret
examples; this is often called “explanation-based learning” because it focuses on the
explanations that the theory is capable of generating for each example (Mitchell, et al. 1986,
DeJon et al., 1986). Some learning methods relax the requirement of a fully comprehensive
domain theory by assuming an incomplete domain theory and augmenting it by processing
new examples and incorporating them into the theory, either to correct erroneous parts or to
add new rules to the theory.
Noise-tolerant versus exact. Some machine learning schemes are robust and tolerant to noise
in the examples presented, whereas others are designed to work with exact information.
Generally speaking, knowledge-based schemes, like other knowledge-rich methods in AI,
tend to be brittle and break down if the input contains errors. Similarity-based methods
generally use much larger numbers of examples and therefore have an opportunity to average
out some effects of noise. This distinction is closely allied to another one: one-shot versus
multi-example learning schemes. Some methods operate by analyzing a single example
intensively, while with others, a large collection of examples is processed together. Clearly,
statistical schemes fall into the second class.
Top-down versus bottom-up. Top-down machine learning methods delineate the space of
concept descriptions in advance and search it for concepts that best characterize the structural
5. 5
similarities and/or difference between the examples that are presented. Some top-down
methods are “exact” in that they guarantee to produce just that set of concepts which are
consistent with the examples (and of course this only makes sense in the case of noise-free
examples), whereas others are heuristic and come up with a “good” concept but not
necessarily the best one. Bottom-up schemes begin by analyzing the individual examples and
building up structures from them. They are sometimes called “case-based” because they focus
on the individual cases. Some bottom-up schemes are one-shot in that they examine single
examples intensively, perhaps interacting with the user to elicit an explanation of any unusual
features that they exhibit. Others are multi-example—for example, nearest-neighbor schemes
that classify new or “unknown” examples on the basis of that old or “known” one that is
closest in some high-dimensional space.
Supervised versus unsupervised. Supervised learning has come to mean learning from a
training set of examples whose desired output patterns are provided, having been assigned by
some expert or “teacher.” It does not imply that the learning process is subject to direct
supervision (that is the purpose of the interactive versus non-interactive distinction below);
indeed, supervised learning often processes a set of examples in batch mode. In contrast,
unsupervised learning is where a set of examples is supplied but there is no indication of the
classes that they belong to (Cheeseman et al, 1988, Fisher, 1987). In this situation, the
learning scheme is expected to analyze the similarities and differences between the examples
and come up with a clustering that, in effect, assigns classes to them. The clustering may be
performed on the basis of either numeric or non-numeric properties, and, perhaps,
background knowledge.
Interactive versus non-interactive. Some learning methods are interactive in that they require a
teacher to monitor the progress of learning, whereas others are not and proceed
autonomously. In some respects, the requirement for a teacher can substitute for the lack of
an adequate domain theory, for the teacher can be consulted to “explain” the situation
whenever a new example fails to fit the system’s expectations (Bareiss et al., 1988). The key
problem here is to implement a dialog between system and teacher that allows information to
be articulated by either party at the appropriate level, and understood by the other. A different
6. 6
possible role for a teacher, which is much easier to arrange, is to have them select the
ordering of examples and choose ones that are most helpful to the learner in its present state
of knowledge. This simplifies communication considerably, but it does require the teacher to
understand the inner workings of the machine learning scheme to some extent.
Single- versus multi-paradigm. A final distinction can be made between single-paradigm
learners and multi-paradigm ones. Because of the various strengths and weaknesses of
current machine learning schemes, there is currently a great deal of interest in combining
learning mechanisms that adopt several approaches (e.g. Pazzani et al., 1992). For example,
similarity-based learning may be used to correct or complete a partial domain theory, or a
judicious combination of bottom-up and top-down learning may outperform either on its
own.
CHARACTERIZING THE PROBLEM
The most important feature of a problem domain, as far as the application of machine learning
is concerned, is the form that the data takes. Most learning techniques that have actually been
applied assume that the data are presented in a simple attribute-value format in which a record
has a fixed number of constant-valued fields or properties. Figure 1a illustrates different
kinds of data types; nominal attributes, which are drawn from a set with no further structure;
linear attributes, which are totally ordered; and tree-structured attributes, which form a
hierarchy or partial order. Figures 1b and 1c show a sample object (or “entity”), and a sample
concept (that in fact subsumes the object), expressed as a vector of generalized attributes.
Attribute vectors cannot describe situations that involve relations between objects. In
actuality, of course, databases are generally expressed as a set of relations, with several
records for a single entity and fields that reference other records or relations. Relations can be
described by functions which, like attributes, may be nominal, linear, or tree-structured.
Some researchers in machine learning are shifting their attention from algorithms that operate
in attribute-value domains to ones designed for more structured relational domains, for
example the field of inductive logic programming (ILP), which seeks to express concepts in a
language such as Prolog, and to infer these programs from data. One system which
7. 7
implements some aspects of ILP is the First Order Inductive Learner (FOIL), described in
Quinlan, (1990).
Another important feature of a problem domain is the quality of the data available. Most
“real” data is imperfect: incomplete (missing values for some attributes and objects), irrelevant
(some fields that do not relate to the problem at hand), redundant (involving unknown, or at
least unexpressed, relations between the attributes), noisy (for example, some attributes have
inherent measurement errors) and occasionally erroneous (e.g. incorrectly transcribed).
Methods of machine learning need to be robust enough to cope with imperfect data and to
discover laws in it that may not always hold but are useful for the problem at hand. The seven
levels of quality shown in Table 1 can be distinguished in a data set (Gaines, 1991). The aim
of a learning system is to discover a set of decision rules that is complete, in that it describes
all of the data; correct, predicting the data accurately; and minimal, i.e. with no redundancy
(level 1), given information at one of the other levels.
Another feature that strongly influences machine learning is whether or not operation
needs to be incremental. In many situations, new examples appear continually and it is
essential that the system can modify what it has already learned in the light of new
information. Learning is often exceedingly search-intensive and it is generally infeasible to
reprocess all examples whenever a new one is encountered.
EXPERT SYSTEMS AND STATISTICS
Often it is assumed that machine learning is proposed as a replacement for expert systems or
statistical methods such as clustering. In reality, the role of learning is to complement both
areas, increasing the set of tools available to the practitioner of either discipline.
In the case of expert systems, machine learning can be applied to the areas of knowledge
acquisition and maintenance. In the creation of a typical expert system, a person with detailed
knowledge about the problem domain and its solution supplies information that is converted
into rule sets to be incorporated into the knowledge base. This process requires that the expert
be able to articulate his or her knowledge clearly and effectively. In some cases, however,
there may be no expert available to supply the rules for the problem. When this happens, the
8. 8
extraction of rules and relationships from the domain may be undertaken using a machine
learning scheme. In the case of supervised machine learning, an expert may supply specially
selected cases or examples for a machine learning scheme and let it generate a model to
explain these examples. This may prove faster and more accurate than having to state each
rule explicitly. The process is incremental in the sense that as each example or case is seen,
the model is adapted to incorporate the new concept. This allows existing rule sets to be easily
updated over time if assumptions about the domain change—as happens very frequently in
practice.
In statistics, as in machine learning, patterns such as trends or clusters are often being
sought in the data. Much of conventional statistics is restricted to continuous, numeric data,
and seeks to test relationships that have been hypothesized in advance. Many machine
learning schemes can work with either symbolic or numeric data, or a combination of both,
and attempt to discover relationships in the data that have not yet been hypothesized. Once a
relationship has been discovered, further statistical analysis can be performed to confirm its
significance. Sometimes, both fields work independently towards the same goal, as in the
case of ID3 (Quinlan, 1986), a machine learning scheme, and CART (Breiman et al, 1984),
standing for “classification and regression trees,” a statistical scheme. These methods both
induce decision trees using essentially the same technique. Machine learning researchers also
incorporate statistics into learning schemes directly, as in the case of the Bayesian
classification system AUTOCLASS (Cheeseman et al, 1988).
AQ11: AN EARLY EXAMPLE OF AN AGRICULTURAL APPLICATION
An often quoted example of the application of machine learning in agriculture is the use of
the AQ11 program to identify rules for diagnosis of soybean diseases. In this early application
the similarity-based learning program AQ 11 was used to analyze data from over 600
questionnaires describing diseased plants (Michalski & Chilausky, 1980). Each plant was
assigned to one of 17 disease categories by an expert collaberator, who used a variety of
measurements describing the condition of the plant. Figure 2a shows a sample record with
values of some of the attributes given in italics.
9. 9
The diagnostic rule of Figure 2b for Rhizoctonia root rot was generated by AQ11, along
with a rule for every other disease category, from a set of training instances which were
carefully selected from the corpus of cases as being quite different from each other—“far
apart” in the instance space. At the same time, the plant pathologist who had produced the
diagnoses was interviewed and his expertise was translated into diagnostic rules using the
standard knowledge-engineering approach. Surprisingly, the computer-generated rules
outperformed the expert-derived rules on the remaining test instances—they gave the correct
disease top ranking just over 97% of the time, compared to just under 72% for the expert-
derived rules (Michalski & Chilausky, 1980). Furthermore, according to Quinlan (in
foreword, Piatetsky–Shapiro & Frawley, 1991), not only did AQ 11 find rules that
outperformed those of the expert collaborator, but the same expert was so impressed that he
adopted the discovered rules in place of his own.
The machine learning workbench
Given the proliferation of machine learning techniques, the task facing a scientist wishing to
apply machine learning to a problem in their own field is immense. Each technique is suitable
for particular kinds of problems, and has particular strengths and weaknesses. The
experimental status of machine learning means that it is impossible to offer one technique as a
general solution, so the key to applying machine learning widely is to simplify access to a
range of techniques.
To this end, the machine learning research group at the University of Waikato has
constructed a software ‘workbench’ to allow users to access a variety of machine learning
techniques for the purposes of experimentation and comparison using real world data sets.
The Waikato Environment for Knowledge Analysis (WEKA 1 ) currently runs on Sun
workstations under X-windows, with machine learning tools written in a variety of
programming languages (C , C ++ and LISP). The workbench is not a single program, but
rather a set of tools bound together by a common user interface.
The WEKA workbench differs from other machine learning environments in that its target
1 The weka is a cheeky, inquisitive native New Zealand bird, about the size of a chicken.
10. 10
user is a domain expert, in this case an agricultural scientist, who wants to apply machine
learning to real world data sets. Other systems such as the MLC++ project at Stanford
University (Kohavi et al,1994), and the European Machine Learning Toolbox project
(Kodratoff et al, 1992) are intended for use by machine learning researchers and
programmers developing and evaluating machine learning schemes, while the Emerald system
(Kaufman et al, 1993) is designed as an educational tool. The WEKA workbench is flexible
enough to be used as in a machine learning research role, and has also been used successfully
in undergraduate courses teaching machine learning. It is important to stress that WEKA is not
a multi-paradigm learner; rather than combining machine learning techniques to produce new
hybrid schemes, it concentrates on simplifying access to the schemes, so that their
performance can be evaluated on their own.
W EKA currently includes seven different machine learning schemes, summarized in
Table 2. In a typical session, a user might select a data set, run several different learning
schemes on it, exclude and include different sets of attributes, and make comparisons
between the resulting concepts. Output from each scheme can be viewed in an appropriate
form, for example as text, a tree or a graph. To allow users to concentrate on experimentation
and interpretation of the results, they are protected from the implementation details of the
machine learning algorithms and the input and output formats that the algorithms use.
The WEKA user interface is implemented using TK/TCL (Ousterhout, 1994), providing
portability and rapid prototyping. The main panel of the workbench is shown in Figure 3. On
the left is the file name and other information about the current data set. The next column
shows a list of the attributes in the data set, along with information about the currently-
selected one. The checkboxes indicate whether or not the attribute will be passed to the
learning scheme, while the diamond indicates which attribute to classify on when using a
supervised learning scheme. In the third column, the values that this attribute can take are
listed. If a particular value is selected, rules will be formed to differentiate tuples with this
value from the others; otherwise, classification rules are generated for each value. This degree
of control is useful for weeding out unused data items. The fourth column lists the available
machine learning schemes. Pressing a button marked ‘?’ displays a short description of the
11. 11
scheme. In the rightmost column, the user can control the way that the data is viewed and
manipulated.
We now briefly discuss the machine learning schemes that are included in the workbench
(Table 2). The first two are for unsupervised learning, or clustering. These are useful for
exploratory purposes when patterns in the data are being sought but it is not clear in advance
what they will be. For example, we have applied clustering to data on human patients with
diabetes symptoms, and discovered that the cases fall naturally into three classes which turn
out to have clinical implications (Monk et al., 1994). As mentioned earlier, AUTOCLASS
discovers classes in a database using a Bayesian statistical technique, which has several
advantages over other methods (Cheeseman et al., 1988). The number of classes is
determined automatically; examples are assigned with a probability to each class rather than
absolutely to a single class; and the example data can be real or discrete. CLASSWEB is a
reimplementation of an earlier system called COBWEB (Fisher, 1987), and also operates on a
mixture of numeric and non-numeric information, although it assigns each example to one
and only one class. Its evaluation criterion is psychologically rather than statistically
motivated, and its chief advantage over AUTOCLASS is that it consumes far fewer resources—
both memory space and execution time.
The other schemes in the workbench are for supervised learning. C4.5 performs top-
down induction of decision trees from a set of examples which have each been given a
classification (Quinlan, 1992). Typically, a training set will be specified by the user. The root
of the tree specifies an attribute to be selected and tested first, and the subordinate nodes
dictate tests on further attributes. The leaves are marked to show the classification of the
object they represent. An information-theoretic heuristic is used to determine which attribute
should be tested at each node, and the attribute that minimizes the entropy of the decision is
chosen. C4.5 is a well-developed piece of software that derives from the earlier ID3 scheme
(Quinlan, 1986), which itself evolved through several versions. OC1, another scheme in the
workbench, also induces decision trees top-down, but each node classifies examples by
testing linear combinations of features instead of a single feature (Murthy et al., 1993).
Although restricted to numeric data, this method consistently finds much smaller trees than
12. 12
comparable methods that use univariate trees.
CNF, DNF, PRISM, and INDUCT all represent the concepts they induce in the form of rules
rather than decision trees. It is easy to convert a decision tree to a set of rules, but more
economical descriptions with smaller numbers of rules and fewer terms in each can usually be
found by seeking rules directly. CNF and DNF are simple algorithms for creating rules that
take the form of conjunctions (the terms in a rule are ANDed together) and disjunctions (terms
OR ed together) respectively. Interesting and surprising results have been reported on
differences between these two seemingly very similar concept representations (Mooney,
1992). PRISM uses a top-down approach, like that of C4.5, for rule rather than decision tree
induction (Cendrowska, 1987); and INDUCT is an improved version that is probabilistically
based and copes with situations that demand non-deterministic rules (Gaines, 1991).
FOIL, for “first-order inductive learner” (Quinlan, 1990), induces logical definitions,
expressed as Horn clauses, from data presented in the form of relations. It begins with a set
of relations, each defined as a set of related values. Given a particular “target” relation, it
attempts to find clauses that define that relation in terms of itself and other relations. This
approach leads to more general, functional definitions that might be applied to new objects.
FOIL, like C4.5, uses a information-theoretic heuristic to guide the search for simple, general
clauses.
The schemes that constitute the current version of the WEKA workbench are not claimed to
be a representative selection of machine learning programs. They are all similarity-based
rather than knowledge-based;. This partly reflects the difficulty of finding a uniform way to
represent background knowledge, but is mainly due to the fact that domain theories are few
and far between in agricultural applications. Not surprisingly, all the schemes included are
noise-tolerant. There are no bottom-up schemes; this is a deficiency that we plan to rectify
shortly. Neither are there any interactive schemes, although the normal mode of operation of
the workbench is fairly interactive, involving as it does manual selection of pertinent attributes
and the synthesis of new ones (see below). We are considering including a fully-interactive
learning scheme; again, the problem is representation of knowledge—and its communication
13. 13
in a form that makes sense to a non-specialist in computer science.
Case study: dairy herd culling
New Zealand’s economic base has historically been agricultural, and while this emphasis has
decreased in recent decades, agriculture is still vitally important to the country’s wealth. Dairy
farming is in turn a large part of the agricultural sector, and the Livestock Improvement
Corporation, a subsidiary of the New Zealand Dairy Board, is an organization whose
mandate is to improve the genetics of New Zealand dairy cows.
THE LIVESTOCK DATABASE
The Corporation operates a large relational database system to track genetic history and
production records of 12 million dairy cows and sires, of which 3 million are currently alive.
Production data are recorded for each cow from four to twelve times per year, and additional
data are recorded as events occur. Farmers in turn receive information from the Livestock
Improvement Corporation in the form of reports from which comparisons within the herd can
be made. Two types of information that are produced are the production and breeding indexes
(PI and BI respectively), which indicate the merit of the animal. The former reflects the milk
produced by the animal with respect to measures such as milk fat, protein and volume,
indicating its merit as a production animal. The latter reflects the likely merit of a cow’s
progeny, indicating its worth as a breeding animal. In a well-managed herd, averages of these
indexes will typically increase every year, as superior animals enter the herd and low-index
ones are removed.
One major decision that farmers must make each year is whether to retain a cow in the
herd or remove it, usually to an abattoir. About 20% of the cows in a typical New Zealand
dairy herd are culled each year, usually near the end of the milking season as feed reserves
run short. The cows’ breeding and production indexes influence this decision, particularly
when compared with the other animals in the herd. Other factors which may influence the
decision are:
• age: a cow is nearing the end of its productive life at 8–10 years;
14. 14
• health problems;
• history of difficult calving;
• undesirable temperament traits (kicking, jumping fences);
• not being in calf for the following season.
The Livestock Improvement Corporation hoped that the machine learning project
investigation of their data might provide insight into the rules that farmers actually use to
make their culling decisions, enabling the corporation to provide better information to farmers
in the future. They provided data from ten herds, over six years, representing 19 000
records, each containing 705 attributes. These attributes are summarized in Table 3.
INITIAL DATA STRUCTURING
The machine learning tools used for the analysis were primarily C4.5 (Quinlan, 1992) and
FOIL (Quinlan, 1990). The initial raw data set as received from the Livestock Improvement
Corporation was run through C4.5 on the workbench. Classification was done on the fate
code attribute, which can take the values sold, dead, lost and unknown. The resulting tree,
shown in Figure 4, proved disappointing.
At the root of the tree is the transfer out date attribute. This implies that the culling
decision for a particular cow is based mainly on the date on which it is culled, rather than on
any attributes of the cow. Next, the date of birth is used, but as the culling decisions take
place in different years, an absolute date is not particularly meaningful. The cows age would
be useful, but is not explicitly present in the data set. The cause of fate attribute is strongly
associated with the fate code; it contains a coded explanation of the reason for culling. This
attribute is assigned a value after the culling decision is made, so it is not available to the
farmer when making the culling decision. Furthermore, we would like to be able to predict
this attribute—in particular the low production value—rather than include it in the tree as a
decision indicator. The presence of this attribute made the classification accuracy artificially
high, predicting the culling decision correctly 95% of the time on test data. Mating date is
another absolute date attribute, and animal key is simply a 7-digit identifier.
The problems with this decision tree stem from the denormalization of the database used
15. 15
to produce the input, and the representation of particular attributes. The solutions to these
problems are discussed below.
The effects of denormalization
Most machine learning techniques expect as input a set of tuples, analogous to one relation in
a database. Real databases, however, invariably contain more than one relation. The relational
operator join takes several relations and produces a single one from them, but this
denormalizes the database, introducing duplication and dependencies between attributes.
Dependencies in the data are quickly discovered by machine learning techniques, producing
trivial rules that relate two attributes. It is therefore necessary to modify the data or the scheme
to ignore these dependencies before interesting relationships can be discovered. In the project
described here, trivial relationships (such as between the fate code and cause of fate attributes)
are removed after inspecting decision trees by omitting one of the attributes from
consideration.
In this particular data set, a more serious problem stemmed from the joining of data from
several seasons. Each cow has particular attributes that remain constant throughout its
lifetime, for example animal key and date of birth. Other data, such as the number of weeks
of lactation, are recorded on a seasonal basis. In addition to this, monthly tests generate
production data, and movements from herd to herd are recorded at various times as they
occur. This means that data from several different years, months, and transfers were included
in the original record which was nominally for one year; data that should ideally be
considered separately (see Table 3).
Although culling decisions can occur at any point in the lactation season, the basic
decision to retain or remove an animal from the herd may be considered, for the purposes of
this investigation, to be made on an annual basis. Annual records should contain only
information about that year, and perhaps previous years, but not “foresight” information on
subsequent data or events as may have been included through the original extract from the
database. The dataset was renormalized into yearly records, taking care that “foresight”
information was excluded. Where no movement information (which included culling
16. 16
information) was recorded for a particular year, a retain decision replaces the missing value.
Monthly information was replaced by a yearly summary. While the data set was not fully
normalized (dependencies between animal key and date of birth still existed, for example), it
was normalized sufficiently for this particular application.
Attribute representation
The absolute dates included in the original data are not particularly useful. Once the database
is normalized into yearly records, these dates can be expressed relative to the year that the
record represents. In general, the accuracy of these dates need only be to the nearest year,
reducing the partitioning process evident in Figure 4.
In a discussion with staff from the Livestock Improvement Corporation, it was suggested
that a culling decision may not be based on a cow’s absolute performance, but on its
performance relative to the rest of the herd. To test this hypothesis, attributes were added to
the database representing the difference in production from the average production over the
cow’s herd. In order to prevent overly biasing the learning process, all the original attributes
were retained in the data set, and derived attributes added to the records were not
distinguished in any way. It was left to the machine learning schemes to decide if they were
more helpful for classification than the original attributes. Throughout this process, meetings
were held with staff at the Livestock Improvement Corporation. Discussions would often
result in the proposal of more derived attributes, and the clarification of the meaning of
particular attributes. Staff were also able to evaluate the plausibility of rules, which was
helpful in the early stages when recreating existing knowledge was a useful measure of the
correctness of our approach.
An obvious step would be to automate the production of derived attributes, to speed up
preprocessing and avoid human bias. However, the space of candidates is extremely large,
given the number of operations that can be performed on pairs of attributes. Typing the
attributes, and defining the operations which are meaningful on each type, would reduce the
space of possible derived attributes. For example, if the absolute dates in the original data are
defined as dates, and subtraction defined to be the only useful operator on dates, then the
17. 17
number of derived attributes would be considerably reduced, and useful attributes such as age
would still be produced. This is an interesting and challenging problem for investigation in
the future.
SUBSEQUENT C4.5 RUNS WITH MODIFIED DATA
After normalizing the data and adding derived attributes, C4.5 produced the tree in Figure 5.
Here, the fate code, cause of fate and transfer out date attributes have been transformed into a
status code which can take the values culled or retained. For a particular year, if a cow has
already been culled in a past season, or if it has not yet been born, the record is removed. If
the cow is alive in the given year, and is not transferred in that year, then it is marked as
retained. If it is transferred in that year, then it is marked culled. If, however, it died of
disease or some other factor outside the farmer’s control, the record is removed. After all, the
aim of this exercise is to discover the farmer’s culling rules rather than the incidence of
disease and injury.
The tree in Figure 5 is much more compact than the full tree shown in Figure 4. It was
produced with 30% of the instances, and correctly classifies 95% of the remaining instances.
The unconditional retention of cows two years or younger is due to the fact that they have not
begun lactation, and no measurements of their productive potential have yet been made. The
next decision is based on the cow’s worth as a breeding animal, which is calculated from the
earnings of the cow’s offspring. The volume of milk that the cow produces is used for the
final decision. The decisions in this tree are plausible from a farming perspective, and the
compactness and correctness of the tree indicate that it is a good explanation of the culling
decision. It is interesting to note that the tree consists entirely of derived attributes, further
emphasizing the importance of the preprocessing step.
Conclusions and future directions
From the work completed on the Livestock Improvement Corporation data it was possible
to isolate three steps that are necessary for the extraction of rules from a database.
The first step is extracting the data from its original form (in this case relational) to a two-
18. 18
dimensional flat-file form suitable for processing by the machine learning programs in the
workbench. This step is basically mechanical, and in principle could be achieved by simple
join operations on the database tables. In practice, however, the quality of the data—in
particular the large number of missing values—makes this non-trivial. This step also involved
the creation of the transformed attributes (e.g. age instead of birth date) and combined
attributes (e.g. not left herd and positive milk test = retained in herd).
The second step involves gaining insight about the problem domain from the extracted
and transformed dataset. It was helpful during this step to try initial runs of the dataset
through the machine learning tools. Some of these runs identified attributes, such as cow
identification number, which clearly have nothing to do with the culling decision. Through
questions directed at the domain experts, a greater understanding of the meaning of these
attributes was obtained, resulting in a better selection of the attributes to be used.
The third step was the use of machine learning tools to generate rules. Once a reasonably
well-structured dataset had been prepared in the standard file format, it was an easy matter to
process the dataset through the different algorithms available in the WEKA workbench, and
compare the resulting rule sets. The rules which resulted were referred to domain experts, and
the feedback used to iterate through all three steps to obtain new results.
In each of these three steps, domain expertise was essential to complement the data
transformation and machine learning processing skills required to prepare and process the
data sets.
Is there a likely future scenario of an automatic and unattended machine learning algorithm
being turned loose for background overnight processing against large relational databases?
We think the potential for this kind of intelligent agent may be there, but that there is much
that has to be done in the interim to create new algorithms and processing techniques that can
discover meaning in large and complex relational data structures, rather than the small and
simple two-dimensional attribute tables that have been used in the tests of machine learning
that are reported in the literature.
In the shorter term, there may be a high payback in making machine learning techniques
19. 19
easily usable by domain experts, who have intimate understanding of the nature of their data
and its relationships. Current developments include an attribute editor to simplify the process
of deriving new attributes from the data. This interactive tool will provide functions to
compute new attributes based on combinations of attributes, as well as inter-record
calculations such as rates of change in time-series data.
Overall, WEKA is fulfilling its role of bringing the potential of machine learning from the
computer science laboratory into the hands of experts in diverse domains.
Acknowledgements
This work is supported by the New Zealand Foundation for Research, Science and
Technology. We gratefully acknowledge the helpful cooperation of the Livestock
Improvement Centre, the work of Rhys Dewar and Donna Neal, and the stimulating research
environment provided by the Waikato machine learning group.
References
Angluin, D. and Smith, C.H., 1983. Inductive inference: theory and methods. Computing
Surveys, 15: 237–269.
Bareiss, A.R. Porter, B.W. and Weir, C.C. 1988, PROTOS: an exemplar-based learning
apprentice. Int. J. Man Machine Studies. 29(5), 549-561
Breiman, L., and Friedman, J.H., Olshen, R., and Stone C., 1984. Classification and
regression trees. Wadsworth International Group; Belmont, California.
Cameron-Jones, R.M., and Quinlan, J.R., 1993. Avoiding Pitfalls When Learning Recursive
Theories Proc. IJCAI 93, Morgan Kaufmann: 1050-1055
Cendrowska, J., 1987. PRISM: an algorithm for inducing modular rules. Int J Man-Machine
Studies, 27: 349–370.
Cheeseman, P., Kelly, J., Self, M., Stutz, J., Taylor, W., and Freeman, D., 1988.
AUTOCLASS: A Bayesian classification system. In: Laird, J. (Editor), Proc. of the Fifth
International Conference on Machine Learning. Ann Arbor, MI: Morgan Kaufmann: 54-
20. 20
64
DeJong, G. and Mooney, R. 1986. Explanation-based learning: an alternative view. Machine
Learning 1(2): 145-176
Fisher, D., 1987. Knowledge Acquisition Via Incremental Conceptual Clustering. Machine
Learning, 2: 139–172.
Gaines, B.R., 1991. The trade-off between knowledge and data in knowledge acquisition. In
Piatetsky–Shapiro and Frawley, 1991: 491–505.
Gennari, J.H., 1989. A survey of clustering methods., Technical Report 89-38, Irvine:
University of California, Dept. of Information and Computer Science.
Haussler, D., 1987. Learning conjunctive concepts in structural domains. Proc. AAAI: 466–
470.
Kaufman, K.A., and Michalski, R.S., 1993. EMERALD 2: An Integrated System of
Machine Learning and Discovery Programs to Support Education and Experimental
Research. Technical Report MLI 93-10, George Mason University.
Kodratoff, Y., Sleeman, D., Uszynski, M., Causse, K., and Craw, S., 1992. Building a
Machine Learning Toolbox. In: Steels, L. and Lepape, B. (Editors), Enhancing the
Knowledge Engineering Process, Elsevier Science Publishers B.V.: 81-108.
Kohavi, R., George, J., Long, R., Manley, D., and Pfleger, K., 1994. MLC++: A Machine
Learning Library in C++. Technical Report, Computer Science Department, Stanford
University.
Lebowitz, M. 1986. Concept learning in a rich input domain: Generalization-Based Memory.
In: R.S. Michalski, J.G. Carbonell and T.M. Mitchell (Editors), Machine Learning: An
Artificial Intelligence Approach, Volume II, Morgan Kaufmann, Los Altos, CA. 193-
214.
MacDonald, B.A. and Witten, I.H. 1989, A framework for knowledge acquisition through
techniques of concept learning. IEEE Trans Systems, Man and Cybernetics 19(3), 499-
512.
21. 21
Michalski, R.S. “Pattern recognition as rule-guided inductive inference.” 1980, IEEE
Transactions on Pattern Analysis and Machine Intelligence 2(4), 349-361.
Michalski, R.S. and Chilausky, R.L., 1980. Learning by being told and learning from
examples: an experimental comparison of the two methods of knowledge acquisition in
the context of developing an expert system for soybean disease diagnosis. International J
Policy Analysis and Information Systems, 4: 125–161
Michalski, R.S., 1983. A theory and methodology of inductive learning. Artificial
Intelligence, 20: 111–161.
Michalski, R.S., Davis, J.H., Bisht, V.S. and Sinclair, J.B., 1982. PLANT/ds: an expert
consulting system for the diagnosis of soybean diseases. Proc. European Conference on
Artificial Intelligence, Orsay, France, July: 133–138.
Mitchell, T.M., Keller, R.M. and Kedar-Cabelli, S.T., 1986. Explanation-based
generalization: a unifying view. Machine Learning 1(1) 47-80
Monk, T.J., Mitchell, S., Smith, L.A. and Holmes, G., 1994. Geometric comparison of
classifications and rule sets, Proc. AAAI workshop on knowledge discovery in
databases, Seattle, Washington, July: 395-406.
Mooney, R.J., 1992. Encouraging Experimental Results on Learning CNF. Technical
Report, University of Texas.
Murthy, S.K., Kasif, S., Salzberg, S., and Beigel, R., 1993. OC1: Randomized Induction
of Decision Trees. Proc. of the Eleventh National Conference on Artificial Intelligence.
Washington, D.C.: 322–327
Ousterhout, J. K., 1994. TCL and TK toolkit, Addison-Wesley.
Pazzani, M. and Kibler, D. 1992, The utility of knowledge in inductive learning. Machine
Learning 9(1), 57-94
Piatetsky–Shapiro, G. and Frawley, W.J. (Editors), 1991: Knowledge discovery in
databases. AAAI Press, Menlo Park, CA. 525 pp.
22. 22
Quinlan, J.R., 1986. Induction of decision trees. Machine Learning, 1: 81–106.
Quinlan, J.R., 1990. Learning Logical Definitions from Relations. Machine Learning, 5:
239-266.
Quinlan, J.R., 1991. Determinate Literals in Inductive Logic Programming. Proc. 12th
International Joint Conference on Artificial Intelligence: 746-750, Morgan Kaufmann.
Quinlan, J.R., 1992. C4.5: Programs for Machine Learning. Morgan Kaufmann. 302 pp.
Quinlan, J.R., and Cameron-Jones, R.M., 1993. FOIL: a midterm report. Proc. European
Conference on Machine Learning, Springer Verlag: 3-20
Schank, R.C., Collins, G.C. and Hunter, L.E., 1986. Transcending inductive category
formation in learning. Behavioral and Brain Sciences, 9: 639–651.
Winston, P.H. 1972. Learning structural descriptions from examples. in P.H. Winston, Ed.,
The Psychology of Computer Vision, McGraw-Hill, New York, 689-702.
Witten, I.H. and MacDonald, B.A. 1988 Using concept learning for knowledge acquisition,
Int. J. Man Machine Studies. 29(2) 171-196
23. 23
color: Tree-structured attribute shape:
regular-polygon oval
Linear attribute
size: 1
2
3
4
5
(a)
Attribute vector Vector of generalized attributes
(sample concept)
color and color and
size = and and
shape = shape = convex
(b) (c)
Figure 1 Attribute domain (adapted from Haussler, 1987)
24. 24
Environmental descriptors Condition of leaves Condition of stem
time of occurrence July leaf spots stem lodging
precipitation above normal leaf spot colour stem cankers
temperature normal colour of spot on other side canker lesion colour
cropping history 4 years yellow leaf spot halos reddish canker margin
damaged area whole fields leaf spot margins fruiting bodies on stem
severity mild raised leaf spots external decay of stem
plant height normal leaf spot growth mycelium on stem
leaf spot size external discolouration
Condition of seed normal shot-holing location of discolouration
mould growth absent shredding internal discolouration
discolouration absent leaf malformation sclerotia
discolouration colour — premature defoliation
size normal leaf mildew growth Condition of roots
shriveling absent leaf discolouration root rot
position of affected leaves root galls or cysts
Condition of fruit pods normal condition of lower leaves root sclerotia
fruit pods normal leaf withering and wilting
fruit spots absent
(a) Diagnosis Brown spot
(b) Rhizoctonia root rot IF [leaves=normal AND stem=abnormal AND
stem-cankers=below-soil-line AND canker-lesion-colour=brown]
OR [leaf-malformation=absent AND stem=abnormal AND
stem-cankers=below-soil-line AND canker-lesion-colour=brown]
Figure 2 Example record and rule in the soybean disease classification problem
26. 26
Transfer out date
<= 900420 > 900420
Transfer out date •••
<= 880217 > 880217
Unknown Animal Date of Birth
<= 860811 > 860811
Transfer out date Died
<= 890613 > 890613
Cause of fate •••
Injury Bloat Calving Grass Injury Low Other Milk Empty Old Udder
Trouble Staggers Producer Causes Fever age breakdown
Sold Died Sold Sold Sold Sold Sold Sold Sold Sold
Mating date
<= 890613 > 890613
Sold Animal Key
<= 2811510 > 2811510
Died Sold
Figure 4: Decision tree induced from raw herd data
27. 27
Age
<= 2 >2
Retained Payment BI
relative to herd
<= -10.8 > -10.8
Milk Volume PI Retained
relative to herd
<= -33.93 > -33.93
Culled Retained
Figure 5. Decision tree from processed data set
28. 28
1 Minimal rules A complete, correct, and minimal set of decision rules
2 Adequate rules A complete and correct set of rules that nevertheless contains
redundant rules and references to irrelevant attributes
3 Critical cases A critical set of cases described in terms of a minimal set of
relevant attributes with correct decisions
4 Source of cases A source of cases that contains such critical examples described in
terms of a minimal set of relevant attributes with correct decisions
5 Irrelevant attributes As for 4 but with cases described in terms of attributes which
include ones that are irrelevant to the decision
6 Incorrect decisions As for 4 but with only a greater-than-chance probability of correct
decisions
7 Irrelevant attributes As for 5 but with only a greater-than-chance probability of correct
and incorrect decisions decisions
Table 1 Levels of quality of input (Gaines, 1991)
29. 29
Scheme Learning approach Reference
Unsupervised AUTOCLASS Bayesian clustering Cheeseman et al. (1988)
CLASSWEB Incremental conceptual clustering Fisher et al. (1987), Fisher (1989)
Gennari (1989)
Supervised C4.5 Decision tree induction Quinlan (1992)
OC1 Oblique decision tree induction for Murthy et al. (1993)
numeric data
CNF & DNF Conjunctive and disjunctive normal Mooney (1992)
form decision trees respectively
PRISM DNF rule generator Cendrowska (1987)
INDUCT Improved PRISM Gaines (1991)
FOIL First-order inductive learner Quinlan (1990), Quinlan (1991),
Quinlan et al. (1993), Cameron-
Jones et al. (1993)
Table 2 Machine learning schemes currently included in the WEKA workbench
30. 30
Relation Number of Recording basis
attributes
Animal Birth Identification 3 Once
Animal Sire 1 Once
Animal 6 Once
Test Number Identification 1 Monthly
Animal Location 3×6 When moved
Female Parturition 5 When calving
New Born Animal 3×4 When calving
Female Reproductive Status 3 Once
Female mating 10 × 3 When mated
Animal Lactation 60 Yearly
Test Day Production Detail 12 × 43 Monthly
Non production trait survey 30 Once
Animal Cross Breed 3×2 Once
Animal Lactation—Dam 12 Once
Female Parturition—Dam 5 Once
New Born Animal—Dam 3×4 When dam calves
Animal—Dam—Sire 2 Once
Table 3: Dairy herd database relations