Module-3
ML and Cloud Computing for IOT
IOT data Analytics, Cloud computing for IOT, Cloud based platforms,
Machine learning for IOT data analytics
IOT Data Analytics
• Existence of IoT devices is heavily dependent on data exchanged by them.
• IoT users require translation of their data, not raw data. Leads to adopting data analytics by
which the raw data can be converted into a productive form that provide better solutions.
• The organizations are using IoT analytics to develop new applications and products for the
contentment of customers.
• Analytics can improve the existing operations and infrastructure and make intelligent
choices related to investments for infrastructure development
Process of IOT Analytics
• Generation of Data: A variety of data is generated by the different kinds of IoT devices.
The data can be of various types like room temperature, humidity, electricity consumption
for smart home applications and, health parameters like blood pressure, body
temperature, pulse rate, etc., for smart healthcare. diversification and volume of data
depend on the type of applications.
• Collection of Data: All the generated data have to be accumulated in a depository
sometimes referred like a data warehouse. As the volume of IoT data is increasing rapidly,
its storage and especially safe storage have become challenging. Most of the organizations
lack the infrastructure for gathering data for which they rely on cloud service providers for
storage facilities.
• Analysis: knowing the value of collected data to deliver better performance in the future.
The data analysis team studies the collected information, the past behavior and, responses,
processes it using different methods and, to derive better insights.
Process of IOT Analytics
• Decisions: After analyzing the data, the organizations take appropriate decisions to deliver
better services for improving their business and relationship with their consumers. The
decisions should promote cost-effective solutions to the problems that were detected in
their previous systems.
• Prediction: The last stage of an analytics process is the prediction of challenges and
reasons for failure that may arise in the future. An upgraded or new application after data
analytics may be successful in solving the existing problems.
Types of Analytics
• Descriptive Analytics: enables reducing voluminous data into a more valuable piece of
information. It uses arithmetic operations on the gathered historical data to convert it into
a form that can be easily interpreted. The results for descriptive analytics are available as
visualizations in the form of graphs, pie charts, bar graphs, etc.
• Predictive Analytics: forecast about happenings in the future to help the organizations to
plan accordingly to achieve their business goals. Predictive analytics applies a different type
of mathematical, data mining, and machine learning methods on the gathered data to
anticipate eventual situations.
• Prescriptive Analytics: predicts multiple consequences for an action to strengthen
decision-making for achieving targets on time. It can be viewed as many predictive models
running simultaneously. Predictive analytics help business users to select the best option.
• Diagnostic Analytics: It uses the gathered data to know the reasons for the occurrence of
certain events or behaviors. Diagnostic analytics helps to take a deeper view of the data by
using correlation, data discovery, and data mining approach
Cloud Computing for IoT
• Cloud computing consists of virtually optimized data centers that are capable of providing
hardware, software, and information reserves to the users according to their requirements.
The organizations working with IoT infrastructure can derive maximum benefits by using
the cloud for the storage and analysis of data.
• Integration of IoT and cloud computing has gained importance with the increase of
demand to manage the available data for the creation of valuable services.
• Cloud computing is a highly reliable and scaling platform that enables the distribution of
resources and costs for a large number of users. The organizations are using the cloud to
store their data as it provides unlimited storage, keeps the data secure, universal access for
the end-users and their devices, and at the same time eliminates the need to maintain big
data centers.
• Cloud computing is a kind of infrastructure where storage as well as the processing of data
are being carried outside the device. Using cloud computing for IoT data enables faster
information exchange, processing of the collected data to meet certain demands, and
arriving at intelligent decisions
Deployment Models for Cloud
• 1. Private Cloud:
• A Private cloud is an infrastructure generally owned and managed by a single
organization. The owner of the cloud can control the activities of purchasing,
maintenance, and support.
• Advantages: organization can have direct access to all the systems, able to troubleshoot
any problems without the intervention of service providers. The owners of the cloud can
update their systems to reduce the chances of systems failure.
• Disadvantages: implementation needs more investments as the infrastructure has to be
competent enough to fulfill all the present as well as future requirements.
Deployment Models for Cloud
• 2. Public Cloud:
set up owned by an external service provider or an organization that provides cloud service.
It can be used by any person or any organization by paying for the services and the limits for
the availability of services.
Primary goal of the public cloud providers is to ensure that their services are accessible by
all devices through the Internet and should not require any additional software.
• Advantages: organizations that use a public cloud do not need to worry about hardware
or software deployments but they have to pay only for the services. economical option.
• Disadvantages: Data security is a major concern while using a public cloud as the service
provider is the owner of data and anyone working with the service provider can access
your data
Deployment Models for Cloud
• 3. Hybrid Cloud:
enables the organizations to move data and applications between private and public clouds
based on operations to be performed
Public cloud can be used to store the data that does not require a secure environment.
sensitive data that are critical for the functioning can be stored in a private cloud.
organizations can store their applications in any one of the clouds.
• Advantages: applications that require the same data, either duplicate copies can be
maintained or data can be moved around.
• Disadvantages: Moving of data can be complicated due to the bandwidth restrictions.
• Implementation of a hybrid cloud is complex as the features of both private and public
cloud has to be considered.
• cost of implementation and maintenance is higher
Deployment Models for Cloud
• 4. Community Cloud:
set up used by multiple organizations that are working for a common goal.
organizations share the computing resources, support, and maintenance activities, and the
cost required for the infrastructure.
• Advantages: cost-effective. Sharing of costs reduces economic burden on the organizations.
• Disadvantages: conflicts for ownership and assigning responsibilities
Service Models for Cloud Computing
• Determine methods that will be used to provide services to the client.
• 1. Software as a Service (SaaS)
• services which are accessible through either web browser working as a thin client or by using a
program interface.
• service provider installs and operates the applications in the cloud and the users gain access to it
through the clients. The users work with the applications by entering their credentials without
installing any additional programs and purchasing software or licenses.
• features of the applications used are stored by the provider
• Advantages:
• scalability in the pricing of applications.
• resistance to software attacks.
• easy to handle
• frequent upgrades
• Examples: Online emails, Google apps, and social media platforms
Service Models for Cloud Computing
• 2. Platform as a Service (PaaS)
• developers can create new applications and offer them as services through the Internet.
• reduces the complexities and extra investments required to maintain and buy the
elemental hardware and operating.
• The user cannot control the infrastructure of the cloud but able to regulate the installed
applications based on available configurations.
• Advantages:
• Integration with third-party services - enhancement of applications based on users.
• Examples: Microsoft Azure, IBM cloud
Service Models for Cloud Computing
• 3. Infrastructure as a Service (IaaS)
• supplies infrastructure components like storage, network devices, firewall, and some
basic resources required for processing data.
• user will be able to control operating systems, applications, and networking components.
• basic infrastructure is provided according to the arrival of demands.
• customers are charged according to the virtual space utilized by them and in some cases
based on their usages like for months, days, or hours.
• preferred by organizations and especially by new entrepreneurs.
• multiple users will be able to share the same infrastructure and allow all the users to
access available resources over the Internet
• Examples: Amazon web services, Google compute engine
Cloud-Based IoT Data Analytics Platform
• Need: Able to handle different types of data, huge volumes of data, provide some security
protocols for the data coming from the edge devices, provide processing capabilities for
the edge as well as the on-premises devices ,able to utilize artificial intelligence and
machine learning for better outputs.
• commonly used IoT data analytics platforms:
• 1. Atos Codex :
• provides an end-to-end data analysis solution for the whole information technology value
chain. maximizing the worth of its data cost-effectively. available as-a-service from the
cloud.
• functioning and profits can be enhanced by analyzing the data obtained from its
employees, partners, and consumers.
• data analytics tools for developing revenue-generating methods by and upgrade their
services by analyzing network activities, the behavior of consumers, and their likes or
dislikes.
Cloud-Based IoT Data Analytics Platform
• 2. AWS IoT:
• procedures for the conversion of raw and unstructured data collected by sensors into a
sophisticated analytical data. AWS steps.
• Collect: collection of data is done by the channels. A channel is an entry point of the
analytics system and collects data from multiple sources and various formats.
• Process: data from channels are sent to the pipeline for processing. performs filtering,
enrichment, and transformation of messages while preprocessing the received data.
• Store: The messages processed by the pipeline are sent to the data store. The types or
number of data stores depends upon the pipeline configuration and the type of
messages that may arrive from various devices or location.
• Analyze: The data from the data store are used to prepare different kinds of datasets
like SQL dataset or container dataset.
• Build: This stage is used for the improved analysis and visualization of the received
analytics which can be used to develop new systems and applications
Cloud-Based IoT Data Analytics Platform
• 3. IBM Watson IoT
• Three basic services: connection, data analysis, and block-chain.
• connects with the devices to gather data and transform them into proper format.
• use predictive and cognitive analysis to help the organizations to identify better
prospects for boosting their business.
• It allows interaction through languages used commonly by the people and then
processes it into a form required by machines or algorithms.
• performs textual analysis to examine the gathered data and tries to generate new
information.
• ML is used to automate the process of transforming or conversions of data
Cloud-Based IoT Data Analytics Platform
• 4. Microsoft Azure IoT
• enables speedy processing of huge data collected by the different types of IoT devices by
applying AI and ML.
• devices possess the ability to develop secure registration on the cloud and are also able
to send and receive data.
• cloud gateway or IoT hub allows safe reception and storage of data along with facilities
of managing the devices.
• stream processing stage obtains a large stream of data records from the cloud gateway
for processing and evaluation.
• user interface for IoT applications uses visualizations and analytics to facilitate the
management of a large number of devices.
Machine Learning for IoT Analytics in Cloud
• IoT sensors continuously sense the data which need to be processed and sent back to
the devices for taking decisions.
• ML and cloud computing are two essential components for data analytics for achieving
cost-effective operation of an application
• ML Algorithms for Data Analytics:
• 1. Classification: A classification task is used for an application to assign class labels to
specific items from the problem domain.
• An example of classification is identifying emails as spam and not spams.
• Classification algorithms: K-Nearest Neighbors (KNN), SVM, and Naïve Bayes.
Machine Learning for IoT Analytics in Cloud
• 2. Regression: training output variables that are real or continuous.
• regression models are identified based on the type of relationship among the
independent and dependent variables.
• Example: Linear regression and SVR
• Linear regression: determine a linear relationship between output and input variables.
• SVR is a regression model based on SVM that is used for the prediction of a continuous
variable.
Machine Learning for IoT Analytics in Cloud
• 3. Clustering: used to form a group of data points. groups are formed according to
similarities in the data points. The data points in a group will have similarities and there
will be many dissimilarities among the data points from two different groups
• Example: K-means, Density-based spatial clustering of applications (DBSCAN)
• K-means: data points are grouped based on the similarity of distance from the center of
the cluster, called a centroid.
• DBSCAN forms a cluster of data points depending on density. A cluster is formed by
grouping data points of constant densities and considers the data points with low density
as outliers. effective for the identification of noise in data
Machine Learning for IoT Analytics in Cloud
• 4. Feature Extraction: reducing the dimensionality of a raw dataset to form a group of
small dimensions to make data processing easier.
• Principal component analysis (PCA) is used to reduce the dimensionality of the dataset to
improve the quality of analytics and minimize loss of information.
• 5. Anomaly Detection: This technique is used for identifying patterns that do not match a
specified pattern or some items of a dataset. The group of indifferent items may cause
errors resulting in the degradation of the performance of algorithms.
• SVM is trained for only one class of data, the normal class. It learns the limits of these data
points known as decision boundary. decision boundary separates the majority of the data
with only a few points on the other side of origin which are called outliers

10-IoT Data Analytics, Cloud Computing for IoT, Cloud Based platforms, ML for Cloud IoT Analytics-07-03-2024.pptx

  • 1.
    Module-3 ML and CloudComputing for IOT IOT data Analytics, Cloud computing for IOT, Cloud based platforms, Machine learning for IOT data analytics
  • 2.
    IOT Data Analytics •Existence of IoT devices is heavily dependent on data exchanged by them. • IoT users require translation of their data, not raw data. Leads to adopting data analytics by which the raw data can be converted into a productive form that provide better solutions. • The organizations are using IoT analytics to develop new applications and products for the contentment of customers. • Analytics can improve the existing operations and infrastructure and make intelligent choices related to investments for infrastructure development
  • 3.
    Process of IOTAnalytics • Generation of Data: A variety of data is generated by the different kinds of IoT devices. The data can be of various types like room temperature, humidity, electricity consumption for smart home applications and, health parameters like blood pressure, body temperature, pulse rate, etc., for smart healthcare. diversification and volume of data depend on the type of applications. • Collection of Data: All the generated data have to be accumulated in a depository sometimes referred like a data warehouse. As the volume of IoT data is increasing rapidly, its storage and especially safe storage have become challenging. Most of the organizations lack the infrastructure for gathering data for which they rely on cloud service providers for storage facilities. • Analysis: knowing the value of collected data to deliver better performance in the future. The data analysis team studies the collected information, the past behavior and, responses, processes it using different methods and, to derive better insights.
  • 4.
    Process of IOTAnalytics • Decisions: After analyzing the data, the organizations take appropriate decisions to deliver better services for improving their business and relationship with their consumers. The decisions should promote cost-effective solutions to the problems that were detected in their previous systems. • Prediction: The last stage of an analytics process is the prediction of challenges and reasons for failure that may arise in the future. An upgraded or new application after data analytics may be successful in solving the existing problems.
  • 5.
    Types of Analytics •Descriptive Analytics: enables reducing voluminous data into a more valuable piece of information. It uses arithmetic operations on the gathered historical data to convert it into a form that can be easily interpreted. The results for descriptive analytics are available as visualizations in the form of graphs, pie charts, bar graphs, etc. • Predictive Analytics: forecast about happenings in the future to help the organizations to plan accordingly to achieve their business goals. Predictive analytics applies a different type of mathematical, data mining, and machine learning methods on the gathered data to anticipate eventual situations. • Prescriptive Analytics: predicts multiple consequences for an action to strengthen decision-making for achieving targets on time. It can be viewed as many predictive models running simultaneously. Predictive analytics help business users to select the best option. • Diagnostic Analytics: It uses the gathered data to know the reasons for the occurrence of certain events or behaviors. Diagnostic analytics helps to take a deeper view of the data by using correlation, data discovery, and data mining approach
  • 6.
    Cloud Computing forIoT • Cloud computing consists of virtually optimized data centers that are capable of providing hardware, software, and information reserves to the users according to their requirements. The organizations working with IoT infrastructure can derive maximum benefits by using the cloud for the storage and analysis of data. • Integration of IoT and cloud computing has gained importance with the increase of demand to manage the available data for the creation of valuable services. • Cloud computing is a highly reliable and scaling platform that enables the distribution of resources and costs for a large number of users. The organizations are using the cloud to store their data as it provides unlimited storage, keeps the data secure, universal access for the end-users and their devices, and at the same time eliminates the need to maintain big data centers. • Cloud computing is a kind of infrastructure where storage as well as the processing of data are being carried outside the device. Using cloud computing for IoT data enables faster information exchange, processing of the collected data to meet certain demands, and arriving at intelligent decisions
  • 7.
    Deployment Models forCloud • 1. Private Cloud: • A Private cloud is an infrastructure generally owned and managed by a single organization. The owner of the cloud can control the activities of purchasing, maintenance, and support. • Advantages: organization can have direct access to all the systems, able to troubleshoot any problems without the intervention of service providers. The owners of the cloud can update their systems to reduce the chances of systems failure. • Disadvantages: implementation needs more investments as the infrastructure has to be competent enough to fulfill all the present as well as future requirements.
  • 8.
    Deployment Models forCloud • 2. Public Cloud: set up owned by an external service provider or an organization that provides cloud service. It can be used by any person or any organization by paying for the services and the limits for the availability of services. Primary goal of the public cloud providers is to ensure that their services are accessible by all devices through the Internet and should not require any additional software. • Advantages: organizations that use a public cloud do not need to worry about hardware or software deployments but they have to pay only for the services. economical option. • Disadvantages: Data security is a major concern while using a public cloud as the service provider is the owner of data and anyone working with the service provider can access your data
  • 9.
    Deployment Models forCloud • 3. Hybrid Cloud: enables the organizations to move data and applications between private and public clouds based on operations to be performed Public cloud can be used to store the data that does not require a secure environment. sensitive data that are critical for the functioning can be stored in a private cloud. organizations can store their applications in any one of the clouds. • Advantages: applications that require the same data, either duplicate copies can be maintained or data can be moved around. • Disadvantages: Moving of data can be complicated due to the bandwidth restrictions. • Implementation of a hybrid cloud is complex as the features of both private and public cloud has to be considered. • cost of implementation and maintenance is higher
  • 10.
    Deployment Models forCloud • 4. Community Cloud: set up used by multiple organizations that are working for a common goal. organizations share the computing resources, support, and maintenance activities, and the cost required for the infrastructure. • Advantages: cost-effective. Sharing of costs reduces economic burden on the organizations. • Disadvantages: conflicts for ownership and assigning responsibilities
  • 11.
    Service Models forCloud Computing • Determine methods that will be used to provide services to the client. • 1. Software as a Service (SaaS) • services which are accessible through either web browser working as a thin client or by using a program interface. • service provider installs and operates the applications in the cloud and the users gain access to it through the clients. The users work with the applications by entering their credentials without installing any additional programs and purchasing software or licenses. • features of the applications used are stored by the provider • Advantages: • scalability in the pricing of applications. • resistance to software attacks. • easy to handle • frequent upgrades • Examples: Online emails, Google apps, and social media platforms
  • 12.
    Service Models forCloud Computing • 2. Platform as a Service (PaaS) • developers can create new applications and offer them as services through the Internet. • reduces the complexities and extra investments required to maintain and buy the elemental hardware and operating. • The user cannot control the infrastructure of the cloud but able to regulate the installed applications based on available configurations. • Advantages: • Integration with third-party services - enhancement of applications based on users. • Examples: Microsoft Azure, IBM cloud
  • 13.
    Service Models forCloud Computing • 3. Infrastructure as a Service (IaaS) • supplies infrastructure components like storage, network devices, firewall, and some basic resources required for processing data. • user will be able to control operating systems, applications, and networking components. • basic infrastructure is provided according to the arrival of demands. • customers are charged according to the virtual space utilized by them and in some cases based on their usages like for months, days, or hours. • preferred by organizations and especially by new entrepreneurs. • multiple users will be able to share the same infrastructure and allow all the users to access available resources over the Internet • Examples: Amazon web services, Google compute engine
  • 14.
    Cloud-Based IoT DataAnalytics Platform • Need: Able to handle different types of data, huge volumes of data, provide some security protocols for the data coming from the edge devices, provide processing capabilities for the edge as well as the on-premises devices ,able to utilize artificial intelligence and machine learning for better outputs. • commonly used IoT data analytics platforms: • 1. Atos Codex : • provides an end-to-end data analysis solution for the whole information technology value chain. maximizing the worth of its data cost-effectively. available as-a-service from the cloud. • functioning and profits can be enhanced by analyzing the data obtained from its employees, partners, and consumers. • data analytics tools for developing revenue-generating methods by and upgrade their services by analyzing network activities, the behavior of consumers, and their likes or dislikes.
  • 15.
    Cloud-Based IoT DataAnalytics Platform • 2. AWS IoT: • procedures for the conversion of raw and unstructured data collected by sensors into a sophisticated analytical data. AWS steps. • Collect: collection of data is done by the channels. A channel is an entry point of the analytics system and collects data from multiple sources and various formats. • Process: data from channels are sent to the pipeline for processing. performs filtering, enrichment, and transformation of messages while preprocessing the received data. • Store: The messages processed by the pipeline are sent to the data store. The types or number of data stores depends upon the pipeline configuration and the type of messages that may arrive from various devices or location. • Analyze: The data from the data store are used to prepare different kinds of datasets like SQL dataset or container dataset. • Build: This stage is used for the improved analysis and visualization of the received analytics which can be used to develop new systems and applications
  • 16.
    Cloud-Based IoT DataAnalytics Platform • 3. IBM Watson IoT • Three basic services: connection, data analysis, and block-chain. • connects with the devices to gather data and transform them into proper format. • use predictive and cognitive analysis to help the organizations to identify better prospects for boosting their business. • It allows interaction through languages used commonly by the people and then processes it into a form required by machines or algorithms. • performs textual analysis to examine the gathered data and tries to generate new information. • ML is used to automate the process of transforming or conversions of data
  • 17.
    Cloud-Based IoT DataAnalytics Platform • 4. Microsoft Azure IoT • enables speedy processing of huge data collected by the different types of IoT devices by applying AI and ML. • devices possess the ability to develop secure registration on the cloud and are also able to send and receive data. • cloud gateway or IoT hub allows safe reception and storage of data along with facilities of managing the devices. • stream processing stage obtains a large stream of data records from the cloud gateway for processing and evaluation. • user interface for IoT applications uses visualizations and analytics to facilitate the management of a large number of devices.
  • 18.
    Machine Learning forIoT Analytics in Cloud • IoT sensors continuously sense the data which need to be processed and sent back to the devices for taking decisions. • ML and cloud computing are two essential components for data analytics for achieving cost-effective operation of an application • ML Algorithms for Data Analytics: • 1. Classification: A classification task is used for an application to assign class labels to specific items from the problem domain. • An example of classification is identifying emails as spam and not spams. • Classification algorithms: K-Nearest Neighbors (KNN), SVM, and Naïve Bayes.
  • 19.
    Machine Learning forIoT Analytics in Cloud • 2. Regression: training output variables that are real or continuous. • regression models are identified based on the type of relationship among the independent and dependent variables. • Example: Linear regression and SVR • Linear regression: determine a linear relationship between output and input variables. • SVR is a regression model based on SVM that is used for the prediction of a continuous variable.
  • 20.
    Machine Learning forIoT Analytics in Cloud • 3. Clustering: used to form a group of data points. groups are formed according to similarities in the data points. The data points in a group will have similarities and there will be many dissimilarities among the data points from two different groups • Example: K-means, Density-based spatial clustering of applications (DBSCAN) • K-means: data points are grouped based on the similarity of distance from the center of the cluster, called a centroid. • DBSCAN forms a cluster of data points depending on density. A cluster is formed by grouping data points of constant densities and considers the data points with low density as outliers. effective for the identification of noise in data
  • 21.
    Machine Learning forIoT Analytics in Cloud • 4. Feature Extraction: reducing the dimensionality of a raw dataset to form a group of small dimensions to make data processing easier. • Principal component analysis (PCA) is used to reduce the dimensionality of the dataset to improve the quality of analytics and minimize loss of information. • 5. Anomaly Detection: This technique is used for identifying patterns that do not match a specified pattern or some items of a dataset. The group of indifferent items may cause errors resulting in the degradation of the performance of algorithms. • SVM is trained for only one class of data, the normal class. It learns the limits of these data points known as decision boundary. decision boundary separates the majority of the data with only a few points on the other side of origin which are called outliers