1. Apriori, Eclat, Upper Confidence
Bound in Machine Learning
Presented by
Ningthoujam Mahesh Singh
B.E. CSE 7th Semester
Roll No: 202012
2. CONTENTS
• Apriori Algorithm
- Definition and explanation
- How it works
- Advantages and Limitation of Apriori
- Application
• Eclat Algorithm
- Definition and explanation
- How it works
- Advantages and Limitation of Eclat
- Application
• UCB Algorithm
- Definition and explanation
- How it works
- Advantages and Limitation of UCB
- Application
• Conclusion
3. -
Apriori Algorithm: Mining Frequent Itemsets
Definition and Explanation
- Used for frequent itemsets in a dataset
- Concept of association rules
- Identify patterns and relationships between different items.
- Uses a breadth-first search and Hash Tree
How it Works
- Two-step process: finding frequent itemsets and generating association rules.
- Bottom-up approach, starting with individual items and gradually expanding to more
complex patterns.
- Finding the frequent itemsets from the large dataset
4. Advantages and limitations of Apriori
Advantages:
- simplicity
- scalability.
Limitations:
- high memory usage
- time-consuming for large datasets
- works slow compared to other algorithms.
5. Application
- Health industry: Detecting patient’s drugs by grouping on ADRs cause on their characteristics
- E-commerce retail shops: Recommending products based on the products already present in the
user’s cart .
- Hydrological systems: Predicting natural phenomena .
- Diabetic study: Analyzing the relationship between different factors that contribute to diabetes .
- Student’s course selection in the E-Learning platform: Recommending courses based on the
student’s previous course selections .
- Stock management: Analyzing the relationship between different stocks and predicting future
trends `
6. Eclat Algorithm: Mining Dense Itemsets
Definition and Explanation
- Stands for Equivalence Class Clustering and Bottom-Up Lattice Traversal
- Another powerful method for mining dense itemsets in a dataset
- Depth-first search technique to find frequent itemsets in a transaction database
- Vertical representation of the dataset, making it more efficient for larger datasets
How it Works
- Exploiting set intersections
- Each item is associated with a tidset (a set of transaction IDs containing that item)
- Utilizes depth-first search to discover dense itemsets
- Their corresponding support values.
7. Advantages and Limitation of Eclat
Advantages over Apriori algorithm
Memory Requirements: Less memory than Apriori algorithm.
Speed: Typically faster than the Apriori algorithm.
Number of Computations: Does not involve the repeated scanning of the data to compute the
individual support values.
Limitation
T-id Sets (Transaction id): The T-id sets can be long, making it expensive to manipulate
8. Application
Market Basket Analysis: Used by big retailers to determine the association between items.
Medical Diagnosis: -Patients can be cured easily
-identifying the probability of illness for a particular disease.
Protein Sequence: Determining the synthesis of artificial Proteins.
9. Upper Confidence Bound (UCB) Algorithm
Definition and Explanation
- Popular reinforcement learning technique
- Method for solving the multi-armed bandit problem
- Solve the exploration-exploitation dilemma
- Maximize rewards by finding the optimal balance between exploring unknown
options and exploiting known information.
How it Works
- Assigns confidence bounds to each option
- Selects the one with the highest upper bound.
- Exploring less visited options initially and gradually shifts towards
- Exploiting options with higher rewards based on accumulated knowledge
10. Advantages and Disavantages of UCB
Advantages:
-Easy to implement.
-self-contained algorithm, meaning it does not require any prior knowledge of the
environment.
-optimal in the sense that it achieves the best possible regret bound for the Multi-Armed
Bandit problem.
Disadvantages:
- stationary, meaning they do not change over time.
- independent and identically distributed (i.i.d.), meaning they are drawn from the same
distribution at each time step.
- suboptimal in some cases, such as when the rewards are correlated or when there
are delayed rewards.
11. Application
Online Advertising: Optimize online advertising by selecting the best ad
to display based on user behavior
Recommendation Systems: Recommend products or services to users
based on their past behavior
Clinical Trials: Optimize clinical trials by selecting the best treatment for
each patient based on their characteristics
12. Conclusion: Harnessing the Power of ML
Algorithms
• Recap of Presented Algorithms:
- Apriori: Discovering frequent itemsets in data
- Eclat: Mining dense itemsets efficiently
- UCB: Optimal decision-making in reinforcement learning
• The Role of These Algorithms in Machine Learning:
- Empowering data analysis and pattern recognition
- Enabling efficient decision-making in diverse domains