Book6.dvi.pdf

This chapter introduces frequent itemset discovery, a technique to identify groups of items that appear together in a large number of baskets (transactions) within a data set, often represented as a market-basket model.

Market-Basket Model:

Assumes a many-to-many relationship between two types of objects: items (e.g., products) and baskets (e.g., shopping carts).
Data is typically stored in a file consisting of sequences of baskets, with each basket representing a set of items.
The support threshold (s) determines the minimum number of baskets an itemset must appear in to be considered frequent.

Frequent Itemsets and Their Applications:

Intuition: A set of items is "frequent" if it appears in many baskets.
Formal Definition: An itemset I is frequent if its support (number of baskets containing I) is greater than or equal to the support threshold s.
Applications:

Retailers: Analyze shopping patterns to understand product associations, optimize shelf layouts, and target advertising.
On-line Retailers: Personalize recommendations based on user purchase history and identify cross-selling opportunities.

Challenges:

Large Data Sets: Algorithms need to be efficient for handling massive datasets with millions of items and baskets.
Approximation: Some algorithms provide approximate results quickly, sacrificing complete accuracy.
Parallel Processing: Techniques like MapReduce can leverage parallel computing to speed up processing.

Key Concepts:

Support: Number of baskets containing a given itemset.
Support Threshold (s): Minimum support required for an itemset to be considered frequent.
Frequent Itemset Algorithms:

A-Priori Algorithm: Eliminates large sets early, focusing on smaller subsets.
Approximate Algorithms: Faster but may not find all frequent itemsets.
* Parallel Algorithms: Utilize MapReduce or similar frameworks for distributed processing.