Unsupervised Learning: Association Rules

Abstract

How does demographic information affect what the customer buys? Is bread usually bought together with milk? Does a specific milk brand make any difference? Where should tomatoes be placed to maximize sales? Is bread bought also when both milk and eggs are purchased? In this shopping basket, the customer has tomatoes, carrots, bananas, bread, eggs, soup, milk, etc. Figure 10.1. Application of association rules in market-basket analysis. 2. Association Rules and Transactional Data Continuing our example of market-basket analysis, we represent each product in a store as a Boolean variable, which represents whether an item is present or absent. Each customer's basket is represented as a Boolean vector, denoting which items are purchased. The vectors are analyzed to find which products are frequently bought together (by different customers), i.e., associated with each other. These cooccurrences are represented in the form of association rules: LHS ⇒ RHS [support, confidence] where the left-hand side (LHS) implies the right-hand side (RHS), with a given value of support and confidence. Support and confidence are used to measure the quality of a given rule, in terms of its usefulness (strength) and certainty. Support tells how many examples (transactions) from a data set that was used to generate the rule include items from both LHS and RHS. Confidence expresses how many examples (transactions) that include items from LHS also include items from RHS. Measured values are most often expressed as percentages. An association rule is considered interesting if it satisfies minimum values of confidence and support, which are to be specified by the user (domain expert). The following examples are used to illustrate the concepts. Example: An association rule that describes customers who buy milk and bread. buys (x, milk) ⇒ buys (x, bread) [25%, 60.0%] The rule shows that customers who buy milk also buy bread. The direction of the association, from left to right, shows that buying milk "triggers" buying bread. These items are bought together in 25% of store purchases (transactions), and 60% of the baskets that include milk also include bread.