Chapter 8. Finding Patterns – Market Basket Analysis Using Association Rules

Think back to your last impulse purchase. Maybe in the grocery store checkout lane you bought a pack of chewing gum or a candy bar. Perhaps on a late-night trip for diapers and formula you picked up a caffeinated beverage or a six-pack of beer. You might have even bought this book on a bookseller's recommendation. These impulse buys are no coincidence, as retailers use sophisticated data analysis techniques to identify patterns that will drive retail behavior.

In years past, such recommendations were based on the subjective intuition of marketing professionals and inventory managers. Now, barcode scanners, inventory databases, and online shopping carts have generated transactional data that machine learning can use to learn purchasing patterns. The practice is commonly known as market basket analysis due to the fact that it has been so frequently applied to supermarket data.

Although the technique originated with...

Understanding association rules

The building blocks of a market basket analysis are the items that may appear in any given transaction. Groups of one or more items are surrounded by brackets to indicate that they form a set, or more specifically, an itemset that appears in the data with some regularity. Transactions are specified in terms of itemsets, such as the following transaction that might be found in a typical grocery store:

The result of a market basket analysis is a collection of association rules that specify patterns found in the relationships among items in the itemsets. Association rules are always composed from subsets of itemsets and are denoted by relating one itemset on the left-hand side (LHS) of the rule to another itemset on the right-hand side (RHS) of the rule. The LHS is the condition that needs to be met in order to trigger the rule, and the RHS is the expected result of meeting that condition. A rule identified from the preceding example transaction might be expressed...

Example – identifying frequently purchased groceries with association rules

As noted in this chapter's introduction, market basket analysis is used behind the scenes for the recommendation systems used in many brick-and-mortar and online retailers. The learned association rules indicate the combinations of items that are often purchased together. Knowledge of these patterns provides insight into new ways a grocery chain might optimize the inventory, advertise promotions, or organize the physical layout of the store. For instance, if shoppers frequently purchase coffee or orange juice with a breakfast pastry, it may be possible to increase profit by relocating pastries closer to coffee and juice.

In this tutorial, we will perform a market basket analysis of transactional data from a grocery store. However, the techniques could be applied to many different types of problems, from movie recommendations, to dating sites, to finding dangerous interactions among medications. In doing so, we will...

Summary

Association rules are used to find useful insight in the massive transaction databases of large retailers. As an unsupervised learning process, association rule learners are capable of extracting knowledge from large databases without any prior knowledge of what patterns to seek. The catch is that it takes some effort to reduce the wealth of information into a smaller and more manageable set of results. The Apriori algorithm, which we studied in this chapter, does so by setting minimum thresholds of interestingness, and reporting only the associations meeting these criteria.

We put the Apriori algorithm to work while performing a market basket analysis for a month's worth of transactions at a modestly sized supermarket. Even in this small example, a wealth of associations was identified. Among these, we noted several patterns that may be useful for future marketing campaigns. The same methods we applied are used at much larger retailers on databases many times this size, and can also...