Database Reference
In-Depth Information
Summary
As an unsupervised analysis technique that uncovers relationships among items,
association rules find many uses in activities, including market basket analysis,
clickstream analysis, and recommendation engines. Although association rules are
not used to predict outcomes or behaviors, they are good at identifying “interesting”
relationships within items from a large dataset. Quite often, the disclosed
relationships that the association rules suggest do not seem obvious; they, therefore,
provide valuable insights for institutions to improve their business operations.
The Apriori algorithm is one of the earliest and most fundamental algorithms for
association rules. This chapter used a grocery store example to walk through the
steps of Apriori and generate frequent
k
-itemsets and useful rules for downstream
analysis and visualization. A few measures such as support, confidence, lift, and
leverage were discussed. These measures together help identify the interesting rules
and eliminate the coincidental rules. Finally, the chapter discussed some pros and
cons of the Apriori algorithm and highlighted a few methods to improve its
efficiency.
