Machine Learning and Natural Language Processing Lab

Master Thesis

Classification Using Association Rules

Stefan Mutter, 2004

---

Association rule mining is a well-known technique in data mining. It is able to reveal all interesting relationships, called associations, in a potentially large database. However, how interesting a rule is depends on the problem a user wants to solve. Existing approaches employ different parameters to guide the search for interesting rules. Classification using association rules combines association rule mining and classification, and is therefore concerned with finding rules that accurately predict a single target (class) variable. The key strength of association rule mining is that all interesting rules are found. The number of associations present in even moderate sized databases can be, however, very large -- usually too large to be applied directly for classification purposes. Therefore, any classification learner using association rules has to perform three major steps: Mining a set of potentially accurate rules, evaluating and pruning rules, and classifying future instances using the found rule set. This thesis compares and combines different approaches for classification using association rules. We use classification using association rules not only to solve classification problems, but also to compare the quality of different confidence-based association rule mining approaches. In this context we show that the quality of rule sets from the standard algorithm for association rule mining can be improved by using a different association rule mining strategy. For this comparison we do a benchmark test using 12 UCI datasets.