Building classifiers with association rules based on small key itemsets

摘要

We present a simple method for building classifiers based on class-association rules. The method uses a prefix tree structure for mining the frequent itemsets and classassociation rules extracted from a training dataset. The rules of a classifier are selected from those built on key itemsets with small sizes, having maximal confidences and maximal supports, and correctly classifying each object of the training dataset. The comparisons with some existing methods in classification, via the experimental results on large datasets, show that on average the present method is better in terms of accuracy and computational efficiency.