An Associative Classification Approach for Enhancing Prediction of Imbalance Data

Data mining has been received considerable interest in recent years because it can yield effective results in many applications. However, while this field has focused on developing class classifiers for rare events (or imbalance data) in some mining applications such as direct marketing, the associative classification has seldom been addressed in rare event prediction although it is essential for such prediction. Therefore, this paper proposes a novel associative classification method ACRE (Associative Classification for Rare Events), which differs from conventional methods in many ways such as pruning, scoring, sorting, and so on. ACRE prunes rules by selecting the rule with the largest hit rate and updating repetitively the remaining cases and hit rates for the remaining rules. Unlike other methods, ACRE scores positive and negative examples respectively in order to obtain better associative classification rules. The experimental results show that ACRE performs better than other conventional classification methods such as PCAR, C5.0, logistic regression classification, and neural network classification do in both prediction accuracies and running times.