A Meta-Heuristic Approach for Enhancing Performance of Associative Classification

Abstract

Associative Classification is an interesting approach in data mining to create more accurate and easilyinterpretable predictive systems. This approach is often built on both association rule mining and classification techniques, to find a set of rules called association rules for classification (CAR) of label attributes. There are many kinds of associative classification such as CPAR, CBA, CMAR but the accuracy is still low on large datasets, and the running time is not reasonable as well. This paper proposes a heuristic approach to significantly enhance the performance of Associative Classification algorithms in running time, reducing the rule set, and accuracy on large data. Experimental results show that heuristic searching the optimal data set makes the associative classification more useful on big data, and reasonable in practice.