ENHANCED TECHNIQUES FOR ASSOCIATION RULE CLASSIFICATION IN LARGE DATASETS

Associative classification performs knowledge discovery by combining the tasks of the Knowledge Discovery in Databases (KDD), association rule mining and classification. It is a promising method that improves classification in terms of accuracy when compared with traditional classification methods. The major issues with associative classification are its inability to work efficiently with huge datasets and the huge number of association rules generated. In this paper, these issues are addressed through the use of hierarchical partitioning with Frequent Pattern List and multiple projections of pruning algorithms with optimized rule selection procedure. Experimental results prove that these solutions provide the advantage of able to work with both small and large sized datasets without degrading the classification accuracy. The proposed algorithm shows an average efficiency gain of 47.90% with respect to reduction in number of rules, while showing a notable improvement in the classification accuracy.