A Rough Set Based Associative Classifier

Associative Classification integrates both association rule mining and classification tasks. Many studies show that associative classifiers give better accuracy than other traditional classifiers. Traditional classification techniques such as decision trees and RIPPER use heuristic search methods to perform classification. Associative classification system is more robust and makes predictions based on entire dataset. In this paper, we use rough sets for feature reduction. We have also introduced two new criteria for ranking the association rules. This improves the overall accuracy of the classifier. Our preliminary results with some UCIML datasets are very encouraging.