Post-processing of associative classification rules using closed sets

For a classifier, besides classification capability, its size is another vital aspect. In pursuit of high performance, many classifiers do not take into consideration their sizes and contain numerous both essential and insignificant rules. This, however, may bring adverse situation to classifier, for its efficiency will been put down greatly by redundant rules. Hence, it is necessary to eliminate those unwanted rules. In this paper, we propose a fast post-processing approach to remove insignificant rules. The basis of this method is the dependent relation between rules regarding to data objects, from which closed sets can be derived. The experimental evaluation on UCI benchmark datasets using two typical classifiers shows that the proposed method is competent for discarding lots of superfluous rules without degrading classification capability greatly. In particular, the computational cost of our approach is extremely lower than the Apriori-like method.