A new rule pruning text categorisation method

Associative classification integrates association rule and classification in data mining to build classifiers that are highly accurate than that of traditional classification approaches such as greedy and decision tree. However, the size of the classifiers produced by associative classification algorithms is usually large and contains insignificant rules. This may degrade the classification accuracy and increases the classification time, thus, pruning becomes an important task. In this paper, we investigate the problem of rule pruning in text categorisation and propose a new rule pruning techniques called High Precedence. Experimental results show that HP derives higher quality and more scalable classifiers than those produced by current pruning methods (lazy and database coverage). In addition, the number of rules generated by the developed pruning procedure is often less than that of lazy pruning.