The Pros and Cons of Pruning in Classification
A number of techniques are presented in the literature for pruning in both decision tree as well as rules based classifiers. The pruning is used for two purposes; namely, Improve performance, and improve accuracy. As the pruning is reducing the set of rules as well as the size of the tree, the probability of improvement in performance is, therefore high. While on the other side, the pruning may eliminate the interesting information which can lead to reducing the accuracy. In this research, the effects of pruning on the accuracy are studied in detail. The experiments were carried out on the same techniques with and without using pruning strategies and the results of both types are compared. The analysis of the five algorithms over fourteen datasets showed that the unwise selection of pruning strategy can reduce the accuracy.
Keywords and phrases - Rule Pruning, Associative Classification, Classification, Association Rules Mining, Data Mining