Generating Classification Rules for Large DataSets

Classification is an important topic in the field of artificial intelligence. Several approaches have been proposed for classification, based on a given set of training instances. One well-known is the decision tree approach, by which a tree is built and classification rules are extracted from the resulting tree by tracking from the root node down to the leave nodes. The extracted rules can then be used for classification. However, overfitting may occur and generalization can be low for large datasets. In this paper, we propose a data mining approach to find the classification rules. First of all, frequent termsets are generated. Then classification rules are derived from the frequent termsets. Through the threshold settings, a desired set of classification rules can be obtained and overfitting can be alleviated and controlled. Examples are shown to illustrate the effectiveness of the proposed approach. The interestingness measures of the developed rules are computed and compared.