An Associative Classifier Adopting Fuzzy Weighted Rules Based on Information Gain

Classification algorithms based on association rules have been proven with higher accuracy and better understandability when compared with classical classifiers. These advantages make them be widely used in the application of intelligent decision systems. However, in some specific fields, such as health care field, people hope to use more prior knowledge and focus more attentions on properties owning strong correlation with class labels in the process of modeling. In this paper, a fuzzy weighted associative classifier based on information gain is proposed. This associative classifier employs an attribute selection strategy based on information gain to determine attribute importance degree and assigns corresponding weights such that the more important attributes are paid more attentions. In addition, the proposed algorithm applies the fuzzy sets to discretizing the numeric variables instead of partitioning directly for avoiding the sharp boundary issues. After implementation, the new classifier is tested with benchmark data from the UCI machine learning repository. Experimental results show that there is an improvement in classification accuracy and reduction in rules redundancy.