An Ensemble Classifier Based on Attribute Selection and Diversity Measure

Ensemble approaches to classification have attracted a great deal of interest in recent years. Many methods have been developed to create the diversity among the classifiers. At present, there are two kinds of diversity creation methods: data partitioning and attributes partitioning. In some applications, attribute partitioning methods are capable of performance superior to data partitioning methods in ensemble learning.In this paper, an ensemble learning algorithm based on attribute selection and diversity measure ASDM is proposed. This algorithm adopts the entire measure of diversity in an ensemble classifier. When a classifier is learned on a random attribute subset, the entire diversity between the learned classifier and all current ensemble members are measured. If the diversity is significant, the learned classifier would be added into the ensemble, otherwise the learned classifier would be discarded. The experimental results show that the ensemble classifier based on attributes selection and diversity measure is effective.