Improving the Performance of Boosting for Naive Bayesian Classification

This paper investigates boosting naive Bayesian classiica-tion. It rst shows that boosting cannot improve the accuracy of the naive Bayesian classiier on average in a set of natural domains. By analyzing the reasons of boosting's failures, we propose to introduce tree structures into naive Bayesian classiication to improve the performance of boosting when working with naive Bayesian classiication. The experimental results show that although introducing tree structures into naive Bayesian classiication increases the average error of naive Bayesian clas-siication for individual models, boosting naive Bayesian classiiers with tree structures can achieve signiicantly lower average error than the naive Bayesian classiier, providing a method of successfully applying the boosting technique to naive Bayesian classiication.