Estimation of Attribute Weights for Naive Bayes Classifier by Using Data Clustering and Decision Tree
Naive Bayes is a classifier which is simple and widely used. Hall has proposed an approach to calculate attribute weight of Na(i)ve Bayes to increase the class prediction. The objective of this study is to improve Halls algorithm by applying a local weighting scheme for optimizing each test data. The proposed approach used data clustering and decision tree algorithm to calculate weights for each attribute and applied to each node of Naive Bayes classifier. Results showed that a data clustering and decision tree algorithms outperformed Halls algorithm upon two parameters which included number of data (large), and number of attribute (large). However the time complexity of this approach might be more than Halls because it uses one more algorithM in calculating attribute weights, clustering algorithm.
attribute weight naive bayes decision tree clustering classification
Hathaichanok Kornchee Anongnart Srivihok
Department of Computer Science, Kasetsart University, Bangkok, Thailand
国际会议
广州
英文
410-415
2008-12-11（万方平台首次上网日期，不代表论文的发表时间）