Improve the Accuracy of One Dependence Augmented Naive Bayes by Weighted Attribute

  • Authors:
  • Siwei Jiang;Zhihua Cai

  • Affiliations:
  • Faculty of Computer Science, China University of Geosciences, Wuhan, China 430074;Faculty of Computer Science, China University of Geosciences, Wuhan, China 430074

  • Venue:
  • ISICA '08 Proceedings of the 3rd International Symposium on Advances in Computation and Intelligence
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Naive Bayes is a effective and widely used data mining algorithm for classification, but its unrealistic attribute conditional independence harm its performance. Selecting attributes subsets is an important approach to extend the Naive Bayes, and the state-of-the-art SBC algorithm has better accuracy in classification. In this paper, we review the weighted attribute method for Naive Bayes, and explain SBC is one of the special case in weighted attributed methods. Interesting this method, we present a new one dependence augmented Naive Bayes with weighted attribute called WODANB, which use the fuzzy Support Vector Machine to optimize the weights. Experiment on whole 36 datasets recommended by Weka, results show that WODANB significant outperforms than NB, SBC, ODANB, TAN.