Title :
Weighted Naive Bayesian Classifier Model Based on Information Gain
Author :
Duan Wei ; Lu Xiang-yang
Author_Institution :
Sch. of Math & Comput. Sci., Jiang xi Sci. & Technol. Normal Coll., Nanchang, China
Abstract :
Regarding to the disadvantage of Naive Bayesian Classifier (NBC), this paper proposes a new weighted Naive Bayesian Classifier model, which is based on information gain theory (IGWNBC). Using information gain of attribute in attribute set in sample space, we can reduce attribute set, and assign relative weight to each classification attribute. And the result of it is that strengthens attributes, which have high relationship with classification and weakens attributes, which have low relationship with classification. By this way, it can keep Naive Bayesian classifier´s easy and effectiveness and improve its classification effect.
Keywords :
Bayes methods; pattern classification; attribute set reduction; information gain theory; weighted naive Bayesian classifier model; Accuracy; Aerospace electronics; Bayesian methods; Classification algorithms; Complexity theory; Correlation; Entropy; Weighted Naive Bayesian Classifier; classification; information gain;
Conference_Titel :
Intelligent System Design and Engineering Application (ISDEA), 2010 International Conference on
Conference_Location :
Changsha
Print_ISBN :
978-1-4244-8333-4
DOI :
10.1109/ISDEA.2010.226