基于属性加权的朴素贝叶斯分类算法改进

朴素贝叶斯分类是一种简单而高效的方法,但是它的属性独立性假设,影响了它的分类性能。针对这种问题,本文提出一种基于属性加权的朴素贝叶斯分类算法。通过分析研究属性之间的相关性,求出条件属性与决策属性的相关系数,同时结合信息论中所涉及的互信息概念,获得新的权重,对不同的条件属性给予不同的权值,从而在保持简单性的基础上有效地提高了朴素贝叶斯算法的分类性能。实验结果表明,该方法可行而且有效。

Naive Bayes Classification Algorithm improved Based on attribute weighted

Nave Bayes classifier is a simple and effective classification method,but its attribute independence assumption affect its classification performance.In response to this problem,a simple method for setting attribute weights for using with Nave Bayes is presented.Through the analysis of the correlation between attributes,computing correlation coefficients between condition attributes and decision attribute,combined with the concept of mutual information to obtain a new weight,different condition attributes are weighted differently.Thus it effectively improve the classification performance of nave Bayes algorithm on the basis of simplicity.Experimental results show that the method is feasible and effective.