降低权重冗余的分类算法CFS-CFW研究

A Classification Algorithm CFS-CFW for Reducing Weight Redundancy

朴素贝叶斯具有强的独立性假设,而特征加权是解决这一问题的方法。CFW算法是一种简单有效的加权算法,但该算法的权重计算公式纳入了特征间冗余性,从而影响为每个特征所赋予的权值,降低了分类精度。针对CFW算法中存在的权重冗余问题,本文提出了CFS-CFW算法。该算法使用特征选择算法CFS有效降低权重冗余性,使得每个特征被赋予更合适的权重。在13个UCI数据集上的实验结果表明,该算法具有更高的分类精度。在UCI的spambase的垃圾邮件分类数据集上,该算法的准确性也更高。

Naive Bayes has a strong independence assumption,and feature weighting is the method to solve this problem.The CFW algorithm is a simple and effective weighting algorithm,but its weight calculation formula incorporates feature redundancy,which affects the weight assigned to each feature and reduces classification accuracy.In response to the weight redundancy problem in the CFW algorithm,this paper proposes the CFS-CFW algorithm.This algorithm uses the feature selection algorithm CFS to effectively reduce weight redundancy,allowing each feature to be assigned more appropriate weights.The experimental results on 13 UCI datasets show that the algorithm has higher classification accuracy.The accuracy of this algorithm is also higher on the spam classification dataset of UCI's spam database.