基于属性选择的改进加权朴素贝叶斯分类算法

Improved Weighted Naive Bayes Classification Algorithm Based on Attribute Selection

朴素贝叶斯分类算法简单且高效,但其基于属性间强独立性的假设限制了其应用范围.针对这一问题,提出一种基于属性选择的改进加权朴素贝叶斯分类算法(ASWNBC).该算法将基于相关的属性选择算法(CFS)和加权朴素贝叶斯分类算法(WNBC)相结合,首先使用CFS算法获得属性子集使简化后的属性集尽量满足条件独立性,同时根据不同属性取值对分类结果影响的不同设计新权重作为算法的加权系数,最后使用ASWNBC算法进行分类.实验结果表明,该算法在降低分类消耗时间的同时提高了分类准确率,有效地提高了朴素贝叶斯分类算法的性能.

Naive Bayes Classification is simple and effective, but its strong attribute independency assumption limits its application scope. Concerning this problem, an improved WNBC algorithm is proposed based on attribute selection. This algorithm combines CFS algorithm with WNBC algorithm, it firstly uses CFS algorithm to get an attribute subset so that the simplified attribute subset can meet conditional independency; meanwhile, the algorithm＇s weighting coefficient is designed on that different attribute values have different influences on the classification result. Finally, it uses ASWNBC algorithm to classify datasets. The experimental results show that the proposed algorithm improves the classification accuracy with lower time consumption, therefore heightens the performance of NBC algorithm.