一种基于遗传算法的加权朴素贝叶斯分类算法  

A weighted Naive Bayes classification algorithm based on a genetic algorithm

作　　者：保玉俊 周莉莉 段鹏[1] BAO Yu-jun;ZHOU Li-li;DUAN Peng(School of Mathematics and Computer Sciences,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650000

出　　处：《云南民族大学学报（自然科学版）》2018年第6期525-529,共5页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：云南民族大学研究生创新基金科研项目(2018YJCXS228)

摘　　要：朴素贝叶斯算法因其分类精度高、模型简单等优点而被得到普遍应用,但因为它需要具备很强的属性之间的条件独立性假设,使得其在实际分类学习中很难实现.针对这个缺点,提出了一种基于遗传算法的加权朴素贝叶斯分类算法(G_WNB).该算法将遗传算法(GA)与加权朴素贝叶斯分类算法(WNB)相结合,首先使用基于Rough Set的加权朴素贝叶斯分类算法,综合信息论与代数论给出的属性权值求解方法,计算出每个属性的权值,以初始权值作为初始种群,加权朴素贝叶斯的分类正确率为适应度函数,采用遗传算法优选,以使适应度函数最高的权值为数据集的最终权值,最后使用G_WNB进行分类.实验表明,该算法提高了分类准确率,同时提高了朴素贝叶斯分类器的性能.The Naive Bayes algorithm has been widely used due to its high classification accuracy and simple model, but it is difficult to be established in practical application because it requires conditional independence assumption between strong attributes. Aiming at this shortcoming, a weighted Naive Bayesian classification algorithm （G_WNB） based on a genetic algorithm is proposed. This algorithm combines the genetic algorithm （GA） with the weighted Naive Bayes classification algorithm （WNB）. First, according to a weighted Naive Bayes classification algorithm based on the Rough Set, the attribute-weight solution method is given by the comprehensive information theory and theory of algebras. The weight of each attribute is calculated, the initial weight is taken as the initial population, the fitness function is the classification accuracy of the weighted Naive Bayes, the genetic algorithm is used for optimization so that the maximum weight of the fitness function is the final weight of the data set, and finally G_WNB is used for classification. Experiments show that the algorithm improves the classification accuracy and the performance of the Naive Bayes classifier.

关 键 词：加权朴素贝叶斯 ROUGH集 属性重要度 遗传算法 适应度函数 分类 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]