作　　者： (肖程望）; (卢军）; (余力耕）; (张弛）;

机构地区： 武汉邮电科学研究院,武汉430074

出　　处： 《计算机系统应用》 2017年第9期200-204,共5页

摘　　要： 传统分类系统往往选择朴素贝叶斯算法作为分类算法,在研究过程中,发现朴素贝叶斯模型(NBC)具有以下前提条件:所有特征互不影响,并且特征属性的权重为1.研究后发现并非如此,在对数据进行分类时,有的特征可能对分类结果的影响较大,有的可能对结果影响较小.为了对算法进行优化,需要给不同的特征附上不同的权重值,才能更加客观的获得分类结果.本文研究了两种传统的基于属性权重的计算方法,同时考虑到手机取证数据的特点,提出两种适应于手机取证数据的改进权重计算方法,并对其改进原理进行研究,比较改进后的权重计算方法与传统的权重计算方法,在使用相同分类算法与相同数据的情况下,对分类结果的影响. The traditional classification system often chooses the Naive Bayesian algorithm as the classification algorithm.In the course of the study, we find that the Naive Bayesian model（NBC） has the following conditions： all the characteristics do not mutually influence each other, and the feature attribute weights is 1. But we find that is not the case after a study. In the classification of data, some features may have a greater impact on the classification results, while some may have little impact. In order to optimize the algorithm, we need to attach different weights to different features, so as to obtain the classification results more objectively. This paper studies two kinds of calculation methods of attributing weight based on the traditional algorithm. At the same time, considering the characteristics of mobile phone forensic data,it proposes the calculation method of two kinds of improved weight suitable for mobile phone forensic data. This paper researches the improvement principle of research, compares the improved calculation method of weight with the traditional calculation method in their impacts on the classification results using the same classification algorithm with the same data.

关 键 词： 手机取证 权重计算 分类算法