基于特征加权的网络不良内容识别方法

Feature weighting based identification method for network undesirable content

网络上不良信息的出现是困扰社会精神健康发展的"顽疾",如果不进行有效的过滤,会给搜索服务带来不良影响,极大的影响了社会的和谐稳定发展。提出一种基于特征加权的网络不良内容识别方法,在对网页上的文本预处理后,引入针对不良内容的加权方法,然后再结合KNN、朴素贝叶斯、SVM三种文本分类方法进行实验对比。对比实验结果表明,所采用的方法在识别网络不良内容上的准确率和召回率都有较大提高。

The emergence of network undesirable information is the＂chronic illness＂which persecutes the healthy development of mental social,if the information isn′t filtered effectively,it will bring undesirably affect on the search service,and influences the harmony and stability development of the society. An identification method for network undesirable content based on feature weighting is proposed. The weighting method for the undesirable content is introduced after text pretreatment on web page,and then the proposed method and three test categorization methods of KNN,Naive Bayes and SVM are compared with the experiments. The contrast experimental results show that the adopted method has great improvement on precision and recall of identifying the network undesirable content.