结合加权特征向量空间模型和RBPNN的文本分类方法

摘要

In this paper, a text classification method combined weighted feature vector space model and the RBPNN are presented. According to the insufficient of traditional text feature extraction method. In the method, the weigthing about text feature is given by the text feature location information and category information, and then the feature frequency is obtained. The characteristic value is calculated using the TFIDF function after that, and the characteristic vector of text is formed. Then the weights between the second network hidden layer and output layer are decided by the least squcre algorithm, so the classification model is built. The experimental results showed that, the good recall and precision are obtained. The performance of text classification method proposed is well.%提出了一种结合加权特征向量空间模型和径向基概率神经网络(RBPNN)的文本分类方法.该方法针对传统的文本特征提取方法的不足,根据文本中特征项的位置信息和所属类别信息定义特征权重,然后,依据特征项的权值计算文档特征项的频数,通过TFIDF函数计算特征值并得到文本的特征向量,最后,采用RBPNN网络分类,通过最小二乘算法求解神经网络的第二隐层和输出层之间的权值,最终训练获得文本分类模型.文本分类实验结果表明,该方法在文本分类中表现出较好的效果,具有较好查全率和查准率.