结合聚类思想神经网络文本分类技术研究*

针对传统的基于神经网络文本分类算法收敛速度慢等缺点,在分析了文本分类系统的一般模型,以及在应用了互信息量的特征提取方法提取特征项后,提出了一种基于样本中心的径向基神经网络文本分类算法;并引入了聚类算法的核心思想,改进误差反向传播神经网络分类算法收敛速度较慢的缺点。实验结果表明,提出的改进算法与传统的BP神经网络分类算法相比,具有较高的运算速度和较强的非线性映射能力,在收敛速度和准确程度上也有更好的分类效果。

Text classification algorithm research based on clustering and neural network

This paper analyzed the general model of text categorization system technology,after the application of the mutual information feature extraction,proposed a text classification algorithm based on a sample center radiate basis function(RBF) neural network,and introduced the core idea of clustering algorithm,to improve the back-propagation(BP) neural network classification algorithm convergence slower shortcomings.Experimental results show that,compared with BP network,RBF network not only has higher speed and stronger nonlinear mapping capacity,but also has faster convergence speed and better accuracy on classification effect.Combining the clustering thought,RBF text classification algorithm has larger theory research value and practical application prospect.