基于小波神经网络的邮件分类算法研究

A study of classification algorithm for E-mails based on the wavelet neural network

小波神经网络模型是将小波理论和神经网络结合起来的一种模型。通过对邮件分类问题的分析,采用由伸缩和平移因子决定的小波基函数代替传统的神经元激励函数的小波神经网络的方法,建立了相应的邮件分类的小波神经网络模型。该模型克服了传统BP神经网络参数不足、隐含层单元数目难以确定、收敛速度较慢等缺点。应用结果表明,该算法在邮件分类中能有效减少平均绝对误差,提高查准率,为邮件分类算法研究提供了一种新的方法。

Wavelet neural network is a kind of neural network. This paper analyses the classification for E-mails and sets up a wavelet neural network model of the classification for E-mails. The classification model is based on retract and translation factors instead of conventional neural network function. This model can overcome the shortcomings of BP networks, such as the shortcoming of the parameters of network, the uncertain unit number of the hidden layer, and the slowly learning rate, etc. The simulated results show that the method of classification for E-mails can decrease the average error and improve the accuracy. It can provide a new way for classification.