| With the advent of the era of big data,mining the potential value of data has become one of the popular research topics.Text data is loaded with a large amount of information,although spam text classification is well developed,which is one of the most well-known subjects of text mining,improving the accuracy of the spam text classifier has been a goal pursued by people.BP neural network is a effective nonlinear model neural network,can better fit the non-linearly separable data,is one of the commonly used models of classification problems.Today,we have a large amount of the text data with high latitude,but there are few labeled data.The traditional BP neural network can not solve these problems well.In this paper,I have done the empirical analysis on the spam text classification with the algorithm of the improved BP neural network and the graph based semi-supervised learning method.The main research contents and results are as follows:(1)For the traditional BP neural network,I added the Elastic Net regularization,with good properties In order to avoid the over-fitting problem of BP neural network faced with high dimensional data.I gave the prove of the group effect of the BP neural network with Elastic Net regularization.On the basis of Zhang’s[7]and Fan Qinwei[8]research,I discussed convergence analysis on the BP neural network with Elastic Net regularization.(2)I discussed the graph based semi-supervised learning algorithm,combined with BP neural network with Elastic Net regularization to make an empirical analysis carried out on the spam text classification.It is found that the BP neural network with Elastic Net regularization term is more effective than the BP neural network in dealing with the over-fitting problem.After combined with the graph based semi-supervised learning algorithm,the accuracy of the classification model has been improved too. |
