An Improved Text Classification Model for Mobile Data Security Testing

In the view of mobile data security detection, text classification model can be realized in the application layer to detect malicious attacks. Since traditional C4.5 decision tree has the disadvantage of no considering about interaction influence between properties in attribute selection, an improved model of C4.5 decision tree based on AdaBoost algorithm is put forward. The problem in measuring the properties of the optimal weak assumptions is to be solved by introducing the weight coefficient of Boosting, which would generate an adaptive adjustment weights at the end of each iteration calculation, so as to reduce the feature subset attribute redundancy and meanwhile, improve the robustness of the classification model. Experimental results illustrate that the proposed text classification model is superior to the traditional method in terms of detection rate and classification accuracy.