Study On Incremental Text Classification Based On Svm Decision Tree

Text Classification is a task which refers to category a text to a predefined one or more categories, this task is greatly needed in many information management systems. Now there are many classification algorithms, such as Support Vector Machine, Naive Bayes, Decision Tree, Neural Networks, K-nearest Neighbor. With the development of the network technology, there have been some new requirements in text classification, such as a large number of scheduled classes, increasing training samples. These new requirements have led to further research on text categorization technology.According to the current research, we focus on the study of two key technologies, which refers to incremental text classification based on SVM (Support Vector Machine) decision tree. We first construct a valid and reasonable hierarchical structure based on SVM decision tree; then we implement incremental learning based on SVM decision tree. The main work includes the following aspects:(1) For the text hierarchy structure aspect, we propose an improved multi-class classification algorithm based on SVM decision tree. The decision tree is constructed based on new inter-class separability formula, which makes texts more separable at the upper node of the decision tree. Experimental results show the effectiveness of this method.(2) In the incremental learning aspect, we use SVM incremental learning method based on KKT conditions to update the hierarchical classifier. Experimental results show that the algorithm not only can obtain better classification results, but also can effectively reduce training and test time. Therefore, the algorithm we proposed has a very good practical value.