Research On The Network Information Filtering Based On SVM

Along with the development of the Internet,users can get the rich and the latest information. At the same time, because of the openness of the Internet, users will inevitably come into contact with pornography,racism,violence,feudal superstition or apparent ideological tendencies information. How to filter the information that has nothing to do with their needs,how to obtain the necessary information quickly and accurately and how to avoid the harmful information has now become an important subject of study.According to the needs of users, Information filtering can search interested information and shield uninterested or harmful information in a dynamic information flow. How to get information, how to express information and how to build user profile are the main issues in information filtering. User profile is constructed through Naive Bayes, KNN, and SVM and so on. Support Vector Machine is a learning machine which is based on Vapnik Chervonenks dimension and structural risk minimization inductive principle, It can solve the problems that are the small sample, nonlinear, high dimension and local minimum points, and applied to text categorization, face recognition and handwriting recognition successfully.On the basis of the network information filtering technology, focusing on accuracy and speed, this paper applies SVM to information filtering, put forward feedback algorithm that is based on SVM, designs and implements the system based on SVM. The main work as follows:1. This paper studies information filtering model and SVM theory.This paper discusses the general model and analyzes the problem that filtering system has, and then focuses on the information filtering key technology: feature selection, text expressing, and classification algorithm and intercepting data packet technology; SVM is an excellent learning method, this paper discusses SVM's training algorithm and multi-classification.2. Research on model selection of Support Vector Machine.Model selection is a problem that has not yet been completely solved. This paper analyzes kernel function and its parameters and points out the method to do model selection, and verifies this method through experiments and the results show that it can find the optimal parameters. 3. This paper proposes information filtering model and researches feedback learning based on SVM.On the basis of information filtering general model and SVM theory, this paper proposes information filtering model based on SVM and analyzes the implementation of each module; Focusing on the problem that information is dynamic, this paper studies feedback learning and incremental learning algorithm, then introduces SVM incremental learning algorithm to feedback learning and constructs feedback incremental learning algorithm. The experimental results show that this method is feasible.4. This paper designs and develops information filtering system based on SVM.On the basis of information filtering model based on SVM, According to layered, modular method, this paper designs and develops information filtering system; The system uses three-layer filtering strategy, including packet intercepted model,SVM training model,filtering model,feedback learning model and so on; The system uses SPI technology to intercept packet, vector space module to express text, SVM method to learn training set to construct user profile and feedback incremental learning algorithm to optimize the user profile.