Font Size: a A A

A Study On The Support Vector Machine Ensemble Learning Mehtod Based On Particle Swarm Optimization

Posted on:2010-03-24Degree:MasterType:Thesis
Country:ChinaCandidate:J S ChenFull Text:PDF
GTID:2178330338975915Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Support vector machine(SVM) is one type of learning machines that is paid wide attention in recent years. Based on statistical learning theory, SVM possesses many merits such as global optimum, well-adapted and excellent generalization performance, so it has been widely applied in the fields such as pattern recognition, regression estimation and density estimation. Generally, almost all researches use single SVM as learner, and multi-SVM learner methods are scarce thought out. Ensemble learning can significantly improve the generalization ability of learning systems through training multiple learning machine and synthesizing the results. If the ensemble learning technology can be introduced to SVM, the generalization performance of SVM may be improved efficiently. Therefore, research on ensemble SVM learning becomes an important research issue.This paper has a systematic research in SVM classification ensemble learning method. It introduces the concept and principles of SVM classification. It systematically researches and analyzes the theoretical analysis, design of achieve method and practical application of ensemble learning technology. It also analyzes the background of ensemble learning. Describes the classification problem in ensemble learning and two main learning methods Bagging and Boosting. Compares the strengths and weaknesses between Bagging and Boosting, and examines the mechanism of their entry into force. Analyzes the principles and background of selective ensemble, and describes a classic selective ensemble method GASEN. At the same time, completed the following research work:1. Propose two SVM approaches, i.e. the classification SVM ensemble learning approach based on Bagging, Bagging_SVM and the classification SVM ensemble learning approach based on AdaBoost, AdaBoost_SVM.2. A selective ensemble algorithm named PSOSEN is proposed to reduce the implementation complexity of selective ensemble approach. PSOSEN uses quickly convergent particles to find the optimal ensemble. By applying the PSOSEN on UCI data sets, the test results demonstrate that the proposed algorithm achieves high speed, and its accuracy and stability are both higher that Bagging and Boosting algorithm. It can become an efficient implementation method of selective ensemble.3. An advanced particle swarm optimization algorithm is proposed to solve the problem which particle swarm optimization easily gets into the local optimal solution. Apply the advanced particle swarm optimization algorithm into classification SVM selective ensemble, and propose a new selective ensemble algorithm based the advanced particle swarm optimization, APSOSEN. By applying the APSOSEN on UCI data sets, the test results demonstrate that the proposed algorithm can effectively solve the local optimization problem of PSOSEN, and improve the correct rate, convergence and ensemble number. It can become an efficient implementation method of selective ensemble.This thesis explores the ensemble SVM learning and selective ensemble SVM learning initially. The problem proposed in the thesis is a new issue for SVM research. Therefore, the obtained results have important significance.
Keywords/Search Tags:support vector machine, ensemble learning, selective ensemble learning, classification, particle swarm optimization
PDF Full Text Request
Related items