Consistent Model Combination for SVR via Regularization Path

摘要：

　　It is well-known that model combination can improve prediction performance of regression model.We investigate the model combination of Support Vector Regression (SVR) with regularization path in this paper.We first define Le-risk of SVR,and prove that SVR regularization path leads to at least one Le-risk consistent fitted model.Then we establish the Le-risk consistency for convex combination of SVR base fitted model,which gives the mathematical justification for model combination of SVR on regularization path.We then propose an effective method for Beyesian model combination of SVR for improving prediction performance,the inherent piecewise linearity of SVR regularization path makes the construction of the candidate model set simple and efficient.Theoretical analysis and experimental results suggest the feasibility of the proposed method.

关键词： Model Combination Consistency Support Vector Regression Bayesian Model Combination

作者: Mei Wang Kaoping Song Hongjun Lv Shizhong Liao

作者单位: School of Computer and Information Technology,Northeast Petroleum University,Daqing,China;Center for Center for Post-Doctoral Studies of Beijing Deweijiaye Science and Technology Ltd,Beijing,China;Key Center for Post-Doctoral Studies of Beijing Deweijiaye Science and Technology Ltd,Beijing,China School of Computer Science and Technology,Tianjin University,Tianjin,China

会议类型: 国内会议

会议名称: 2014全国理论计算机科学学术年会

会议地点: 济南

会议语种:英文

页码: 1-10

在线出版日期: 2014-10-16（万方平台首次上网日期，不代表论文的发表时间）