文献类型：会议论文

英文题名：Research on Support Vector Regression in the Stock Market Forecasting

作者：Cai, Chun[1];Ma, Qinghua[1];Lv, Shuqiang[1]

第一作者：蔡春

通讯作者：Cai, C[1]

机构：[1]Beijing Union Univ, Coll Arts & Sci, Dept Informat & Sci, Beijing 100191, Peoples R China

第一机构：北京联合大学应用文理学院

通讯机构：[1]corresponding author), Beijing Union Univ, Coll Arts & Sci, Dept Informat & Sci, Beijing 100191, Peoples R China.|[114172]北京联合大学应用文理学院;[11417]北京联合大学;

会议论文集：Conference on Electronic Commerce, Web Application and Communication

会议日期：MAR 17-18, 2012

会议地点：Wuhan, PEOPLES R CHINA

语种：英文

外文关键词：support vector regression; time series; kernel function; parameter selection

摘要：As the stock market has high noise, nonlinearity, uncertainty characteristics as well as traditional neural network forecasting method deficiencies of the problem, this paper presents a support vector regression (SVR) and financial time series methods to forecast future stock market; follows the value investment philosophy, reduces dimensional space by support vector machines the input vector mapped to a high dimensional space, and nonlinear problem into a linear problem, establish the value of the equity investments of prediction models under the principle of structural risk minimization. For an empirical test of the model to get the stock opened at 0.02% error of 3.66%, closing error of 0.00% to 3.41%, the results showed that SVR prediction in the stock market index has certain advantages.