L1-ε -TWIN SUPPORT VECTOR REGRESSION FOR THE DETERMINANTS OF INFLATION:A COMPARATIVE STUDY OF TWO PERIODS IN CHINA
To do the feature selection byε - twin support vector regression for inflation problem,a well known L0 norm optimal problem is introduced and approximated by L1 to get the solutions,which called L1-ε -twin support vector regression (L1- ε -TSVR).L1- ε -TSVR explores and compares the important determinants of the inflation in China between the periods 2001:01 to 2005:12 and 2006:01 to 2012:12.Compared with ε -TSVR,our L1-ε - TSVR not only fits function well,but also does feature selection efficiency.Our experimental results show that Chinese yuan exchange rate and the interest rate have no effect on the inflation during the period 2001:01 to 2005:12.However,with the reform of the exchange rate flexibility and interest rate rule in China,the interest rate and the exchange rate are the most significant explanatory factor for the Chinese inflation over the period 2006:01 to 2012:12.Besides,ferrous metals price index is three as high in the latter period as in the first period,which indicates that the cost driven inflation may appear in China.
support vector regression L1-ε -TSVR feature selection determinants of inflation
Ya-Fen Ye Xiao-Jun Liu Yuan-Hai Shao Jiao-Yu Yuan
Zhijiang College, Zhejiang University of Technology, Hangzhou 310024, China Hualong Bishui Kindergarten, Binjiang, Hangzhou 310052, China
国际会议
11th International Symposium on Operations Research and its Applications(第11届运筹学及其应用国际研讨会)
安徽黄山
英文
212-216
2013-08-23（万方平台首次上网日期，不代表论文的发表时间）