期刊名称:IAENG International Journal of Computer Science
印刷版ISSN:1819-656X
电子版ISSN:1819-9224
出版年度:2019
卷号:46
期号:2
页码:264-274
出版社:IAENG - International Association of Engineers
摘要:An improved wavelet transform based weightedε-twin support vector regression (WW-ε-TSVR) is proposed inthis paper. In our WW-ε-TSVR, to reduce the impact of outliers,the wavelet weight matrix is introduced to give differentpenalties for the samples located in different places. Further, byusing the ‘plus’ function, a pair of unconstrained minimizationproblems is solved in primal space rather than dual space, inwhich three smooth functions are introduced to replace thenon-differentiable non-smooth ‘plus’ function. To speed up thetraining procedure, the generalized derivative iterativeapproach and Newton iterative approach are used to obtain theapproximate solution, and five more detailed iterativealgorithms are given. At last, the experimental results onseveral artificial and UCI datasets indicate that the proposedmethod is of effectiveness and applicability, it not only givessimilar or better generalization performance with otherpopular methods such as TSVR and ε-TSVR, but also requiresless computational time.