文献类型：期刊文献

英文题名：Prediction and dispatching of workshop material demand based on least squares support vector regression with genetic algorithm

作者：Wang, Shuai[1];Wang, Qingming

机构：[1]E China Univ Sci & Technol, Dept Integrated Human Sci Phys, Sch Mech Engn, Shanghai 200237, Peoples R China

年份：2012

卷号：15

期号：1

起止页码：213

外文期刊名：INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL

收录：;WOS:【SCI-EXPANDED(收录号:WOS:000301809700024)】；

语种：英文

外文关键词：material demand; genetic algorithm; least squares support vector regression; prediction technique

摘要：Prediction and dispatching of workshop material demand is very important to material distribution. In order to solve the problem of low forecasting and dispatching accuracy of traditional methods for workshop material demand, prediction and dispatching of workshop material demand based on least squares support vector regression with genetic algorithm is presented in the paper. As the choice of the parameters of LSSVR has a great influence on the performance of LSSVR and it is difficult to know beforehand what values of the parameters are appropriate. Therefore, genetic algorithm (GA) is applied to optimize the parameters of LSSVR in the paper, and the optimal parameters are adopted to construct the LSSVR model. Finally, gummed wire is employed as the experimental object, and the gummed wire demands from 2006.10 to 2008.11 are employed to testify the GA-LSSVR model compared with traditional methods. In order to show the superiority of the GA-LSSVR model, the comparison of prediction error among GA-LSSVR, traditional LSSVR and traditional SVR is performed. The MAPE of GA-LSSVR is 0.0205, the MAPE of traditional LSSVR is 0.0308,and the MAPE of traditional SVR is 0.0392.The experimental results show that GA-LSSVR has best prediction results for material demand among GA-LSSVR, traditional LSSVR and traditional SVR.