A hybrid model of grey relational analysis, particle swarm optimization and support vector regression in forecasting perishable commodities sales volume

• Main Authors: LAN, HAO-HUA, 藍浩華
• Other Authors: Koo, Tong-Yuan
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/97365700742364723942

碩士 === 國立雲林科技大學 === 工業工程與管理系 === 104 === Today Taiwan has convenience stores everywhere, to inroads into outside food market, fresh species also provide more diverse, leading to daily produce scrapped commodity very much. In practice, the demand for perishable commodities tend to be because of the impact of external environmental factors cause large fluctuations, so this study chose support vector regression (SVR) as the main predictive models because it considers the structural risk minimization, and in the sales forecast has a good result.. However SVR often subject to the combined effects of parameters, in order to solve this shortcoming, this study will use grey relational analysis screening influence sales factor, then use the particle swarm algorithm to find the SVR parameter combination to construct model (PSO-SVR). Finally, using Mean Absolute Error (MAE) for evaluation of the model. The results show , predictions GRA-SVR model and PSO-SVR model is even better than the SVR model, which the PSO-SVR performed better. Therefore, it can prove the validity of the research methods.