Application of Back-Propagation Neuron Network on the Demand casting of Fitness Equipment

• Main Authors: YUAN MING CHIANG, 江元明
• Other Authors: Yu Wen Chen
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/88813025181099670944

碩士 === 大葉大學 === 工業工程與科技管理學系碩士在職專班 === 95 === Fitness equipment is one of the rapid growth sport products in Taiwan. Because of the change of global economic environment, trade liberalization and product diversification, world consumer markets of fitness equipment are innovated and aligned rapidly no matter in logistics or in fashion. Moreover, facing the servere challenge of main China and other countries, Taiwan’s export-oriented fitness equipment industry need to find out an appropriate forecasting method to reduce the difficulties such as, delay delivery due to shortage of goods or increase of management cost owing to the excess stock. Thus, the accuracy of forecasting method will influence the operation cost and management quality directly. This research is based on historical data and monthly (BPNN) sales revenue (1996~2006) of fitness equipment industry, and we use the back propagation neuron network’s of (BPNN’s) two conversion functions (Logsig & Tansig) for forecasting Data of Three month are used as an basis to predict the current value. Matlab is used for our example by minimizing Mean Square Error to get a standard, and export the predicted value. Results of (BPNN) are compared with Moving Average, Simple Exponential Smoothing and Holt’s Model, and we use MAPE as evaluation of predict results, then get the minimum Error as the product demand model of prediction. We conclude that the MAPE has the minimum by using (BPNN) two conversion functions; in addition, Tansig is better than Logsig. Results also prove that (BPNN) is the best method in this study, which is better than other three traditional forecasting methods.