| Summary: | 碩士 === 國立虎尾科技大學 === 工業管理系工業工程與管理碩士班 === 106 === The bicycle peripheral product manufacturing industry of Taiwan is extremely competitive in the global market. Except for Original Design Manufacturer, many companies have begun to create their own brands. To satisfy the flexible demand of customers and the market, strike a balance between continuously supply and alleviate the pressure of finished goods inventory. As for enterprises, development the sales forecast model is an important key to pursue profits and reduce waste at the same time. Based on the data provided by the case companies, there are seven product categories as the research objects in this study, and the induction made from the characteristics of the industry for 45 factors affect the prediction. The purpose of this study is applying Gaussian Processes, Linear Regression, Multilayer Perceptron, SMOreg of time series forecasting tools based on data mining to product sales forecasting and compared the forecast results with product sales in 2017. The study results show that with the four algorithms as the base, SMOreg got the best results. All of the study results with forecasting error are from 0.85% to 27.92%. In this study, none of tools is compatible to every product, the most suitable method for each object still depends on the decision by user.
|
|---|
