| Summary: | 碩士 === 國立成功大學 === 工業與資訊管理學系碩博士班 === 97 === Because of the age of global competition is coming, every manufacturer would like to provide low price to attract customers and get order in the market. For providing lower selling price or getting higher profit, manufacturer must be control the production cost, not only finds cheaper material , but also improves the yield in manufacturing process to decrease the production cost relatively. In the beginning of new product of developing step, the experience and knowledge about manufacturing process is not sufficient. It is helpful for decreasing the researching time, if using the pre-manufacturing experience for new manufacturing process.
This thesis researches a problem about forecasting yield in new manufacturing process. Because of it is a new manufacturing process, in the situation that datasets is small, we use mega trend diffusion according to the experience about pre-manufacturing process to construct the experience about new manufacturing process. We make some virtual sample that is similar to the data in new manufacturing process to train forecast model. The result shows that, comparing to other forecast model, our model is better than others in accuracy and stability.
|
|---|
