The Application of Neural Network to Simulate Design Thinking Research

• Main Authors: Chun-Chieh Cheng, 鄭俊傑
• Other Authors: Sei-Wo Tseng
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/49912969858434172328

碩士 === 雲林科技大學 === 工業設計系碩士班 === 98 === In a competitive product market, a product life cycle is very short. We have to develop products at the same time and fit the market need. Hence, how do designers promote design efficiency and design different styles quickly and create various styles to compete with competitors In a product design process, we don’t know design thinking about designers. Hence the process we call “Black box work ”. Researchers try to know this by sciences. We know design behaviors and processes by design knowledge and design strategy and a sketch process and use software to simulate it . The research use Neural Network to simulate designer’s thinking in a design process. When it uses each vocabulary values and evaluative values of focus group to verify, the MEAN errors and MSE is in a rang. we set up. So we can use Neural Network simulate each style vocabulary evaluations in a sketch process. When the input is a group of point coordinates data and the output is the vocabulary evaluations and the output nodes is greater than the input, it can reduce MSE and we can gain better evaluations. On the other hand, we use each vocabulary to evaluate styles, because input nodes is smaller than output nodes, so it produces difference between an evaluative result and a real style. Although Study 6 input nodes equals output nodes, but not enough samples can’t evaluate correct values.