An XGBoost based evaluation methodology of product color emotion design

Product color emotion design has become an important new design direction following the development of user-centered design. In this paper, we integrated Kansei engineering (KE) and extreme gradient boosting (XGBoost) algorithm to explore a new evaluation method of product color emotion design. Firstly, based on KE, a user emotion image perception space was established using big data, semantic differential (SD), and factor analysis (FA). Secondly, the product color emotion evaluation model based on XGBoost was established using XGBoost algorithm and experimental data. In application, we established an XGBoost based evaluation model of exercise bike color emotion to verify the effectiveness of this evaluation model. We then analyzed the graph of feature importance which was output by the XGBoost model and learned the weight of each color parameter. Next, we used the case of dust mite controller to verify effectiveness of this color design method in different type product. Moreover, we compared predictions of the XGBoost model, multiple linear regression (MLR) model, and the back propagation neural network (BPNN) model. The result proved that the color emotion evaluation model based on XGBoost had a better performance. All the above results indicated that the product color emotion design method based on XGBoost could effectively reveal the user's color emotion and help designers make color scheme decisions. So, it has certain applicability and practicability, and this method can be applied to a variety of color design cases.