A study on modeling customer preferences for conceptual design

초록

In this paper, we propose a concept selection method to evaluate future market performances of concept candidates, and to choose the best concept among those. The main and interaction effects of product performance factors, economic factors, and time on a market performance are modeled using a Bayesian framework-based Artificial neural network (ANN). The Bayesian framework is employed to measure the potential risk of wrong selection in using a trained ANN model. Based on the measured uncertainty bounds in the predicted future market performance, the most promising and robust concept may be selected. To validate our concept-selection method, we employed an automobile concept selection problem in the U.S. market. Seventeen concepts were assumed to compete in 2013, and the future market share with error bounds was predicted using the trained model based on sale data.

키워드