A Comparative Study of Data-Intensive Demand Modeling Techniques in Relation to Product Design and Development

Abstract

This paper presents a comparative study of choice modeling and classification techniques that are currently being employed in the engineering design community to understand customer purchasing behavior. An in-depth comparison of two similar but distinctive techniques — the Discrete Choice Analysis (DCA) model and the C4.5 Decision Tree (DT) classification model — is performed, highlighting the strengths and limitations of each approach in relation to customer choice preferences modeling. A vehicle data set from a well established data repository is used to evaluate each model based on certain performance metrics; how the models differ in making predictions/classifications, computational complexity (challenges of model generation), ease of model interpretation and robustness of the model in regards to sensitivity analysis, and scale/size of data. The results reveal that both the Discrete Choice Analysis model and the C4.5 Decision Tree classification model can be used at different stages of product design and development to understand and model customer interests and choice behavior. We however believe that the C4.5 Decision Tree may be better suited in predicting attribute relevance in relation to classifying choice patterns while the Discrete Choice Analysis model is better suited to quantify the choice share of each customer choice alternative.