A Data-Driven Network Analysis Approach to Predicting Customer Choice Sets for Choice Modeling in Engineering Design

In this paper, we propose a data-driven network analysis based approach to predict individual choice sets for customer choice modeling in engineering design. We apply data analytics to mine existing data of customer choice sets, which is then used to predict choice sets for individual customers in a new choice modeling scenario where choice set information is lacking. Product association network is constructed to identify product communities based on existing data of customer choice sets, where links between products reflect the proximity or similarity of two products in customers' perceptual space. To account for customer heterogeneity, customers are classified into clusters (segments) based on their profile attributes and for each cluster the product consideration frequency is computed. For predicting choice sets in a new choice modeling scenario, a probabilistic sampling approach is proposed to integrate product associations, customer segments, and the link strengths in the product association network. In case studies, we first implement the approach using an example with simulated choice set data. The quality of predicted choice sets is examined by assessing the estimation bias of the developed choice model. We then demonstrate the proposed approach using actual survey data of vehicle choice, illustrating the benefits of improving a choice model through choice set prediction and the potential of using such choice models to support engineering design decisions. This research also highlights the benefits and potentials of using network techniques for understanding customer preferences in product design.