Recommendation for custom product via probabilistic relevance model

Product recommendation system has been widely used in industry especially for e-Commerce companies to solve the problem of information overload. Nonetheless, information overload is also a severe issue in custom product development practice. Sometimes customers can easily get overwhelmed by the vast number of product varieties and it is hard for them to make choices. However, the established product recommendation approaches are primarily for off-the-shelf products, adaptation for custom products has been difficult due to the different scenarios of custom product design. In this paper, a new recommendation method for custom product design is proposed based on probabilistic relevance model. The idea is to calculate the probability that each product meets an active customer´s specifications based on partial product specifications given by the customer. Then the recommendation is presented according to the ranking of probabilities of relevance. Experiments are carried out and the result shows that the presented approach can improve the recommendation efficiency significantly comparing with random recommendation.