Consumer preference prediction by using a hybrid evidential reasoning and belief rule-based methodology

Abstract

Consumer preference prediction is a key factor to the success of new product development. This paper presents a hybrid evidential reasoning (ER) and belief rule-based (BRB) methodology for consumer preference prediction and a novel application to orange juices. The orange juices are distinguished by their values of sensory attributes, which are grouped for simplicity into different categories such as appearance, aroma, texture, flavour, and aftertaste. The ER approach is used to aggregate consumer preferences for category attributes into an overall preference, and the BRB methodology is used to model the casual relationships between category attributes and their sensory attributes. The casual relationships between the overall preference and the sensory attributes of orange juices are trained and tested using real data and memorized for prediction or new product design. A case study involving 16 orange juices is conducted using the proposed hybrid ER and BRB methodology to demonstrate its novel applications. The results show that the hybrid ER and BRB methodology can fit and predict consumer preferences with high accuracy.