Augmenting feature model through customer preference mining by hybrid sentiment analysis

A feature model is an essential tool to identify variability and commonality within a product line of an enterprise, assisting stakeholders to configure product lines and to discover opportunities for reuse. However, the number of product variants needed to satisfy individual customer needs is still an open question, as feature models do not incorporate any direct customer preference information. In this paper, we propose to incorporate customer preference information into feature models using sentiment analysis of user-generated online product reviews. The proposed sentiment analysis method is a hybrid combination of affective lexicons and a rough-set technique. It is able to predict sentence sentiments for individual product features with acceptable accuracy, and thus augment a feature model by integrating positive and negative opinions of the customers. Such opinionated customer preference information is regarded as one attribute of the features, which helps to decide the number of variants needed within a product line. Finally, we demonstrate the feasibility and potential of the proposed method via an application case of Kindle Fire HD tablets.