Service Objective Evaluation via Exploring Social Users' Rating Behaviors

With the boom of e-commerce, it is a very popular trend for people to share their consumption experience and rate the items on a review site. The information they shared is valuable for new users to judge whether the items have high-quality services. Nowadays, many researchers focus on personalized recommendation and rating prediction. They miss the significance of service objective evaluation. Service objective evaluation is usually represented by star level, which is given by a large number of users. The more user ratings, the more objective evaluation is. But how does it work for new items? It is lack of objectivity if there are few users have rated to the item, such as there are just two ratings. In this paper, we propose a model to solve service objective evaluation by deep understanding social users. As we know, users' tastes and habits are drifting over time. Thus, we focus on exploring user ratings confidence, which denotes the trustworthiness of user ratings in service objective evaluation. We utilize entropy to calculate user ratings confidence. In contrast, we mine the spatial and temporal features of user ratings to constrain confidence. We conduct a series of experiments based on Yelp datasets. Experimental results show the effectiveness of proposed model.