A Feature-Based Reputation Model for Product Evaluation

Knowing the strengths and weaknesses of a product is very important for manufacturers and customers to make decisions. Several sentiment analysis systems are proposed to determine the opinions of customers about products and product features. However, the aggregation methods used are not able to estimate a true reputation value and to reflect the recent opinions quickly. Most of these systems are based on single source and therefore suffer from availability and susceptibility issues. In this paper, we propose a multi-source reputation model where several aggregation methods are introduced in order to evaluate product based on features. In addition, we also propose a method which uses four parameters in order to rank the reputability of each rating before considering it for reputation values. The results show that the proposed model estimates good reputation values even in the presence of biased behaviors, robust to false ratings and reflects the newest opinions about product rapidly.