Article Preview
Top1. Introduction
E-commerce has become one of the major way of shopping for products ranging from simple electronics to valuable items. Almeroth, and Zhao (2010) where collective unfair ratings are referred to as collusion Sun and Liu (2012); Swamynathan et al. (2010) and are more complicated and much difficult to detect than the single unfair ratings Sun and Liu (2012). For that reason, in the present work, we are focusing on understanding and identifying unfair rating scores, all good reputation problems, and collusion and manipulation detection. In online shopping, the customers often depend on other customers’ feedback posted through a rating system, before deciding on buying a product (Mukherjee, Liu, & Glance, 2012). Online feedback-based rating systems, also known as online reputation systems, are systems in which users provide ratings to items they bought. Based on the feedback of the product, the consumer decides whether to buy the product or not. This motivates the seller to promote or demote a product of their interest depending on their competitors’ product, by posting rating scores which are unfair (Brown and Morgan, 2006; Harmon, 2004). For example, fraudulent sellers may try to increase their income by submitting positive feedback which increases their product rating (Harmon, 2004). In addition, occasional sellers on eBay boost their reputations unfairly by selling or buying feedbacks (Brown and Morgan, 2006). Unfair rating scores are given singularly or collectively Swamynathan,
In the present work, we are focusing on understanding and identifying unfair rating scores, all good reputation problems, and collusion and manipulation detection.
Figure 1. Generic process of “all good reputation” problem
An online rating system needs truthful feedback in order to work properly. An important part of a rating system is creating honest and representative feedback. It should not only have qualitative, quantitative facts opinion-based process, but also should detect situations where some users may try to mislead the system by providing unfair positive and negative reviews, which is likely to lead to collusion and manipulation. Figure 1 Generic process of a reputation system and consists of three main components: feedback collection, computation and rating scores presentation (Noorian and Ulieru, 2010; Sun and Liu, 2012). Feedback Collection is responsible for collecting feedback from community and feeding it to the rating system. A feedback is the opinion of an evaluator on the quality of an item or a person. Generally, a feedback can be expressed either as a number, such as a textual review, or a rating score (Adler, de Alfaro, Kulshreshtha, and Pye, 2011). A numeric feedback can be a negative or positive number chosen from either a discrete list of options, for example, an integer choice between 1 and 5 representing quality of an item. Feedback computation is responsible for computation of all feedbacks received from evaluators to calculate correct rating scores for people and items. Investigating the feedback computation part after using sentiment classification algorithm is the focus of this study.
According to Reyes-Menendez, Saura, and Martinez-Navalon (2019), it is improbable for all consumers to be satisfied with a product, very positive ratings might be interpreted by users as incredible information. Consequently, it is recommended that companies ensure that the feedback made to them on platforms such as TripAdvisor, contains only actual comments, as well as several less positive comments; this can generate greater customer credibility and reliability of an offered product.
Figure 2 shows how unfair positive and negative reviews affects negatively on reputation scores computation. If the collection part collects unfair positive and negative reviews, the computation part will compute unfair positive and negative reviews and then the rating scores presentation part will present unfair rating scores as output.