Ranking Online Services by Aggregating Ordinal Preferences

Abstract

With the increase of the number of online services, it is more and more difficult for customers to make the decision on service selection. Services ranking mechanism is an important service for e-commerce that facilitates consumers’ decision-making process. Traditional services ranking methods ignore the fact that customers cannot rate services under the same criteria, which leads to the ratings are actually incomparable. In this paper, we propose to exploit the ordinal preferences rather than cardinal ratings to rank online services. Ordinal preferences are elicited from filled ratings so that the intensity of ratings is ignored. We construct a directed graph to depict the set of pairwise preferences between services. Then, the strongest paths of the directed graph are identified and the evaluation values of services are calculated based on the strongest paths. We prove our method satisfies some important conditions that a reasonable services ranking method should satisfy. Experiments using real data of movie ratings demonstrate that the proposed method is advantageous over previous methods, and so the proposed method can rank services effectively even the ratings are given by customers with inconsistent criteria. In addition, the experiments also verify that it is more difficult to manipulate the ranking result of our method than existing methods.