A Graph-Based Reputation Assessment System for Online Review Communities
Athanasios I. Salamanis, Grigorios Christainas, Dionysios D. Kehagias, Dimitrios K. Tzovaras · 2020
One major concern regarding the operation of online communities is the trustworthiness of the users, and a reliable way for estimating it is by using reputation systems. These systems aim to provide reliable rankings of users, items and businesses based on feedback provided by the rest of the users in the online community. However, a reliable reputation system should be robust against different types of attacks orchestrated by malicious users. In this paper, we propose a robust graph-based reputation assessment system for online review communities. This system is based on the construction of a weighted bipartite graph that includes both users' and businesses' clusters as nodes. After the users and the businesses of the online community have been grouped into clusters and the bipartite graph has been constructed, the reputation of businesses based on user ratings and the topology of the graph is estimated. The proposed system was evaluated in terms of robustness against malicious attacks using a real-world review dataset, and compared with the simple-average reputation system. Preliminary results indicate that the proposed system yields more robust behaviour against a specific type of malicious attack compared to the simple-average reputation system.