A robust rating aggregation method based on temporal coupled bipartite network

Abstract

As rating expresses preferences in online or offline evaluation tasks, aggregating diverse ratings provided by raters is an essential process for thoroughly assessing the quality of an object, which can aid in decision-making and recommendation. Eliminating the impact of rating distortion on certain objects has attracted significant attention from researchers to design robust rating aggregation methods. However, existing methods are constrained by massive distorting ratings, which usually emerge mainly in specific temporal ranges, namely temporal burstiness. Therefore, we propose a novel robust rating aggregation method based on a temporal coupled bipartite network, which can effectively model the segmentation of ratings to deal with the burstiness. Experimental results and analyses indicate that our method exhibits greater robustness than state-of-the-art methods, particularly in handling significant disturbances occurring within specific temporal intervals. This novel approach holds potential for application in real-time rating platforms.