HIN-RNN: A Graph Representation Learning Neural Network for Fraudster Group Detection With No Handcrafted Features

Social reviews are indispensable resources for modern consumers’ decision making. To influence the reviews, for financial gains, some companies may choose to pay groups of fraudsters rather than individuals to demote or promote products and services. This is because consumers are more likely to be misled by a large amount of similar reviews, produced by a group of fraudsters. Semantic relation such as content similarity (CS) and polarity similarity is an important factor characterizing solicited group frauds. Recent approaches on fraudster group detection employed handcrafted features of group behaviors that failed to capture the semantic relation of review text from the reviewers. In this article, we propose the first neural approach, HIN-RNN, a heterogeneous information network (HIN) compatible recurrent neural network (RNN) for fraudster group detection that makes use of semantic similarity and requires no handcrafted features. The HIN-RNN provides a unifying architecture for representation learning of each reviewer, with the initial vector as the sum of word embeddings (SoWEs) of all review text written by the same reviewer, concatenated by the ratio of negative reviews. Given a co-review network representing reviewers who have reviewed the same items with similar ratings and the reviewers’ vector representation, a collaboration matrix is captured through the HIN-RNN training. The proposed approach is demonstrated to be effective with marked improvement over state-of-the-art approaches on both the Yelp (22% and 12% in terms of recall and F1-value, respectively) and Amazon (4% and 2% in terms of recall and F1-value, respectively) datasets.