An individual-group-merchant relation model for identifying fake online reviews: an empirical study on a Chinese e-commerce platform

• Published in: Information technology and management Vol. 20; no. 3; pp. 123 - 138
• Main Authors: Yu, Chuanming, Zuo, Yuheng, Feng, Bolin, An, Lu, Chen, Baiyun
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.09.2019; Springer Nature B.V
• Subjects: Business and Management; Business competition; Computer Communication Networks; Consumer behavior; Content analysis; Data Structures and Information Theory; Electronic commerce; Empirical analysis; Identification methods; Indicators; IT in Business; Iterative algorithms; Iterative methods; Operations Research/Decision Theory; Performance enhancement; Product reviews; Regression analysis; Regulatory mechanisms (biology); Training; China; Unsupervised machine learning; Opinion mining; IGMRM; User behaviour modelling; Fake review identification
• ISSN: 1385-951X; 1573-7667
• DOI: 10.1007/s10799-018-0288-1

During the online shopping process, customer reviews strongly influence consumers’ buying behaviour. Fake reviews are increasingly utilized to manipulate products’ reputations. Automatically and effectively identifying fake reviews has become a salient issue. This study proposes a novel individual-group-merchant relation model to automatically identify fake reviews on e-commerce platforms, which focuses on the behavioural characteristics of the stakeholders. Three groups of indicators are proposed, i.e., individual indicators, group indicators and merchant indicators. An unsupervised matrix iteration algorithm is utilized to calculate the fake degree values at individual, group and merchant levels. To validate the model, an empirical study of fake review identification on a Chinese e-commerce platform is implemented. A total of 97,804 reviews related to 93 online stores and 9558 different reviewers are randomly selected as the test data. The experimental results show that the F-measure values of the proposed method in identifying fake reviewers, online merchants and groups with reputation manipulation are 82.62%, 59.26% and 95.12%, respectively. The proposed method outperforms the traditional methods (e.g. Logistic Regression and K nearest neighbour) in identifying fake reviews. It suggests that the combinations of the behaviour indicators with content analysis can effectively improve the performances of the fake review identification. The proposed method is more scalable to large datasets and easier to be employed, as it does not require manual labelling training set and it eliminates the training of classification models. This study greatly contributes to purifying the Chinese environment of business competition and establishing a better regulatory mechanism for credit manipulation in China.