Fake Online Reviews Detection and Analysis Using Bert Model

Abstract

The impact of user reviews on the revenue of an e-commerce organization cannot be overstated. Consumers heavily depend on online reviews while taking decisions on purchasing, making credibility of these reviews crucial to businesses. Unfortunately, some companies resort to pay some people to post deceiving reviews, which can mislead consumers and harm a company's reputation. While various techniques have been developed to detect fake reviews over the past decade, there is a lack of comprehensive surveys that analyze and summarize these approaches. This paper aims to bridge the gap by focusing on detecting the fake reviews. It provides an overview of the current datasets available and their collection methods, as well as an analysis of the feature extraction techniques utilized in prior research. We employ statistical ML techniques, including Naïve Bayes, Support Vector Machines, and the transformer BERT, to conduct experiments on Twitter review datasets. Our results show that the Naïve Bayes algorithm achieved an accuracy of 97.14%, while Support Vector Machines achieved an accuracy of 100%. These results provide a baseline for future studies in this area.