A Comparative Study of Machine Learning Methods for Detection of Fake Online Consumer Reviews

Abstract

Online product reviews provide valuable information for consumer decision making. Customers increasingly rely on the reviews and consider them a trusted source of information. For businesses, it is therefore tempting to purchase fake reviews because competitive advantage can be easily achieved by producing positive or negative fake reviews. Machine learning methods have become a critical tool to automatically identify fake reviews. Recently, deep neural networks have shown promising detection accuracy. However, there have been no studies which compare the performance of state-of-the-art deep learning approaches with traditional machine learning methods, such as Naive Bayes, support vector machines or decision trees. The aim of this study is to examine the performance of several machine learning methods used for the detection of positive and negative fake consumer reviews. Here we show that deep neural networks, including convolutional neural networks and long short term memory, significantly outperform the traditional machine learning methods in terms of accuracy while preserving desirable time performance.