A Comprehensive Survey on Fake Review Detection System with Future Directions

Reviews on the web tend to sway the decision of prospective consumers, who are planning to buy that particular product or service. Many positive reviews make the customer inclined to buy that product even if it is costlier than other products in the same category. However, many negative reviews make the customers wary of buying it. This gave birth to the fake review creation business. Companies buy fake reviews in bulk to downgrade their competitors or artificially improve their credibility. Fake reviews may be generated by a machine or by people who have been paid for them. Fake reviews are near to impossible to be detected by the human eye. Hence, the need for fake review detection is gaining popularity. Researchers have worked on machine learning and deep learning techniques. But these algorithms suffer from many limitations, such as the small size of training data, little or no consistency of solution on different datasets, concept drift, the adaptation of fake reviews over a period of time, etc. This survey has summarized the existing work based on research techniques as well as on the identifying features. It has also analyzed the challenges and presented future directions for further research.