A System Using Deep Learning and Fuzzy Logic to Detect Fake Yelp Reviews

Date of Award

2019

Degree Name

Master of Computer Science (M.C.S.)

Department

Department of Computer Science

Advisor/Chair

Advisor: James Buckley

Abstract

With the prevalence of online searching, looking up online reviews of businesses, such as restaurants, hotel and other services, is a major factor in people's decision making. However, fake reviews cause the sentiment analysis of a corpus of reviews to be clouded. This research uses the YELP data set that is publicly available on the internet. I connect both review content, user information and business information to optimize the fake review detection accuracy. In terms of my solution, I use both feature extraction and deep learning-based classification to detect fake reviews. In feature extraction, features are extracted from reviews using term frequency and frequency-inverted document frequency. I extract features from users' information like number of reviews, review length, number of fake reviews generated, review date, and review helpfulness. I also extract features from businesses like star number, number of reviews, number of associated fake reviews, and type of business. All of these features will be input to different deep learning models (CNN and RNN.) and results compared to determine which machine learning model has the best performance. I then use fuzzy logic to classify the Yelp dataset into 3 groups: fake review, neutral review and true review. This fuzzy classification allows the user to know when a review is real or fake, and to look at those reviews in between to determine their authenticity.

Keywords

Computer Science, opinion mining, deep learning, fake review, data mining

Rights Statement

Copyright © 2019, author

Link to Full Text

Share

COinS