Prediction of Review Sentiment and Detection of Fake Reviews in Social Media

摘要

For the past dozen years, social media has been widely used and has slowly become major sources of various information. Some social media sites specialize in people's reviews and making recommendations. Some reviews are positive while others negative. Either way, they have a significant impact on users' decisions of whether purchasing the products or services or not. However, fake reviews and fake users can greatly twist the true opinions that are out there. The detection of fake reviews is critical for any progress toward making great usage of the web. In this work, we analyze the statistics of a Yelp Challenge dataset and propose a simple-but-powerful new detection algorithm based on locations of the reviewers and businesses. We also investigate the prediction of the sentiments of the reviews. We compare the accuracy and false alarm rates of several prediction algorithms. While the results are less than optimum, but there are hopes for strong performance with minor revisions.