Fake Review Detection on Yelp Dataset

Abstract: Customer reviews are crucially significant in today's modern era. It is preferable for a consumer to interpret good or service or store reviews before having to decide that which and where to purchase. Consumers may well be misled into purchasing low-quality items given the prevalence of spam feedback/reviews, while satisfactory businesses may very well be defamed by fraudulent feedback. Unlike, say, advertisements, online consumer reviews contain experiences from actual people. It thus has a tremendous influence on the level of customers as well as indirectly on firms. Concerningly, these monetary inducements have produced a market for spammers to generate evaluations in order to falsely boost or criticize firms, practices known as opinion spam. To solve this issue, we discover that the usual reviewers' arrival pattern is consistent and generally indifferent to their ranking patterns. In contrary, spam operations are typically brief and associated with the ranking either favorably or unfavorably. Hence, we advise using abnormally correlated temporal variations to spot such threats. In order to view and exploit such associations, we identify and create multivariate data set based on aggregate data