Consumers are dependent on online directory websites such as Yelp on a daily basis to find a reasonable service provider, be it a dentist or a good restaurant for dinner! Yelp most primarily focuses on simple reviews and ratings which means consumers might have to sift through dozens of reviews to find a good business. Yelp gets constant feedback on the usefulness of the reviews posted by allowing users to vote the review “useful”. However, this would mean multiple users sifting through them and providing a feedback. The goal of the project is to predict the number of “useful” votes a review will get and thereby have them higher up the review listings. By implementing this project we aim to get an understanding of how to approach a real world business problem with traditional machine learning algorithms to gain useful insights.

Yelp provided data on the following : 11,537 businesses 8,282 checkin sets 43,873 users 229,907 reviews

Each file is composed of a single object type, one JSON object per line. The training data was recorded on 2013-01-19. The testing data contains reviews, businesses, users, and checkins from the period between 2013-01-19 and 2013-03-12