Identifying Influential Factors for Yelp Business Ratings

In this paper, we investigate potential factors that may influence business performance on Yelp. We considered businesses’ overall star ratings as a measure of their performance. In order to account for user sentiment and location dynamics we constructed additional features from business and user review data. We experimented with regression (Linear and Decision-Tree) as well as classification (Naive Bayes, Decision Tree and Random Forest) models and found that regression models achieved lower error that classification models. We found that across feature selection techniques, the important factors included sentiment of reviews, business location, neighbourhood, ambience of place, etc. However, user review sentiments tend to greatly influence star ratings in comparison to other factors.