This is part of the Yelp Data Challenge that you can find here: https://www.yelp.com/dataset/challenge My goal is to determine the features that is most predictive of a rating. This would allow restaurants to gain meaningful insight on the kinds of customer experiences that corresponds to ratings. What is the more important, service or food? This project introduces a method for finding the most informative features for scale based classification (rating of 1-5). Classification is then performed with SVM.
FilterRestaurants.RSubset only restaurant reviews from the datasetgeneratewordclouds.RExploratory analysis with word clouds for each ratingcorrelationwordlist.RFor a list of features in term document matrix form, assign a Coincidence Strength Factor score to quantify how informative a feature is (see yelpdatachallengecoincidence.pdf for an explanation)naivebayesevensample2.RUse naive bayes classifier to predict ratingssvmmodel2.RUse SVM classifier to predict ratings
Mike Huang, contact: [email protected]