Yelp rating and review provide useful information of business to its users, and they have significant effect on how users make decisions. How can Yelp help users to choose business that is consistent with his/her taste? We hope to predict which businesses will be favored by user given his/her review history. With this knowledge, Yelp can provide accurate personal recommendations to its users, and thus gain more popularity among users as a handy application.
The majority of our work falls on the generation of sets of features from raw datasets. Due to the high dimension of data, we perform feature selection techniques to get more compact sets of features. The prediction of users' preference in our simple model is indicated as 0(dislike) and 1(like). Then, we extend it to regression model to predict the score of users' preference. The higher score indicates stronger preference. Related work on Yelp dataset includes: sentimental analysis of users reviews, prediction of business attention in future, prediction of rating for subtopics and exploration of hidden topics and hidden topics.