Yelp + + : 10 Times More Information per View

In this project we investigate two machine learning methods, one supervised and one unsupervised, that will allow the information content of Yelp data to be efficiently conveyed to the users. The first is matrix completion via the novel ”max-norm” constraint which out results show to be more powerful than the traditional nuclear norm minimization. The second is text summary via sparse PCA which can provide a concise summary of the available immense text reviews. We implement and run these algorithms on actual Yelp data and provide results.