Latent Factor-based Rating Feedback Learning for Restaurants Recommendation

Abstract

Nowadays, when people go out to eat, their choice of restaurant depends not only on taste, but also on many other factors. Therefore, mining what factors of the restaurant the users care about is a key problem for restaurant recommendation. This paper is engaged to mining the latent theme factors of restaurant the users care about and applying the result to restaurant recommendation. In this paper, we used LDA model to extract the latent theme features of the restaurants, calculated the similarity based on latent factors and rating feedbacks to make rating prediction and restaurants recommendation. This paper conducted an experiment with the review data from Yelp dataset, exploring the performance of the algorithm and the optimal theme number K. The results of the experiment showed that the algorithm achieved some improvement in rating prediction. To some content, applying the latent theme distribution to the problem of restaurant recommendation can solve the problem of data sparsity, decrease the computational dimension and raise the accuracy of rating prediction.