Recommendation Algorithm Based On Cross-Domain

As we now live in a highly information-oriented era, recommender system is to help us to seek information from the wide ocean of data more accurately. Recommender system aims to predict the unknown information based on existing user history information. However, due to inactivity of users that data set is sparse comparing to its size. Thus the user behavior could not be predicted accurately based on one single dataset. Recently, recommender system which integrating information from multiple data sets has been a hot topic. And, many algorithms have been proposed to deal with the problem of cross-domain recommendations. Algorithm of cross-domain recommendations can be divided into text-based and collaborative filtering based.This paper contains the following context:Firstly, the introduction of algorithm of recommender system. Secondly, the introduction of various models using latent factor in cross-domain and single-domain. Thirdly, unlike the previous cross-domain algorithm, our model considers both the sharing factors and individual ones among data sets. Finally, applying rating matrix generated model into single domain. We propose a novel recommendation system based on ratio.