A review on cross domain recommendation

Abstract

Social market, consumers frequently connect from ecommerce websites to social networking sites such as Facebook and Twitter. There have been few determinations on accepting the connections between users' community media profiles and their e-commerce activities. Consumers can also post their newly bought products on micro blogs with links to the e-commerce product web pages. Review on Prediction user's buying activities on user's social media profile from the e-commerce. Extract all feature and use for recommendation. Collaborative Filtering does not have several user ratings to base recommendations on, which indications to the cold-start problem. Influence merchandise adopter information for recommendation, we are facing two major challenges. First, review data are actual deafening and often contain dialect, mistakes and emoticons. Product Demo graphic info of many product adopters can be used to describe both products and users, which can be unified into a recommendation. Predict a user's follow-up buying behavior at a specific period with lineage accuracy. Purchase possibility can be leveraged by recommender systems in different circumstances, as well as the zero-query pull-based endorsement consequence. Matrix Factorization to consider user aspects, and show that our protracted yields better analytical correctness compared to traditional Matrix Factorization and to a non-personalized baseline for cold-start product recommendation.