Ranking Based Recommendation using Online Social Users Data

Social voting system is a new segment in online casual networks. This is helpful in providing correct recommendations with the help of factors like social trust etc. Here we propose Matrix factorization (MF) and nearest neighbor-based recommender framework which will help the factors of user activities and also compare them with the peer reviewers, to provide an accurate recommendation. Through investigation with real social voting traces, we test that social network and group affiliation information can automatically improve the accuracy of popularity-based voting recommendation, and social network information control group affiliation information in NN-based methodology. We likewise inspect that social and gathering data is significantly more important to the cold user than to heavy users. In our experiments, simple meta path-based nearest-neighbor (NN) models outperform computation-intensive MF models in hot-voting recommendation, while users interest for cold votings can be better mined by MF models. We further propose a hybrid recommender systems (RS), bagging different single methodology to achieve the best top-k hit rate.