Pre-study: Trust and Smart Prediction Using Social Relationships for Effective Recommendation Systems
Taewan Noh, Hayoung Oh · The Smart Computing Review · 2015
Trust prediction in social networks has become an important issue, since an increasing number of online users spend time on the services that are provided. Either a similarity or a social relationship is used to predict a trust relationship among users, and our proposed method shows that both can be simultaneously used to incorporate a prediction with low-rank matrix factorization. Our experiment uses Epinionss movie ratings and the review ratings of a famous movie site as a dataset to compare the new method with a random method, which generates a trust relationship randomly, according to prediction accuracy. The results show that 1) the new method has higher accuracy, and that 2) the prediction accuracy remains constant instead of increasing when the number of test subjects increases.