Product Rating Prediction Using Centrality Measures in Social Networks

Online recommendation systems provide useful information to users on various products and also allow the users to rate the products. However, they do not usually consider the fact that users trust their connections more than others and that the trusts vary from connection to connection i.e., we value the opinions of our connections differently. Moreover, the importance of connections' opinion changes over time. Thus, there is a need to consider the evolving trust relationships among users. In this work, we use both the user's social connections and non-connections to predict how a user would rate a particular product. We argue that we not only trust our connections more but also the trust varies over time, which we capture using a time-dependent trust matrix. We use the degree and eigen-vector centrality measures in conjunction with the user-item rating matrix to find how the social connections impact how one rates a product. To test the validity of the proposed framework, we use Epinions dataset which provides the ratings for products and trust matrix over 11 time periods. We show the accuracy our predictive model using the mean absolute error.