Mining relational context-aware graph for rater identification
Yue Shi, Martha Moon Larson, Alan Hanjalić · 2011
This paper studies the rater identification problem in recommender systems. We propose to approach rater identification by fusing influence from various factors that relate to users. The rater likelihood is modeled from a probabilistic point of view as the conditional probability of a user given sources of contextual information, resulting in an aggregation model that fuses all the available information sources pertaining to a particular user. The result is a relational context-aware graph. A random walk with restart is used to calculate the proximity scores over this graph, which are used to identify raters. We compare our approach with several baselines in a set of experiments performed on the CAMRa2011 challenge dataset. The results demonstrate the superiority of our approach in predicting the identity of a rater who rated a particular movie within a given household.