Title: Item-Based Collaborative Filtering Using the Big Five Personality Traits

Abstract:
Collaborative filtering is a popular technique that makes high quality recommendations. It is also able to recommend items from multiple domains (e.g. books vs. movies). However, we found that it makes more accurate recommendation within one domain than across domains. The proposed system uses the personality of users to generate the recommendations. It develops a profile for each item that reflects the personality of users who like it. Item profiles are used to give item-based collaborative filtering recommendations. In the experiments, it is presented that the pure version of the proposed system permits high accuracy. Moreover, the prediction accuracy becomes greater when the system makes cross-domain recommendations than when it works in one domain.