Time-Aware Recommendation Based on User Preference Driven
Thitiporn Neammanee, Saranya Maneeroj · 2018
Time-Aware Recommender System (TARS) is a type of Context-Aware Recommender System that consider time for predicting the rating of the target item. In the current TARS, they give a high importance to the data that is nearby time to the current time or nearby time to the target user data but did not consider preference change, or concept driven, of the target user. Therefore, the result of the current TARS may not good enough. This paper proposes the TARS that can detect the individual preference driven by using FCM algorithm and entropy to find the preference change in the rating timeline. Then, we select the period in the past of the target user that has a similar preference to the current preference period to find the neighbors and predict the target item's rating. The proposed method does not necessary use the entire user data in the past but uses only the data in the past period of the target user that is similar to the current period for prediction. In the experiment, the proposed method is compared with the two current TARS methods on MovieLens dataset. The evaluation results can be confirmed that the proposed method provides more accuracy and coverage than the two current methods. Moreover, the proposed method uses smaller data than the other two methods. Thus, this makes the calculation time faster.