Solvıng The Cold-Start Problem Usıng User Profıle Based Bagged Ensemble Model For Movıe Recommendatıon
Abstract
Effective and accurate recommendations is a key-challenge for today’s ecommerce dominated world. Accurate recommendations can not only improve sales for the organizations, but also provides easy selection opportunities for consumers. This ultimately reduces the choice overload for consumers. Cold start is a major challenge for any recommendation system. This work presents a user-profile based model, UPBEM, to handle the cold start issue in recommendation systems. Correlation between users is identified based on their profile and predictions are made based on the current user’s information, along with the information of similar users. A bagging ensemble is used for the recommendation process. Experiments were conducted on the MovieLens data. Comparisons and results show reduced MAE and RMSE, exhibiting improved performances of the UPBEM model