Using OWA Approach to Solve Cold-Start Problem of Recommender Systems

The proliferating electronic commerce has led recommender systems to become impressive tools that their ability to leverage the power of data to benefit any enterprise is non-negligible. They are purposed to effectively proffer those items that meet the users' preferences best. A variety of techniques and methods have been designed and developed for recommender systems such as collaborative filtering and demographic-based filtering. This study proposes a new hybrid recommender system that its concentration is mainly on improving the performance and efficiency of operation under an undesirable condition called the "new user cold-start" which is caused by the existence of users that happen to have no ratings or only a small number of ratings. In this hybrid method, we have applied the optimistic exponential type of ordered weighted averaging (OWA) operator to combine the outcoming results of four recommender system strategies. Experiments were conducted over the MovieLens dataset and resulted in a predominance of the proposed hybrid approach dealing with the cold-start conditions.