Abstract
Recommender systems are software tools that play an important role of generating a list of recommendations for unseen items based on the past users experience and interactions. One of the most popular approaches is Collaborative Filtering (CF) that considers the users similarities to generate the recommendation. Although, recommender systems have been discovered in many aspects, the popularity bias is still one of the challenges that need to be considered. Therefore, we proposed a novel model that applies a switching technique to solve the long tail recommendation problem (LTRP) when collaborative filtering fails to find the target case using a multi-level method. We evaluate the results using the public dataset 100K Movielens. Our result outperforms all the existing methods through reducing the recommendation error rates for the items in the long tail.