Reducing the Cold-Start Problem by Explicit Information with Mathematical Set Theory in Recommendation Systems

Abstract

The explosive growth of enormous information makes our world as a global village. Recommendation systems are widely used in many different kinds of commercial web sites. A key challenge is how to provide recommendations when historical data for a user is missing and mining relationship between user and recommender system, a problem generally known as cold-start recommendation. This issue is a well-known problem for recommender systems and much research has been done to find the best way to overcome this. The proposed idea is querying user through an initial interview to get explicitly information, this process proposed as new user preferences to make user profile. In this paper, we present mathematical set theory to solve cold-start recommendation problem. Experiment will perform on Movie-lens dataset and user preferences can be formed into mathematical set. The combinations of multiple preferences represent as subsets and the largest subset will be the first choice as input for the recommender system. User Item matrix strategy will be used to get predicted rating for cold users. Experimental results on mathematical set approach for recommendation on the Movie-lens dataset demonstrate that the proposed approach significantly outperforms existing methods for cold-start recommendation.