Improving the Performance of Recommender Systems by Alleviating the Data Sparsity and Cold Start Problems

Recommender systems, providing users with personalized recommendations from a plethora of choices, have been an important component for e-commerce applications to cope with the information overload problem. Collaborative filtering (CF) is a widely used technique to generate recommendations. The basic principle is that recommendations can be made according to the ratings of like-minded users. However, CF inherently suffers from two severe issues, which are the problems targeted in this research. • Data sparsity refers to the difficulty in finding sufficient reliable similar users since in general the active users only rated a small portion of items; • Cold start refers to the difficulty in generating accurate recommendations for the cold users who only rated a small number of items.