Evaluation of Collaborative Filtering and Knowledge Transfer Based Cross Domain Recommendation Models

Nikita Taneja and Dr. Hardeo K Thakur

Cross domain recommendation which aims to transfer knowledge from auxiliary domains to target domains has become an important way to solve the problems of data sparsity and cold start in recommendation systems. Cross domain recommendation aims to alleviate the sparsity problem in domains by transferring knowledge among related domains. This paper gives a brief survey of the various studies in this research line in two dimensions Collaborative Filtering Domains and Knowledge Transfer Styles and proves by making a comparison of various recommendation algorithms that the knowledge transfer can be better solutions to recommendation problem compared with other methods. Collaborative filtering and Knowledge Transfer remain the most used types of recommendation models, However as knowledge transfer enables to extract and transfer knowledge from some auxiliary sources in order to assist the learning task on the target data, these models perform well. As not much have been done in respect to comparing these models side by side. This paper evaluates most commonly used collaborative filtering models including User Similarity (CRUS), Nearest Neighbors (knn) and Matrix Decomposition (SVD) with Knowledge Transfer (CBT) based models to evaluate which model is best for CDRS. Experimental results on real data sets are drawn to find out the superior model which can perform well especially in the presence of cold start users in target domain.

Volume 11 | 10-Special Issue

Pages: 1146-1153

DOI: 10.5373/JARDCS/V11SP10/20192958