A Comparative Evaluation on Different Types of Recommender Systems

Dhiraj Khurana and Sunita Dhingra

Recommender is the effective data processor that observes various aspects of information for predicting the item, event or behavior. The user history, behavior and element specific characteristics can be observed under different phenomenon for optimizing the performance of recommender system. Content-based, collaborative and hybrid are the common-types of recommender systems. In this paper, an analytical evaluation of these recommender systems is provided for the Movielens dataset. This paper also explored the functional behavior of Correlation-based, Weighted correlation based and Frequency weighted correlation based collaborative filters. This paper has characterized these recommender systems with relative functional behavior. Experimentation is conducted on three different sample sets. Analytical observations are conducted using RMSE and MAE parameters. Analytical observations show that the hybrid recommender with correlation measure has achieved the most accurate and reliable results.

Volume 11 | 10-Special Issue

Pages: 1252-1259

DOI: 10.5373/JARDCS/V11SP10/20192970