Machine Learning Paradigm for Internet of Things Applications

7Offline and Online Performance Evaluation Metrics of Recommender System: A Bird’s Eye View

R. Bhuvanya* and M. Kavitha

Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, India

Abstract

Recommender system (RS) plays a major role in e-commerce sites and social media. In today’s world, the usage of data has increased with a large database, and it is becoming a difficult task for people, to find a relevant item of their interest. So, the RS helps the people to find the most relevant item from the available database, and there exist various algorithms for prediction and recommendation. To determine the effectiveness of algorithms in RS, there are numerous ways to evaluate the goodness of recommendation and a recommender algorithm. The efficiency of the recommendation can be evaluated through various performance evaluation metrics. This paper analyzes various evaluation methods and metrics and demonstrated the ranking prediction and top-N movie recommendation for the users with different algorithmic approaches which will help the designers to understand the ways to choose a good recommendation algorithm under various scenarios. In addition to the above, this paper presents the accuracy calculation of our system based on real-world data.

Keywords: Evaluation metrics, offline, online, recommender system, user study, SVD, SVD++

7.1 Introduction

The influence of the recommender system (RS) has started in the mid-1990s, and it is still active in the ...