Interaction Data Analysis for Personalized Recommendation System

As the volume of information exploded, the recommendation system emerged as an effective way to address information overload. In particular, personalized recommendation systems based on user-item interaction data (e.g. click rate and purchase history) are used in most recommendation systems such as e-commerce, video streaming service, and social media. These interaction data have a significant impact on the computing resources as well as the performance of the algorithm. However, in most personalized recommendation system studies, the computing resource problem of interaction data has received relatively less scrutiny. In this paper, we show in detail the computing resource problems caused by interaction data and define them as the long-tail phenomenon. Also, to analyze the long-tail phenomenon, a social phenomenon, we propose the analytical method applying graph theory.