An Interpretable Classification Method Based on User Behaviour in Online Game

Abstract

For online game management, it is important to understand user behavioral characteristics. It is possible to increase user satisfaction for games by classification based on their behavioral characteristics and designing optimal game contents for each cluster having different types of motivation for games. We propose an interpretable classification method by using principal component analysis, K-means, and decision trees. As a result of applying the proposed method to actual in-game behavior data, we found user retention rate or Average Revenue Per User were different between the clusters. Furthermore, we confirmed that it could classify interpretable clusters and designed concrete game contents for each cluster.