A Method of Distinguishing Abnormal Player Based on Steam User Profile for Game Disorder Risk Analysis

Abstract

In this paper, we proposed a rating index to determine the game-disorder risk of a game based on game attributes and player attributes. We collected 24,950 active user profile data for 20 games from a popular game platform called Steam. Using an unsupervised machine learning clustering approach -DBSCAN and game-disorder prevalence index as its threshold, we found that the game-disorder risk is not concerned with their popularity. However, it has more relation to game features such as "Open World", "Action" and "Multiplayer".