Clustering Algorithms Analysis Based on Arcade Game Player Behavior

Abstract

The purpose of this study is to investigate the feasibility of using different clustering algorithms in grouping arcade game data for player behavior profiling. Using 3 clustering algorithms namely K-Means, Hierarchical Agglomerative Clustering, and DBSCAN, recorded game data for 6 games were clustered and the performance of each clustering algorithm was measured and compared. K-Means was shown to produce the highest quality and well formed clusters among all other algorithms used, and it also scored the highest on two of the evaluation metrics used. This study definitely answered the question regarding the utilization of different clustering algorithm with the use of arcade game data. Further studies are needed in order to generalize the idea of player profiling on games as a whole, with no regards in genres.