Profiling Players Using Real-World Datasets: Clustering the Data and Correlating the Results with the Big-Five Personality Traits

Abstract

Computer games provide an ideal test bed to collect and study data related to human behavior using a virtual environment having real-world-like features. Studies regarding individual players' actions in a gaming session and how this correlates with their real-life personality have the potential to reveal great insights in the field of affective computing. This study profiles players using data collected from strategy games. This is done by taking into account the gameplay and the associations between the personality traits and the subjects playing the game. This study uses two benchmark strategy game datasets, namely, StarCraft and World of Warcraft. In addition, the study also uses the Age of Empire-II game data, collected using 50 participants. The IPIP-NEO-120 personality test is conducted using these participants to evaluate them on the Big-Five personality traits. The three datasets are profiled using four clustering techniques. The results identify two clusters in each of these datasets. The quality of cluster formation is also evaluated through the cluster evaluation indices. Using the clustering results, the classifiers are then trained to classify a player, after a gameplay, into one of the two profiles. Results show that the gameplay can be used to predict various personality features using strategy game data.