17th International Conference on Advances in Computer Games (ACG), ELECTR NETWORK, 23 - 25 November 2021, vol.13262, pp.197-207
Modeling players based on their in-game events is essential for predicting their future behaviors. Player modeling studies mostly target a specific game or genre. This makes it difficult to transfer existing methods from one game to another. In this study, we propose a generic event-trait mapping and unsupervised learning approach for player modeling that extends our earlier modeling method with Principal Component Analysis (PCA). We present a case study of this new approach on a dataset of ten thousand players of World of Warcraft (WoW), a massive multiplayer online role-playing game (MMORPG). The base and the extended approaches are compared with an AutoEncoder (AE) based approach on this dataset. The methods generate clusters (persona) as mixtures of different character traits. The best results are obtained with the extended event-trait mapping approach for player modeling.