Strategy-Based Player Modeling during Interactive Entertainment Sessions by Using Bayesian Classification

Strategy-based player modeling is to recognize player´s strategy pattern during the gameplay per se. In this paper, Pac-Man game is used as a test-bed. Different Bayesian classifiers like naive Bayes and Bayesian net are chosen to analyze off-line data from gamers who are choosing different strategies, in other words the classifiers are trained with sample data from players using different strategies. The method attempts to use the constructed classifier to predict strategy type of a future player based on the data captured from its gameplay. This paper presents the basic principle of the strategy-based player modeling by using the Bayesian classification theoretic approach and discusses the results of the experiments. The hypothesis proposed in this paper is that Bayesian classification could be used as an approach with excellent performance to recognize player´s strategy pattern during real-time game genre.