Swords, Data and Balls: Extracting Extreme Behavioural Prototypes with Kernel Minimum Enclosing Balls

Extracting behavioural prototypes plays an important role in player profiling. Understanding the type of players present in the game goes alongside improving game-play experience as well as player engagement over time. In this paper, we introduce the application of Kernel Minimum Enclosing Balls (KMEBs) as a tool to extract meaningful extreme prototypes in games and present an example use-case analyzing a behavioural dataset from a Massively Multiplayer Online Role Playing Game. Unlike the majority of the methods covered in this context, our approach allows for modelling nominal and numerical behavioural features, extending the scope and capability of the profiling methods as well as improving the interpretability of the results.

[1]  Carlos Garćıa Márquez.  Multivariate Kernel Functions for Categorical Variables , 2014 .

[2]  Christian Bauckhage,et al.  Guns, swords and data: Clustering of player behavior in computer games in the wild , 2012, 2012 IEEE Conference on Computational Intelligence and Games (CIG).

[3]  Claus Bahlmann,et al.  Online handwriting recognition with support vector machines - a kernel approach , 2002, Proceedings Eighth International Workshop on Frontiers in Handwriting Recognition.

[4]  R. Tibshirani,et al.  Prototype selection for interpretable classification , 2011, 1202.5933.

[5]  Patrick Haffner,et al.  Support vector machines for histogram-based image classification , 1999, IEEE Trans. Neural Networks.

[6]  Claus Bahlmann,et al.  Learning with Distance Substitution Kernels , 2004, DAGM-Symposium.

[7]  Christian Bauckhage,et al.  Profiling in Games: Understanding Behavior from Telemetry , 2017 .

[8]  Rafet Sifa,et al.  Towards Structural Hyperparameter Search in Kernel Minimum Enclosing Balls , 2021, LION.

[9]  Nuno Vasconcelos,et al.  A Kullback-Leibler Divergence Based Kernel for SVM Classification in Multimedia Applications , 2003, NIPS.

[10]  Rafet Sifa,et al.  Joint Selection of Central and Extremal Prototypes Based on Kernel Minimum Enclosing Balls , 2019, 2019 IEEE International Conference on Data Science and Advanced Analytics (DSAA).

[11]  Bernard Haasdonk,et al.  Tangent distance kernels for support vector machines , 2002, Object recognition supported by user interaction for service robots.

[12]  Diego Klabjan,et al.  Guns and guardians: Comparative cluster analysis and behavioral profiling in destiny , 2016, 2016 IEEE Conference on Computational Intelligence and Games (CIG).