Research Article
Deep Learning for Classifying Battlefield 4 Players
@INPROCEEDINGS{10.1007/978-3-319-49616-0_15, author={Marjolein Vries and Pieter Spronck}, title={Deep Learning for Classifying Battlefield 4 Players}, proceedings={Intelligent Technologies for Interactive Entertainment. 8th International Conference, INTETAIN 2016, Utrecht, The Netherlands, June 28--30, 2016, Revised Selected Papers}, proceedings_a={INTETAIN}, year={2017}, month={1}, keywords={Deep learning Computer games Player classification}, doi={10.1007/978-3-319-49616-0_15} }- Marjolein Vries
Pieter Spronck
Year: 2017
Deep Learning for Classifying Battlefield 4 Players
INTETAIN
Springer
DOI: 10.1007/978-3-319-49616-0_15
Abstract
In our research, we aim to predict attributes of human players based on observations of their gameplay. If such predictions can be made with sufficient accuracy, games can use them to automatically adapt to the player’s needs. In previous research, however, no conventional classification techniques have been able to achieve accuracies of sufficient height for this purpose. In the present paper, we aim to find out if deep learning networks can be used to build accurate classifiers for gameplay behaviours. We compare a deep learning network with logistic regression and random forests, to predict the platform used by Battlefield 4 players, their nationality and their gaming culture. We find that deep learning networks provide significantly higher accuracies and superior generalization when compared to the more conventional techniques for some of these tasks.
