Study of gamers’ behaviours through game data mining and analytic

• Main Author: Ngoh, Vincent Chang Chiat
• Other Authors: Ong Yew Soon
• Format: Final Year Project
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/59217
• Institution: Nanyang Technological University

The current trend in gaming is moving from single player gaming to multiplayer gaming. This is seen as a threat to the games industry, as multiplayer games yield less revenue by game hour as compared to single player games. As a result, games developers are moving onto AI agents’ development to create a more believable and challenging single player experience, in hope of attracting and retaining customer base. In this project, investigation on mining of gamers’ behaviours in terms of game tactic in the game StarCraft II will be conducted. Association rule mining will be applied on the data collected from professional gamers to mine for rules associated to rush tactic used in the game. Rules mined will then be compared against game guides and online resources for their accuracy. Finally, a discussion on how AI agent will benefit from the data mined will also be conducted. The final result shows up to 80% accuracy in the rules mined in terms of classifying different itemset to the tactic used by the players, thus proving it is possible to mined and understand gamers’ behaviours in game. It is concluded that if given enough data, it will be possible for an agent to improve game play over time and predict players’ tactics in a game, thus making the game more believable and challenging to human player.