| Summary: | Agent-based modelling (ABM) is increasingly being used as a way of understanding complex phenomena and systems. As traditional agent-based modelling relies on the designer's knowledge of the model, this thesis focuses on methods to improve the accuracy of traditional ABM using machine learning (ML). Firstly, a traditional ABM is developed using Counter-Strike: Global Offensive as the subject of study, followed by data mining using linear regression and k-mean clustering to develop an ML-based ABM. Finally, k nearest neighbours is used to measure the accuracy of the traditional ABM and the ML-based ABM. The results show that the traditional ABM achieve 78.8% accuracy and the ML-based ABM accuracy improve by 14.07% accuracy to 92.95%. This demonstrates the difficulty in achieving a high level of accuracy in the traditional way of modelling ABM, which will result in less accurate simulations of ABM. Machine learning can help researchers to better understand the characteristics of the target model in order to achieve a higher level of accuracy.
|
|---|