Abstract—Multi-agent based simulation (MABS) is an important approach for studying complex systems. The Agent-based model often contains many parameters, these parameters are usually not independent, with differences in their range, and may be subjected to constraints. How to use MABS investigating complex systems effectively is still a challenge. The common tasks of MABS include: summarizing the macroscopic patterns of the system, identifying key factors, establishing a meta-model, and optimization. We proposed a framework of experimental design and data mining for MABS. In the framework, method of experimental design is used to generate experiment points in the parameter space, then generate simulation data, and finally using data mining techniques to analyze data. With this framework, we could explore and analyze complex system iteratively. Using central composite discrepancy (CCD) as measure of uniformity, we designed an algorithm of experimental design in which parameters could meet any constraints. We discussed the relationship between tasks of complex system simulation and data mining, such as using cluster analysis to classify the macro patterns of the system, and using CART, PCA, ICA and other dimensionality reduction methods to identify key factors, using linear regression, stepwise regression, SVM, neural network, etc. to build the meta-model of the system. This framework integrates MABS with experimental design and data mining to provide a reference for complex system exploration and analysis.
Index Terms—Agent, complex systems, data mining, simulation, uniform design.
Fa Zhang is with the Beijing Institute of Technology, Zhuhai School, Zhuhai, China (e-mail: richter2000@163.com).
Shi-hui Wu and Zhi-hua Song are with the Equipment Management and UAV Engineering College at the Air Force Engineering University, Xi’an, China (e-mail: wu_s_h82@sina.com, szhele@163.com).
[PDF]
Cite: Fa Zhang, Shi-Hui Wu, and Zhi-Hua Song, "Framework of Experimental Design and Data Mining in Multi-agent Simulation," International Journal of Modeling and Optimization vol. 12, no. 1, pp. 1-7, 2022.
Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).
