Calibrating Agent-Based Models Using Behavioral Experiments
Scott Heckbert
2008 Spring Simulation Multiconference (SpringSim'08)- Poster Sessions (SCS-Poster sessions 2008)
Ottawa, Canada, April 14 - 17, 2008
Summary
Validating agent-based models (ABM) involves testing behavioural rules for simulated individuals against empirical data. However, validation is often difficult using individual-scale data because of data limitations, and cost of generating robust data sets. As a result, researchers commonly use a combination of theoretical assumptions and empirically-based functions. Empirical data for ABMs can be drawn from a range of sources and elicited by a variety of techniques. Specifically, this paper describes techniques using experimental economics and the calibration of agent behaviours in ABM. A method of using models interactively with human participants is proposed as a method for efficient and robust calibration of ABMs.
|
START
Conference Manager (V2.54.5)
|