An Efficient Algorithm for the Calibration of Agent-Based Models Using Machine Learning

Abstract

A new algorithm for calibrating agent-based models is proposed, which employs a popular gradient boosting framework. Machine learning techniques are not used to develop a surrogate model, but rather assist in narrowing down the parameter space during the search for optimal parameters. Our approach is shown to achieve accuracy which compares favorably to the current state-of-the-art methods, while having a smaller computational cost.

Keywords: Agent-Based Models, Heterogeneous Agent Models, Model Calibration, Computation Time, Optimisation, Gradient Boosting, Machine Learning, XGBoost, Shapley Values, SHAP

JEL Classification: C02, C13, C15, C61, C63