The complexity of addressing disruptive trends and disaggregated management strategies in transportation has led to the increased adoption of Agent-Based and Activity-Based modeling. However, broad implementation is hindered by the computational demands of calibrating intricate behavioral parameters in Activity-Based models. This paper introduces a novel Bayesian Optimization approach, utilizing an improved Random Forest surrogate model, to automate the calibration process and overcome these challenges. The proposed method sets a benchmark by calibrating the largest set of parameters yet, employing a sequential model-based algorithm configuration theory. Tested in Tallinn, Estonia, with 477 behavioral parameters, the calibration process yields satisfactory results for major indicators, including an OD matrix average mismatch of 15.92 vehicles per day and a 4% error in the overall number of trips.
Our Paper on 'Bayesian Optimization Approach for Calibrating Large-Scale Activity-Based Transport Models' is Now Available!
A_Bayesian_Optimization_Approach_for_Calibrating_Large-Scale_Activity-Based_Transport_M
.pd
Download PD • 3.85MB
