AbstractProbabilistic and nonlinear models have been used to accurately model various phenomena in asset management systems (AMS). With a commonly adopted framework using Monte Carlo simulation and heuristic algorithms, AMS proposed in the literature aim to maintain the functionality of assets in their life-cycle by optimally allocating limited resources to different intervention actions. However, due to their high computational costs, upscaling complex project-level AMS to a multitude of assets currently is far from practical. To address this gap between the literature and the practice of project-level AMS, this paper presents a new machine learning–based methodology to estimate (near-)optimal intervention timings which usually are derived by optimization algorithms. To illustrate, an ensemble of random forests models was trained on optimal maintenance timings of more than 1.6 million semisynthesized bridges. The trained model yielded optimized maintenance, rehabilitation, and reconstruction (MRR) plans with greater than 95% accuracy on the test set and greater than 89% accuracy on more than 4,600 highway bridges in Indiana, and did so 6 orders of magnitude faster than the conventional framework of complex MRR optimization. Practitioners can adopt the proposed methodology to enhance their decision-making systems, obtain optimal maintenance plans without sacrificing complex and accurate models, and take another step toward sustainability objectives.