Increasing travel demand, complex traffic patterns, and unpredictable disruptions have made urban road network resilience a critical issue in traffic management. Signalized intersections, as key network nodes, significantly influence congestion formation, queue propagation, and performance degradation. Although full-scale intelligent control can improve traffic operations, it is not always feasible or optimal due to economic, technological, and implementation constraints. This study aims to identify the optimal combination of intelligent intersections by considering operational performance, environmental impacts, and implementation costs simultaneously.

A hybrid framework integrating traffic simulation, a Genetic Algorithm (GA), and a machine learning surrogate model is proposed. The studied network consists of six signalized intersections modeled in the SUMO environment. Each intersection’s control strategy was represented as a binary variable, where one indicates intelligent control and zero represents conventional fixed-time control. All 64 possible combinations were simulated, and a multi-objective function was developed based on travel delay, time loss, carbon dioxide emissions, fuel consumption, and implementation cost. The GA was then employed to optimize the control configuration, while the Gradient Boosting model was used as a surrogate model to efficiently estimate the resilience index and reduce computational costs.

The results showed that the fully intelligent scenario achieved the best performance in terms of operational and environmental indicators; however, its higher implementation cost prevented it from being the optimal solution. The configuration [1,1,0,0,1,1] was identified as the optimal scenario, providing performance close to the fully intelligent strategy with lower implementation costs. Moreover, the Gradient Boosting model accurately estimated the resilience index, achieving MAE=0.0248, RMSE=0.0462, and R²=0.927, while reducing the computational burden of scenario evaluation. The findings highlight that targeted intelligent control deployment can provide a cost-effective alternative to full network automation and support resilience-oriented urban traffic management.