<p>The accurate prediction of longitudinal cracking in Continuously Reinforced Concrete Pavement (CRCP) is crucial for effective pavement management and maintenance strategies. This study explores the application of advanced machine learning techniques, particularly Support Vector Regression (SVR) optimized by Particle Swarm Optimization (PSO), to predict longitudinal cracking using a comprehensive set of predictor variables. These variables include structural factors, traffic loads, and climatic conditions. The optimal hyperparameters for the PSO-SVR model, identified through extensive optimization, significantly enhanced predictive performance. The model was evaluated using five-fold cross-validation, resulting in a mean Root Mean Square Error of 5.5313 and a mean R-squared value of 0.9426, outperforming not only traditional approaches like Linear Regression and Decision Trees, but also competitive machine learning baselines like SVR with Grid Search, Random Forest, Gradient Boosting, and Gaussian Process Regression. The convergence of the PSO algorithm illustrated efficient optimization, and the predicted versus measured values plot confirmed the model’s reliability. Further analysis included Partial Dependence Plots (PDPs) and 3D scatter plots, which provided insights into the relationships between Age and other key variables, highlighting their collective impact on longitudinal cracking. The study found that Age, traffic loads (Equivalent Standard Axle Loads and Annual Average Daily Traffic), and structural thicknesses are significant predictors of pavement performance. The results indicate that the PSO-SVR model is highly effective for predicting longitudinal cracking in CRCP, offering valuable insights for pavement engineers and researchers. This study demonstrates the potential of integrating advanced machine learning techniques with optimization algorithms to improve pavement performance prediction models, guiding more efficient design, maintenance, and rehabilitation strategies.</p>