Data-Driven Optimization of Freight Train Operations on the Dammam–Riyadh Corridor Using Regression Modeling and Hybrid Intelligent Algorithms

This study addresses the high costs and emissions associated with diesel freight operations on the busy Dammam-Riyadh corridor by developing a hybrid, data-driven optimization framework that combines regression modeling, the Taguchi method, and a Genetic Algorithm (GA).First, a multiple linear regression model was trained on 30 real freight trips validated via 5-fold cross-validation and reporting R = 0.87 and RMSE = 3,200 SAR to predict total trip cost from six operational variables.Next, a Taguchi L9 orthogonal array was used to perform a sensitivity analysis under the "smaller-is-better" Signal-to-Noise (S/N) ratio, identifying wagon count and trip duration as the most influential factors, with a minimum predicted cost of 42,388.64SAR.Finally, we applied a DEAP-based GA (population = 50; generations = 100; blend crossover; Gaussian mutation) to globally optimize all six variables within empirically derived bounds, achieving a predicted cost of 34,054.33SAR ( 44% reduction versus the dataset mean).Key assumptions include linear cost relationships in the regression and fixed stop/truck counts during Taguchi screening; limitations stem from the single-corridor dataset.This combined approach balances rapid factor screening with precise global optimization, offering both strategic insights and actionable recommendations for reducing freight transportation costs while maintaining operational reliability.