Groundwater salinization is a critical issue in deltaic aquifers, particularly in semi-arid regions, due to seawater intrusion (SWI) and human activities. Effective Groundwater Vulnerability (GwV) mapping remains challenging in predicting salinization risks globally. This study integrates the GALDIT-NUTS framework with machine learning models-Random Forest Regression (RFR) and Generalized Linear Models (GLMs)-as well as hydrogeochemical analysis to address this challenge. Focusing on the Sharqia aquifer in Egypt's eastern Nile Delta, the study refines GwV indices derived from GALDIT-NUTS factors using a conditioned vulnerability index (CVI) based on electrical conductivity (EC) measurements. The RFR model (R-2 = 0.995, RMSE = 0.014) outperformed GLM (R-2 = 0.942, RMSE = 0.052) and the basic GALDIT-NUTS model, confirmed by Pearson correlation analysis (RF: r = 0.995, GLM: r = 0.924, basic: r = 0.603). The refined GALDIT-NUTS-RFR map accurately pinpointed high GwV areas affected by SWI, characterized by high TDS, Na+, Cl-, and Sr levels, and a low Na/Cl ratio, indicating Na-Cl type water. Moderately vulnerable zones, mostly in central areas, showed higher Na/Cl ratios, possibly linked to sewage or fertilizer inputs. The least vulnerable zones, located in southern regions, exhibited freshwater facies, with Ca-HCO3 water type and low Seawater Mixing Index (SMI). These findings are crucial for developing effective groundwater management strategies in stressed deltaic regions worldwide.