<p>In the present study, PVC membranes were fabricated using the NIPS method with Tween-20 as a pore-forming additive. The aim was to examine the effects of polymer and additive concentrations on membrane performance and to optimize fabrication parameters using machine learning. Previous applications of machine learning in membrane fabrication have largely relied on single-step or isolated optimization strategies and have lacked integrated hybrid frameworks for hyper parameter tuning and process optimization, particularly for surfactant-modified PVC systems. An ANN–GA–PSO pipeline was developed in which a genetic algorithm (GA) was first used to optimize the artificial neural network (ANN) architecture for flux prediction, followed by particle swarm optimization (PSO) to further refine the model parameters. Experimental data were generated using PVC and Tween-20 concentrations, resulting in 48 data points across varying pressures. The data set was split 80/20 into training and test sets. The machine learning models, optimized using GA and PSO, accurately predicted membrane flux, achieving a test MSE of 0.0138 and R2 of 0.9934, and a training MSE of 1.87 × 10⁻⁵ with R² of 0.99998. Model performance was further supported by a fivefold cross-validation mean R² of 0.995 ± 0.002. PSO-based optimization identified an ideal formulation consisting of 15.18 wt.% PVC and 6.0 wt.% Tween-20 at 0.46 bar, predicting a flux of 50 L/m².h; however, experimental validation of this optimum has not yet been conducted. Overall, this study demonstrates the potential of machine learning to refine membrane fabrication parameters and enhance performance for water purification applications.</p>