This study presents a methodology for optimizing process parameters in complex polymer filling simulation model using neural networks, focusing on the critical roles of parameters like temperature, pressure, and cooling time. The neural network model is trained on historical data to predict the optimal settings, improving accuracy and reducing defects such as warpage and sink marks. Through this method, the optimization process becomes more efficient, minimizing defect rates and enhancing product quality. This approach not only improves injection molding processes but also highlights the potential for real-time optimization and adaptive control in complex manufacturing systems.