This study introduces a novel Deep Marketing Mix Model (DMMM) leveraging machine learning (ML) and deep learning (DL) to overcome the limitations of traditional methods in handling granular data. By employing ML/DL with derivative-free optimization, DMMM enables accurate daily budget allocation across media channels. Comparative analyses show that DMMM achieves superior performance in prediction accuracy, budget allocation efficiency, and computational speed compared to competing models. This research makes a significant contribution to marketing analytics in academia and industry.