This thesis explores the application of artificial intelligence in developing and optimizing marathon training plans. Leveraging a hybrid model combining Long Short-Term Memory (LSTM) networks with convolutional neural network (CNN) layers, this research aimed to harness detailed, sequential athlete data to generate personalized training strategies. An experimental study involving 100 athletes compared the effectiveness of AI-optimized training plans against traditional methods over six months. The findings indicate that AI-enhanced plans significantly improve race times and reduce injury rates. The model’s accuracy, precision, recall, and F1 score were rigorously evaluated, showing high reliability and the potential for precise adjustments to training regimens based on individual performance metrics. The successful implementation of this model demonstrates AI’s potential in transforming marathon training by providing data-driven, customized, and scientifically grounded training insights.