Design of a Basketball Player Shooting Trajectory Prediction and Training Assistance System Based on LSTM Neural Network

With the current necessity of the correct prediction of shooting trajectories and individual training instructions in the training of basketball players in shooting, this paper presents a basketball shooting trajectory prediction and assisted training system using Long Short-Term Memory (LSTM) networks. To begin with, the multidimensional data of motion of the athlete in the shooting process are extracted and then their features are obtained. Second, a neural network model is built on the basis of LSTM, and the model is utilized to model the information that depends on time and identify the shooting path. The network parameters are optimized to reconstruct the trajectory of a ball with high precision by using a loss function. Lastly, the system integrates the probability of the prediction outcomes with the history of the performance of the athlete with the aim of offering customized training recommendations. The experimental findings indicate that the suggested method records an average MSE of 0.0191 and an average MAE of 0.106 on the shooting landing position which is far much better than the correlation between prediction and training.