Abstract—This study tackles the challenge of image recognition in badminton due to its small size and complex movement trajectory by developing a new dataset and predicting shot sequences. The experimental procedure involves two main steps: detailed pre-processing of original data (extracting key features, normalizing, and dividing into training and test sets) and using deep learning algorithms to predict future shots. We validated the model with 20 repeated experiments and multiple evaluation metrics (Accuracy, Precision, Recall, F1-Score) and confirmed statistical significance and stability through T-tests. The results demonstrate that the Gated Recurrent Unit (GRU) effectively predicts shot sequences, benchmarking with Long Short-Term Memory (LSTM) and One-Dimensional Convolutional Neural Networks (1D-CNNs). The proposed architecture model with joint Rectified Linear Unit (ReLU) achieved 59.77% accuracy compared to a random guessing baseline of 16.67%, showcasing superior performance and promising applications. Therefore, this research provides a new tool for badminton data analysis and lays the groundwork for broader sports data analysis and behavior prediction.
 
Keywords—sports analytics modelling, time-series analysis, sequential structured dataset, supervised learning


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).