HLGABM: Ensemble Generative Adversarial Network-Based Light-Gradient-Boosting Machine for Effective Injury Prediction

In sports, injuries are the most common occurrence, resulting in psychological, physical, and financial consequences. Therefore, predicting sports injuries in athletes is essential for their well-being and leading a successful sporting life. Although the existing deep learning and machine learning methods have achieved efficient results, the lack of sufficient generalization, poor interpretation, and other computational complexities make the prediction task more challenging. Therefore, this research proposes aHybrid Loss ensemble GenerativeAdversarial network-based light-gradient-boosting Machine (HLGABM) model for accurate injury prediction on athletes. In the proposed approach, the utilization of the HLGABM model addressed the existing challenges, including the computation complexity, low accuracy, overfitting, and poor generalization. Specifically, the application of the hybrid distance loss function in the HLGABM model enhances the overall performance, resulting in faster convergence, better generalization, and more robust training. Moreover, the model utilizes logarithmic data normalization and missing data imputation within the pre-processing stage, which eliminates the potential bias during computation. Further, the utilization of GAN based augmentation technique maximized the size and the diversity of the data. In line with this, the GAN architecture is ensemble with the HLGABM approach thatassisted in better efficiencyand accuracy. Further, the proposed HLGABMmodel achieves an accuracy of 96.89%, sensitivity of 96.29%, and specificity of 97.49%.