Optimizing Injury Detection with Autoencoder-Based Classifiers and Feature Selection

Many machine learning applications, such as injury detection systems, have made extensive use of autoencoders. For instance, it was suggested to use improved representative features in a deep autoencoder-based injury detection system to increase detection accuracy. Similarly, a feature selection based on the agricultural fertility algorithm was used to enhance injury detection systems, demonstrating the potential of feature selection techniques in improving detection performance. This study investigates the combination of autoencoder-based classifiers for injury classification and training. This method is used on the most significant feature chosen using the chi-square test (for binary values) and Pearson correlation (for continuous values). For the experiment, we have used the dataset. The study included 250 athletes, 150 of whom were women and 100 of whom were men. The average age of the study participants ranged from 18 to 22 years old. The quiz's response rate is 90.30%. The results of the trial show that the Injury Detection System outperforms previous studies and other classifier techniques, achieving a high classification accuracy of 92.27%.