This study investigates the use of machine learning approaches, notably the Extra Tree Classification (ETC) model, Pelican Optimization Algorithm (POA), and Wild Geese Optimization (WGO) optimizers, to anticipate people's vulnerability to disc hernia. The primary goal is to evaluate the effectiveness of these models in correctly identifying patients at risk of developing a disc hernia. To achieve this goal, a dataset containing values for six biomechanical features was used to classify orthopedic patients into normal, disc hernia, or spondylolisthesis, and to train the ET-based models. The dataset includes clinically relevant spinal measurements that describe posture, pelvic alignment, and lumbar spine mechanics, which are widely used indicators in orthopedic diagnostics. Before training, the data were prepared and organized to ensure that the models could effectively learn patterns associated with spinal abnormalities. The ETC model paired with the POA (ETPO) yields the maximum accuracy. With an accuracy of 0.952, ETPO is the most reliable model tested. ETWG comes close behind, with a significant accuracy of 0.932. This makes ETWG the second-best predictor of disc hernia susceptibility. In addition to accuracy, the models demonstrate strong classification capability in distinguishing healthy individuals from patients with spinal disorders. The findings highlight the potential of optimization-enhanced machine learning techniques to improve diagnostic prediction in clinical settings. Overall, this research shows that machine learning models, especially the ETC method combined with optimization approaches such as WGO and POA, can effectively predict disc hernia susceptibility. It also provides valuable information on the relative efficacy of different models overall.
Bhuiyan et al. (2026) studied this question.