Hyperparameter Tuning on Classical Machine Learning Models in Orthopedic Disease Prediction on Biomechanical Features

Using machine learning in healthcare is increasingly becoming an advanced method for predicting and treating diseases early. The significant increase in orthopedic diseases has made early disease detection more crucial than ever, allowing for more effective disease prevention. This study aims to support healthcare professionals in the early prediction and classification of orthopedic diseases. To achieve this goal, we used data visualization methods to analyze the data and assess visualizations using statistical results from charts. Subsequently, machine learning methods, including Random Forest, Logistic Regression, k-Nearest Neighbor, and LightGBM, were applied to a dataset containing information on 310 patients, comprising six biological features describing each patient’s pelvic status and spine. The results of these algorithms were then compared, with Logistic Regression considered the algorithm that yielded the best performance, achieving an accuracy of up to 87%. In contrast, other algorithms ranged from 85% and above.