A Comparative Study of Mainstream Machine Learning Algorithms for Spine Health Prediction

Spine sub-health and spine-related diseases are common among modern people. The diagnosis and treatment of spinal diseases require doctors with extensive clinical experience, while machine learning can effectively and in large quantities predict spine health, thus assisting doctors in making decisions and reducing the burden of medical staff. In this study, we used seven mainstream machine learning methods - Logistic Regression (LR), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Naive Bayes (NB), Decision Tree (DT), Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), to construct classification models for spine health prediction, and compared the advantages and disadvantages of the seven models using multiple evaluation metrics in order to select the appropriate model for the issue of concern in practice. The results show that among the seven machine learning methods, Random Forest and XGBoost perform more outstandingly in each evaluation metric (accuracy and precision are higher than 0.9), while the K-Nearest Neighbor algorithm demonstrate superior performance (0.92) when AUC was used as the evaluation metric. These results suggest that the use of machine learning methods for spine health prediction has good prospects and that the most suitable algorithm can be selected according to our concerns.