A Hybrid Ensemble Approach with Feature Selection and Interpretability to Optimize Osteoporosis Risk Prediction

Osteoporosis is a prevalent disease characterized by reduced bone density and an increased susceptibility to fractures. Early detection is crucial for effective intervention. This study used a large dataset to construct a machine learning based framework for osteoporosis risk prediction. The preprocessing pipeline includes SMOTE normalization, scaling and class imbalance correction, followed by mutual information feature selection. On an 80:20 train-test split, several models and apply ensemble technique were trained and assessed, including random forest (RF), support vector machine (SVM), adaboost (ABC), extra trees classifier (ETC), decision tree (DTC) and XGBoost. Metrics like accuracy, precision, recall, F1-score and AUC score were used to evaluate the models. Ensemble of random forest and gradient boosting outperformed the other models, demonstrating its robustness in predicting osteoporosis risk with an AUC score of 83.05%. The model's interpretability was further improved by the incorporation of SHAP analysis which shed light on the contributions of important elements. This framework provides scalable solution for wider healthcare applications by showcasing the potential of machine learning to enhance osteoporosis prediction and help in making clinical decisions.