Project DetailsIn this work, predictive models were developed to identify patients at high risk for complications following hip and knee replacement surgeries, including blood transfusion, venous thromboembolism, acute kidney injury, and infections. The workflow involved extensive data preprocessing, conversion of medical codes into binary features, and addressing class imbalance through appropriate class weighting. Classical machine learning methods (Logistic Regression and Random Forest) as well as a Deep Neural Network (DNN) with advanced hyperparameter tuning were employed. Model performance was assessed using metrics such as AUC, Recall, Accuracy. All models achieved good and very close performance levels, with Logistic Regression occasionally slightly outperforming the DNN, and vice versa. Overall, the results were closely matched, underscoring both the potential and challenges associated with applying advanced machine learning techniques to medical prediction tasks.
