Machine Learning-Based Prediction of Chronic Kidney Disease with Logistic Regression

Chronic Kidney Disease (CKD) remains a critical health issue on a global scale, often progressing undiagnosed until advanced stages. In this research, a machine learning (ML)-based predictive system was developed to assess CKD risk with a Logistic Regression (LR) model. A dataset containing relevant medical attributes, such as blood urea levels, blood glucose readings, and indicators of conditions like anemia and diabetes, was utilized. Data preprocessing techniques, including cleaning, encoding, and feature selection, were applied to enhance model performance. The trained model achieved an accuracy of 97.5%, with precision and recall values exceeding 95%, ensuring high reliability in CKD prediction. A web-based application was integrated with the model using the Flask framework, allowing users to input medical test values and receive instant risk assessments. Comparative analysis with other ML algorithms revealed that Random Forest (RF) and Gradient Boosting (GB) exhibited slightly higher accuracy, but LR remained a competitive choice due to its simplicity and interpretability. The confusion matrix analysis indicated minimal misclassifications, confirming the model's effectiveness in real-world applications. By providing an accessible and user-friendly interface, early recognition and timely medical intervention were facilitated. The research demonstrated that ML significantly enhances disease prediction and proactive healthcare management. However, limitations such as dataset size and demographic diversity were identified, suggesting the need for further research. The findings emphasized that ML, when combined with web technologies, offers an efficient approach to CKD risk assessment, contributing to the advancement of predictive healthcare solutions.