Kidney Stone Prediction Based on Urine Analysis: A Comprehensive Study of Machine Learning Models

Kidney stones, medically known as renal calculi, are solid crystal aggregations formed in the urinary tract, affecting millions globally. They are a common yet painful urological condition that can lead to severe health complications such as infections, urinary obstruction, and chronic kidney disease if not diagnosed and treated promptly. Identifying kidney stones early allows for more effective intervention strategies, minimizing patient discomfort and reducing healthcare costs. This paper explores a comprehensive evaluation of machine learning models for predicting kidney stone occurrence based on urine analysis. Using a dataset comprising 79 samples with six physical features such as specific gravity, pH, osmolarity, conductivity, urea concentration, and calcium concentration and ten machine learning models were evaluated: K-Nearest Neighbors, Support Vector Machine, Naive Bayes, Logistic Regression, Random Forest, Decision Tree, Adaboost, XGBoost, Gradient Boosting and Stacking. Logistic Regression achieved the highest accuracy of 0.93 with a precision of 0.87, demonstrating its suitability for such predictions. This research highlights the potential of ML based diagnostics in healthcare applications.