COMPARATIVE ANALYSIS OF CLASSIFICATION ALGORITHMS FOR EARLY PREDICTION OF CHRONIC KIDNEY DISEASE

Abstract

The need for information increases along with rapid technological advances. Accurate information is very important for predicting diseases in the medical field. Converting collected data into knowledge requires a specialized approach. This process is called data mining. Data mining is typically used to predict specific diseases based on patient medical records, particularly chronic kidney disease. The chronic kidney disease data were obtained from the Kaggle dataset. This study aims to examine the use, evaluate, and compare the performance of various classification algorithms, such as Gradient Boosted Trees, k-NN, Naive Bayes, Decision Tree, Random Tree, Generalized Linear Model, and Logistic Regression. The main objective is to find the algorithm with the highest AUC, accuracy, precision, and recall. The results of classifying the dataset using the Decision Tree and Generalized Linear Model algorithms showed outstanding performance, with the highest accuracy of 98.50%. On the other hand, the Naive Bayes and Generalized Linear Model algorithms achieved 100% recall, demonstrating an extraordinary ability to find all positive disease cases. Overall, these results provide important insights into the effectiveness of various data mining algorithms in predicting chronic kidney disease and lay the foundation for more accurate and reliable diagnostic systems in the future.