Chronic kidney disease diagnosis using decision tree algorithms

Area of Science:

• Nephrology
• Medical Informatics
• Machine Learning

Background:

• Chronic Kidney Disease (CKD) is a progressive, asymptomatic decline in renal function over months to years.
• CKD is staged based on Glomerular Filtration Rate (GFR), influenced by factors like age, sex, race, and serum creatinine.
• The CKD-EPI linear model is efficient for GFR estimation and CKD stage detection.

Purpose of the Study:

• To predict various stages of Chronic Kidney Disease (CKD) using machine learning classification algorithms.
• To develop a sustainable and practicable model for detecting CKD stages with high medical accuracy.
• To leverage machine learning for early symptom detection and diagnosis of CKD.

Main Methods:

• Utilized machine learning classification algorithms, specifically Random Forest and J48.
• Applied algorithms to a dataset derived from medical records of individuals with CKD.
• Compared the performance of J48 and Random Forest algorithms for CKD stage prediction.

Main Results:

• The J48 algorithm demonstrated superior performance in predicting all stages of CKD compared to Random Forest.
• J48 achieved an accuracy of 85.5% in detecting CKD stages.
• Comparative analysis confirmed J48's improved performance over the Random Forest algorithm.

Conclusions:

• The J48 algorithm shows significant potential for accurate CKD stage detection.
• An automated system for detecting CKD severity can be developed using these machine learning models.
• Early and accurate detection of CKD stages is crucial for preventing adverse health outcomes.