Early Risk Factor Prediction in Chronic Kidney Disease Diagnosis Using Feature Selection and Machine Learning...

Area of Science:

• Nephrology
• Biomedical Informatics
• Machine Learning

Background:

• Chronic kidney disease (CKD) involves irreversible kidney function decline.
• Early stages of CKD are often asymptomatic, complicating diagnosis.
• Accurate identification of CKD risk factors is crucial for timely intervention.

Purpose of the Study:

• To identify significant risk factors for CKD using feature selection techniques.
• To improve the predictive diagnosis of CKD using machine learning classifiers.
• To enhance patient care through earlier and more accurate CKD risk assessment.

Main Methods:

• Utilized a CKD dataset with 1,032 patient records and 14 features.
• Employed feature importance (tree-based) with Sequential Feature Selector (SFS) and ReliefF for risk factor identification.
• Trained and evaluated eight supervised and ensemble machine learning classifiers using cross-validation.

Main Results:

• Identified top 10 significant risk factors, with hemoglobin emerging as the most critical.
• Achieved high performance across classifiers: 86-98% accuracy, AUC > 0.96, precision 92-98%, recall 90-99%, F1 score 93-98%.
• Gradient boosting demonstrated superior performance in accuracy, precision, AUC, recall, F1 score, specificity, and bias.

Conclusions:

• Feature selection algorithms effectively identified key CKD risk factors.
• The proposed machine learning pipeline shows strong diagnostic performance for CKD risk.
• This methodology offers a promising approach for earlier and more accurate detection of CKD risk factors compared to conventional methods.