Utilizing Machine Learning for Early Detection of Chronic Kidney Disease
Tanjim Mahmud, Md. Faisal Bin Abdul Aziz, Borhan Uddin, Avishek Majumder, Tahmina Akter, Nahed Sharmen, Mohammad Shahadat Hossain, Karl Andersson
Abstract
The study employed a range of machine learning and deep learning techniques to predict Chronic Kidney Disease (CKD) using clinical data containing 24 predictive parameters. CKD, characterized by abnormal kidney function or renal failure over an extended period, necessitates accurate early-stage prediction for effective management. Machine learning emerged as a robust tool for this purpose, exhibiting superior performance compared to deep learning and other methods. The methodology encompassed data preprocessing, addressing missing values, and feature extraction before applying various machine learning algorithms. These included K-nearest neighbor, Gradient Boosting, as well as deep learning models like CNN and ANN. The dataset comprised 400 individuals, with 250 diagnosed with CKD. Among the techniques evaluated, Gradient Boosting classification stood out as the most accurate, achieving a remarkable 97% accuracy in predicting Chronic Kidney Disease (CKD) status. This method streamlined feature selection, enabling the identification of crucial predictors while maintaining high predictive performance.