Enhancing Chronic Kidney Disease Diagnosis using Machine Learning Classifiers: A Comparative Analysis

Abstract

One of the most common and dangerous illnesses affecting people on a global scale, chronic kidney disease (CKD), does not manifest itself until the kidneys of a particular person have sustained irreparable harm. The progression of CKD is linked to many serious side effects, such as an increased risk of different diseases, kidney failure, nerve harm, pregnancy problems, anemia, and hyperlipidemia. This illness claims the lives of millions of individuals each year. Since there are no significant symptoms that can be used as a benchmark to identify CKD, diagnosing the condition might be difficult. Occasionally, data may be interpreted wrongly when the diagnosis is persistent. To diagnose CKD in patients, this study employs a machine learning classifier. Six machine learning (ML) techniques are used in this study, including Random Forest (RF), Random Tree (RT), Decision Table (DTa), Decision Tree (DTr), Naïve Bayes (NB), and Hoeffding Tree and multiple performance metrics are considered such as accuracy, TPR, FPR, recall and mean absolute error (MAE). To select the most accurate classifier for predicting CKD, these predictive models are created using a dataset on chronic kidney disease containing 279 attributes acquired from Kaggle. Our objective is to ease the introduction of machine learning techniques for precisely detecting CKD by learning from dataset attribute reports. The main contribution of the research is an MLbased model for diagnosing chronic renal disease that outperforms common diagnosing techniques and reaches the highest predicted accuracy. This study also contrasted how well each model performed. We were able to predict this disease with the Random Forest model more accurately than ever before, at a 76.23% accuracy level.