Study on Optimal Machine Learning Approaches for Kidney Failure Detection System Based on Ammonia Level in the Mount

Abstract

Patients with kidney failure can emit bad breath with a certain level of ammonia content. The interpretation of ammonia levels in patients with renal failure was identified in this work. This research used the Chronic Kidney Disease dataset from UCI Machine Learning Repository. The dataset was processed first to get the value of eGFR (estimated Glomerular Filtration Rate) and ppb (parts per billion) of ammonia. Based on the eGFR value, the severity of kidney failure was divided into 5 categories, namely normal (Stage 1), mild (stage 2), moderate (stage 3), severe (stage 4) and failure (stage 5). The values of eGFR features are used as input for the machine learning technique in order to predict the level of kidney failure. Four different types of machine learning techniques, namely Support Vector Machine (SVM), Naïve Bayes (NB), Artificial Neural Network (ANN) and K-nearest neighbors (KNN), are applied and compared. The last process was to identify kidney failure using the K-Nearest Neighbors (KNN) method based on eGFR and ppb of ammonia dataset. AI-based kidney failure severity identification system with KNN algorithm had an average accuracy of 89.9% and 95.65% for training and testing accuracy, respectfully.