MACHINE LEARNING IN HEALTHCARE: PREDICTING CHRONIC KIDNEY DISEASE THROUGH FEATURE-DRIVEN HEURISTIC MODELS

Abstract

Chronic kidney disease (CKD) is amid the substantial backer to morbidity and mortality from non-infectious syndromes that can affect 10 to 15% of the universal inhabitants. Premature and accurate recognition of the stages of CKD is whispered to be vibrant to minimize the influences of patient’s healthiness complications, for instance hypertension, low blood count-anemia, bone disorder, poor nutritional vigor, neurological, and acid base abnormalities worries with timely intrusion through proper medications. Machine learning models can very well help clinicians to accomplish exactly this goal because of their prompt recognition performance. There have been different studies conducted based on machine learning methods in the identification of CKD during its premature phase. They did not pay more attention to the prediction of the given stage. In this research, the binary, and multi classification for phase prediction has been performed. The prediction models used include K-Nearest Neighbor ‘KNN’, Naïve Bayes ‘NB’, and Multilayer Perceptron ‘MLP’. An evaluation of the models was done and the results from the experiment indicated that KNN has a better performance of 99.17% than NB and MLP.