A Study on Detection of Chronic Renal Failure Based on Machine Learning

The most critical problem in the health community is the detection of kidney disease. In developing countries like Pakistan, Bangladesh, Nepal, Sri Lanka, and India either the government hospitals do not have enough facilities or private health centers are too expensive to give medical care for kidney patients. The primary objective of this research is to detect kidney disease at an early stage. A solution found on statistical analysis and machine learning, to predict renal failure is proposed. Sample data of 800 patients having an age group limit of 20-60 years are collected through a questionnaire from different hospitals in Punjab, Pakistan. A questionnaire has been designed to obtain a variety of data such as gender, age, diabetes, heart disease, hypertension, ESRD, eGFR rate, smoking, swelling of feet or ankles, anemia, insomnia, and kidney stones, etc. The contributions of this paper include different pre-processing techniques, including data normalization. Multiple machine learning algorithms were applied, including Decision Tree, KNN (K-Nearest Neighbor), SVC (Support vector Classifier) Neural Network, Naive Bayes, and Random Forest. Random forest, SVC, Decision tree, and Naive Bayes gives the accuracy of 96%, 91%, 93%, and 97%, respectively. Neural Network gives the best results with 98%accuracy.