A Detailed Analysis on Kidney and Heart Disease Prediction using Machine Learning

Abstract

Detection of disease at earlier stages is the most challenging one. Datasets of different diseases are available online with different number of features corresponding to a particular disease. Many dimensionalities reduction and feature extraction techniques are used nowadays to reduce the number of features in dataset and finding the most appropriate ones. This paper explores the difference in performance of different machine learning models using Principal Component Analysis dimensionality reduction technique on the datasets of Chronic kidney disease and Cardiovascular disease. Further, the authors apply Logistic Regression, K Nearest Neighbour, Naïve Bayes, Support Vector Machine and Random Forest Model on the datasets and compare the performance of the model with and without PCA. A key challenge in the field of data mining and machine learning is building accurate and computationally efficient classifiers for medical applications. With an accuracy of 100% in chronic kidney disease and 85% for heart disease, KNN classifier and logistic regression were revealed to be the most optimal method of predictions for kidney and heart disease respectively.