HEART DISEASE PREDICTION USING DATA MINING CLASSIFICATION ALGORITHMS

Abstract

A range of conditions which affect heart is called heart diseases or “cardiovascular diseases”. This disease can
bring out heart attack, chest pain, stroke etc. By reviewing some research paper related to heart disease
prediction it was identified that most of the paper using singular algorithm to predict the disease using
machine learning algorithm. Some of them indicates that they can’t use optimization techniques to improve
their model performance. For these results, they have faced some problem to predict heart disease in an
efficient manner by using their proposed system. To overcome these problems and for getting more accurate
results in this medical study is very crucial that’s why four different classification algorithms were
implemented to predict heart disease and find out the effectiveness of these algorithms. In this study Principal
Component Analysis (PCA) dimensionality reduction technique was applied which helped to get better results
with the aim of better accuracy by using these algorithms since medical diagnosis is sensitive. For this
approach data was collected from UCI repository which was found in Kaggle and it is named as “Heart
Disease UCI”. It was observed that without Principal Component Analysis (PCA) Logistic Regression
performed best to predict heart disease with Principal Component Analysis (PCA) k-Nearest Neighbors
(KNN) achieved greater accuracy compared to other classification algorithms. By applying PCA it was
identified that accuracy for other algorithms like Decision Tree and Naive Bayes also increased compared to
originally.