A Multi-source Based Healthcare Method for Heart Disease Prediction by Machine Learning

Abstract

Accurate prediction of heart disease can save thousands of lives and de-crease health care cost significantly. In order to increase prediction accuracy-cy, we need to analyze data from multiple sources. However, current prediction methods based on machine learning do not consider the benefit of multiple sources. In this article, we combine four sensors with the electronic medical records (EMR), and perform feature extraction, preprocessing, feature fusion to predict heart disease by the support vector machines (SVM) and the convolutional neural network (CNN). The four sensors, including the medical sensor, the activity sensor, the sleeping sensor, and the emotion sensor use feature extraction techniques that are tailored for each sensor, considering their characteristics. Through analysis, it is demonstrated that the proposed method can increase the accuracy of heart disease prediction.