Heart disease diagnosis using deep neural network

Abstract

Heart disease is one of the leading causes of death, accounting for 16 percent of the world’s total deaths. For patients, early detection and timely treatment of heart disease are significantly important. Traditional methods of diagnosing heart disease, however, are both expensive and time consuming for low-income groups. Existing research shows that by analysing large amounts of complex healthcare data, the use of machine learning in medicine can improve efficiency and lower cost. The author wants to develop a model that reads patients’ physiological characteristics to predict if they have heart disease or not. This technique provides affordable and fast access to heart disease diagnosis for low-income people. The author uses the deep neural network as model and also uses 40,000 heart disease cases from the Centers for Disease Control and Prevention (CDC) as our data set. This data set contains 16 physiological features of patients. In the beginning, our model has the accuracy of 75.26%. After increasing the data set, our model has the accuracy of 76.63%.