AutoML Based IoT Application for Heart Attack Risk Prediction

Abstract

Machine Learning is used worldwide, for many applications, healthcare is also one such application. Machine learning can be crucial in determining whether or not there will be locomotors abnormalities, heart ailments, and other conditions. If foreseen far in advance, such information can offer crucial intuitions to professionals, who can then modify their detection and approach as per patient. The real-time data from IoT based healthcare systems is obtained, which is further processed to predict the potential risk of cardiovascular diseases. Developing a machine learning algorithm to predict potential heart diseases in people is focussed in this paper. In this paper, a comparative analysis of various classifiers, including decision trees, logistic regression, SVM, and random forests is conducted. Also, an ensemble classifier that performs hybrid classification by using both strong and weak classifiers is proposed because it can have a large number of training and validation samples. Current classifier and proposed new classifiers, such as Ada-boost, Grid Search CV is analysed. The accuracy outcome can be improved by using Auto ML. Auto ML provides its suggestions on which algorithm is to be used to get the best accuracy outcome. From the results obtained, it is evident that Auto ML can reduce the time taken to perform data analytics and it act s as a more accurate method. Along with Real-time IoT data observation and auto ML, the paper provides a complete aspect to a healthcare device.