Deep Learning Technique to Predict Heart Disease using IoT Based ECG Data

Abstract

Scientific Knowledge and Electronic devices are growing day by day. In this aspect, many expert systems are involved in the healthcare industry using machine learning algorithms. Deep neural networks beat the machine learning techniques and often take raw data i.e., unrefined data to calculate the target output. Deep learning or feature learning is used to focus on features which is very important and gives a complete understanding of the model generated. Existing methodology used data mining technique like rule based classification algorithm and machine learning algorithm like hybrid logistic regression algorithm to preprocess data and extract meaningful insights of data. This is, however a supervised data. The proposed work is based on unsupervised data that is there is no labelled data and deep neural techniques is deployed to get the target output. Machine learning algorithms are compared with proposed deep learning techniques using TensorFlow and Keras in the aspect of accuracy. Deep learning methodology outfits the existing rule based classification and hybrid logistic regression algorithm in terms of accuracy. The designed methodology is tested on the public MIT-BIH arrhythmia database, classifying four kinds of abnormal beats. The proposed approach based on deep learning technique offered a better performance, improving the results when compared to machine learning approaches of the state-of-the-art.

Related Papers

• → A Tour of Unsupervised Deep Learning for Medical Image Analysis(2021)94 cited
• → Semi-supervised Learning(2020)30 cited
• → Self-Supervised Learning for Large-Scale Unsupervised Image Clustering(2020)
• → Analysis of recent advancement in unsupervised deep learning(2022)
• → Deep-learning-based raw data generation for deep-learning-based image reconstruction(2023)