Chapter 9 - Class imbalance data handling with deep learning–based ubiquitous healthcare monitoring system using wearable devices

Abstract

Rapid development in the production of wearable devices has taken place for healthcare monitoring system. At the same time, the existence of imbalanced data poses a major influence on the prediction model, and many of the under sampling models require maximum time and decreased performance. In this view, this chapter develops a new class imbalance data handling (CIH) with optimal deep belief network (ODBN) model, named CIH-ODBN for ubiquitous healthcare monitoring system. To handle the class imbalance problem in healthcare data, adaptive synthetic sampling (ADASYN) technique is employed. Besides, the ODBN model is applied to determine the presence of diseases. In addition, rainfall optimization algorithm (ROA) is introduced to tune the hyperparameters of the deep belief network (DBN) model. An extensive implementation analysis was performed to signify the effectual outcome of the CIH-ODBN model. The outcome for the experimental validation verified the effective classification outcome of the CIH-ODBN model with the accuracy of 0.916 and 0.932 on the test diabetes and heart disease dataset.