Wearable IoT (w-IoT) artificial intelligence (AI) solution for sustainable smart-healthcare

Abstract

Smart technologies, specifically wearables are cutting edge innovation of design science with an emerging Artificial Intelligence (AI) capability for sustainable healthcare. Wearable IoT (w-IoT) applications, solutions and systems can promote early warning measures for physiological parameter monitoring and other vital health observation while addressing, streamlining and enhancing emergency response procedures in the provision and deliverance of healthcare services. These solutions exhibit real-time responses with underlying machine-learning (ML) methodologies alongside ubiquitous, context-aware, pervasive and advance software features. AI frameworks, for the development and implementation of solutions are well covered in this study adopting design science (DS) principles for new product development (NPD), comprising various healthcare scenarios for distributed numbers and environments. Physiological or health activity-related data produced by embedded optical smartwatch sensors can instigate sustainable and economical health-oriented solutions for continuous monitoring, semantic predictions for constrained, intractable and autonomous environments to address cardiac disorders. This paper addresses, the practical implementation of the w-IoT health technology solution prototype for real-time applicability, covering problem identification and utilizing design science guidelines, evaluation and contribution by emphasizing on the experimental stage in general and with specificity. It covers performance results rendering research science communication on machine learning models for time series analysis, regression and classification to implement defined and adaptive thresholds, adopting standard deviation and moving average, computing mean square error (MSE), root mean square error (RSME) and mean absolute error (MAE) values, utilizing exponential moving average results on multiple features, prominently targeting resting heartrate data. Machine Learning algorithms for classification with higher F-score or performance metrics adopted are Decision Trees (DT), K-Nearest Neighbours (KNN), XGboost, One-class SVM and Logistic Regression. In Binary classification, KNN achieved F-score of 91 %, followed by DT at 81 % which seems an effective algorithm with flexibility on overfitting with high accuracy result. This study will cover all stages of design science methodology, guidelines for w-IoT healthcare solution development, by presenting experimental prototype towards pipeline implementation to address healthcare needs, alleviating previously prevalent Body Area Networks (BANs) solutions precision with advancing w-IoT smart technologies or Wireless Body Sensor Networks (WBSNs).