Enhancing personalization in IoT-based health monitoring via generative AI and transfer learning
Rupali Atul Mahajan, Rajesh Dey, Mudassir Khan, Mazliham Mohd Su’ud, Muhammad Mansoor Alam, Pratibha Vijay Jadhav
Abstract
Owing to the rapid expansion of Internet of Things (IoT) devices, the health care sector is responsible for immense amounts of real-time data, which provides an impetus for custom health metrics. In this context, the current research seeks to fill this gap by proposing a groundbreaking system that employs generative AI technologies and transfer learning in the field of IoT-based health monitoring. Before examining the IoT health data, we must remove any potential discrepancies and errors through data cleaning. An adaptive filter referred to as the delayed error normalized LMS (DENLMS) is a highly sophisticated method that essentially contributes to increasing the precision and accuracy of these particular data. By applying analysis in the frequency domain to the data, we were able to extract features via the fast Fourier transform (FFT) and subsequently review sessions that contained, for example, heart rate variability or respiratory signals over time. The process of developing a generative AI model for personal health monitoring involves selecting suitable models, such as generative adversarial networks (GANs) or variational autoencoders (VAEs), owing to their ability to generate and simulate health data patterns effectively. To facilitate functional data analysis, the system design integrates machine learning techniques with generative models for patient data from various IoT devices. Importantly, the accuracy rate of this technique is 95.6%, the precision rate is 96.4%, the recall rate is 94.7%, and the F1 score is 95.5%. These metrics surpass those of most other techniques described in this study, demonstrating the superior performance of this research technique over other generic algorithms and its implementation with Python software. Future research could also focus on addressing the seemingly trivial challenge of enhancing model adaptability and scalability to meet individual health requirements and integrate multiple data sources.