Adaptive Neuro-Fuzzy Inference System for Real-Time Health Monitoring and Sleep Optimization

Abstract

The study addresses the important problem of accurately predicting health status from physiological data, which is critical to healthcare monitoring. Patient data is inherently uncertain and variable, so medical diagnoses become complicated. Traditional diagnostic systems based on rigid thresholds may not have the capability to capture the complexity of dynamic physiological states. An alternative that promises to work well is fuzzy logic, which handles uncertainty and imprecision well. This work explores the use of fuzzy logic, combined with an Adaptive Neuro-Fuzzy Inference System (ANFIS), to classify health status based on key physiological parameters: heart rate, SpO2, and body temperature.  The dataset in this study comprises physiological data of 5 individuals collected over the course of 30 iterations. These inputs are used by the fuzzy model, converted into fuzzy membership functions, and processed by a set of fuzzy rules to provide a health status of Poor, Average, or Good. The achieved classification accuracy is 100%, which is confirmed by the high value of precision, recall, and f1 score and corresponding confusion matrix. The promising outcomes are limited by the small sample size and with a static dataset, which may not represent the wide range of actual health conditions occurring in real life. Moreover, the model is sensitive to input feature quality and selection. We believe that future research should expand the dataset to other conditions and explore hybrid models to improve robustness and generalizability. Although the findings are important because they demonstrate how fuzzy logic has a significant role to play in healthcare, they also signal a route to intelligent health monitoring systems that could be used to aid clinical decision-making.