Edge Computing-Enhanced Machine Learning Framework with Anomaly Detection-Aware Predictive Modelling in IoT Healthcare Systems

The implementation of IoT in healthcare systems allows for continued patient monitoring; however, issues such as latency, data privacy, and efficient anomaly detection arise. To solve these problems, this study presents an Edge Computing Enhanced Machine Learning Framework and an Anomaly Detection-Aware Predictive Modeling solution to expedite and effectively process patient data with security at the network edge. Anomaly detection is done using Z-score and DBSCAN methods, and the results show high detection accuracy using the mean absolute error (MAE) and root mean square error (RMSE) metrics: MAE of 0.1396 and RMSE of 0.3737 for the Z-score; and MAE of 0.0003 and RMSE of 0.0184 for the DBSCAN. For predictive modeling, Random Forest (RF) and Support Vector Machine (SVM) were used; RF offers a better approach to the anomaly classifications with 0.97 of precision, 0.95 of recall, and 1.00 AUC than SVM with 0.91 of precision, 0.77 of recall, and 0.97 AUC. These outcomes verify that the proposed framework is efficient in quickly detecting health abnormalities and estimating risky situations, which makes it perfect for such healthcare applications.