Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{197813,
        author = {AYESHA SAYYAD and RUSHIKESH NARESH BABAR and ADITYA RAJ and SHLOK KUMAR and SAHIL YADAV},
        title = {A Self-Adaptive Machine Learning Framework for Continuous Health Risk Assessment Using IoT Data Streams},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {6974-6979},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=197813},
        abstract = {Continuous health monitoring through Internet of Things (IoT) devices has revolutionized remote patient care by enabling real-time acquisition of physiological signals. Despite significant advancements, most machine learning (ML) models deployed in such systems remain static and assume stable data distributions. In practice, patient physiological conditions evolve over time due to disease progression, therapeutic interventions, and lifestyle modifications, resulting in distributional shifts commonly referred to as concept drift. This paper presents a self-adaptive machine learning framework engineered to maintain predictive reliability in dynamic healthcare environments. The proposed architecture integrates an Adaptive Windowing (ADWIN) drift detection module with an automated incremental retraining pipeline, enabling continuous self-calibration without manual intervention. The framework was evaluated on two benchmark healthcare datasets under simulated streaming conditions incorporating both gradual and abrupt drift scenarios. Experimental results demonstrate that the proposed adaptive system achieves a mean F1-score of 0.934, outperforming a static baseline by 14.2 percentage points. Mean post-drift recovery was observed at 3.2 batch intervals. These findings establish the framework as a robust, scalable solution for sustained predictive accuracy in IoT-enabled health monitoring systems.},
        keywords = {Adaptive machine learning, concept drift, continuous learning, health risk assessment, Internet of Things, IoT healthcare, remote patient monitoring, streaming data.},
        month = {April},
        }