Towards precision in IoT-based healthcare systems: a hybrid optimized framework for big data classification
Abstract The rapid expansion of IoT-enabled healthcare systems has resulted in the generation of large-scale, heterogeneous, and high-dimensional medical data. However, traditional machine learning models often struggle with issues such as data imbalance, redundant features, and noise, limiting thei...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Big Data |
| Online Access: | https://doi.org/10.1186/s40537-025-01243-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The rapid expansion of IoT-enabled healthcare systems has resulted in the generation of large-scale, heterogeneous, and high-dimensional medical data. However, traditional machine learning models often struggle with issues such as data imbalance, redundant features, and noise, limiting their reliability in clinical diagnostics. To overcome these limitations, this study proposes a hybrid framework that combines Lionized Heap Optimizer (LHO) for intelligent feature extraction with a Hierarchy Golden Eagle-Based Self-Constructing Neural Fuzzy (HGE-SNF) classifier for robust disease prediction. The proposed model was developed in MATLAB R2023b and tested using real-world datasets involving five disease categories: COVID-19, heart disease, diabetes, kidney disorders, and cancer. After preprocessing and feature selection, 114 key features were retained from an initial high-dimensional dataset exceeding 200 attributes, involving over 8000 total patient records. Performance evaluation was conducted using standard classification metrics, such as Accuracy, Precision, Sensitivity, Specificity, AUC, F-measure, MCC, and NPV. While the model achieved high scores across these metrics (e.g., AUC up to 0.995, Sensitivity and Specificity above 98%), we report these results conservatively and contextualize them with statistical validations, including cross-validation and multiple-run averaging. Comparative analysis against state-of-the-art models such as PSO-DNN, WOA-BRNN, and Adaptive E-Bat DBN is presented in detail, demonstrating the proposed model’s consistent superiority in classification performance and computational efficiency across diverse disease types. This research presents a scalable, interpretable, and resource-conscious solution for real-time medical data classification in IoT-based healthcare settings. Its modular design and adaptive behavior make it suitable for integration into edge-based clinical monitoring systems and decision support applications. KEYWORDS: Big data, Healthcare, Optimization, Classification, Feature Extraction and COVID-19 database. |
|---|---|
| ISSN: | 2196-1115 |