Optimized disease prediction in healthcare systems using HDBN and CAEN framework

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Classification and segmentation are crucial for healthcare, IoT, and edge computing but face challenges with accuracy and specificity on large, diverse datasets.
• Existing methods struggle to minimize false positives and negatives, limiting their real-world applicability.

Purpose of the Study:

• To develop a robust hybrid framework that enhances classification and segmentation performance.
• To address limitations in accuracy, precision, and specificity in current methodologies.

Main Methods:

• A hybrid framework combining a Hybrid Deep Belief Network (HDBN) for feature extraction and a Custom Adaptive Ensemble Network (CAEN) for dynamic prediction aggregation.
• Incorporation of an optimization mechanism for adaptability and robustness across varied datasets.

Main Results:

• Achieved 93% accuracy, 87% precision, 95% specificity, and 91% recall on four diverse datasets.
• Demonstrated high reliability with a Matthews Correlation Coefficient of 0.8932.
• The framework establishes a new benchmark for scalable, high-performance classification and segmentation.

Conclusions:

• The proposed hybrid framework offers a significant advancement in classification and segmentation, providing robust solutions for real-world applications.
• The framework's adaptability and performance pave the way for future integration with explainable AI and real-time systems.