Deep neural network model for enhancing disease prediction using auto encoder based broad learning.

Area of Science:

• Bioinformatics
• Healthcare Informatics
• Machine Learning in Medicine

Background:

• Big Data has transformed disease prediction models, enabling early illness detection.
• Deep neural networks offer high accuracy but face challenges like gradient instability and slow training.
• Traditional Broad Learning (BL) excels in incremental learning but struggles with complex feature extraction in healthcare.

Purpose of the Study:

• To address the limitations of existing models in complex healthcare data analysis.
• To introduce a hybrid model, Adaptive Broad Learning (ABL), for improved disease prediction.
• To enhance feature extraction capabilities in intricate medical environments.

Main Methods:

• Developed ABL by integrating the Broad Learning system with Denoising Autoencoders (AE).
• Employed incremental learning strategies to avoid gradient descent and accelerate training.
• Focused on robust feature extraction from complex medical datasets.

Main Results:

• The ABL model demonstrated superior performance in extracting complex features from medical data.
• Achieved a high prediction accuracy of up to 98.50% across various datasets.
• Validated the model's effectiveness in intricate healthcare scenarios.

Conclusions:

• ABL offers a robust solution for disease prediction in complex healthcare settings.
• The model's adaptive nature and high accuracy support its application in clinical decision-making.
• ABL provides an agile and accurate approach to forecasting diseases, overcoming limitations of prior methods.