Effective clinical decision support implementation using a multi filter and wrapper optimisation model for Internet...

Area of Science:

• Medical Informatics
• Computer Science
• Data Science

Background:

• Internet of Things (IoT) in healthcare generates vast, complex data.
• Feature selection (FS) is critical for accuracy, efficiency, and interpretability in IoT-based Clinical Decision Support Systems (CDSS).
• Data redundancy and noise negatively impact system performance, necessitating effective FS.

Purpose of the Study:

• To propose a novel two-phase feature selection model for IoT-based CDSS.
• To improve the accuracy, efficiency, and interpretability of healthcare decision-making systems.
• To address challenges of data volume, redundancy, and noise in healthcare data.

Main Methods:

• A two-phase FS model combining Filter Methods (FM) and Wrapper Methods (WM).
• Phase I: Ensemble of five FMs followed by Pearson Correlation Method (PCM).
• Phase II: Binary Optimized Genetic Grey Wolf Optimization Algorithm (BOGGWOA) as WM, utilizing Support Vector Machine (SVM) for accuracy prediction.

Main Results:

• The proposed model effectively integrates valuable features from filter methods.
• Pearson Correlation Coefficient (PCC) is used to eliminate irrelevant features.
• BOGGWOA optimizes feature selection for enhanced classification accuracy (CA).

Conclusions:

• The developed two-phase FS model is critical for practical and dependable CDSS in IoT healthcare.
• This approach enhances decision-making system quality by reducing computational overhead and improving interpretability.
• The integration of filter and wrapper methods, including BOGGWOA, offers a robust solution for healthcare data analysis.