A Feature-Driven Decision Support System for Heart Failure Prediction Based on χ ² Statistical Model and Gaussian...

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Heart failure (HF) is a leading global cause of mortality.
• Existing intelligent decision support systems for HF detection often suffer from low prediction accuracy.
• Improving the accuracy of HF detection is crucial for patient outcomes.

Purpose of the Study:

• To develop a novel feature-driven decision support system to enhance heart failure prediction accuracy.
• To identify and utilize an optimal subset of features for improved HF detection.
• To evaluate the performance of the proposed system against existing methods.

Main Methods:

• A two-stage approach was developed, beginning with a chi-squared (χ²) statistical model to rank 13 common HF features.
• An optimal feature subset was selected using a forward best-first search strategy based on χ² test scores.
• A Gaussian Naive Bayes (GNB) classifier was employed as the predictive model in the second stage.

Main Results:

• The proposed chi-squared Gaussian Naive Bayes (χ²-GNB) method achieved a prediction accuracy of 93.33% on an online heart disease database of 297 subjects.
• The χ²-GNB model demonstrated a 3.33% improvement in HF prediction performance compared to the standard GNB model.
• The developed method outperformed existing literature methods, which reported accuracies ranging from 57.85% to 92.22%.

Conclusions:

• The feature-driven χ²-GNB system significantly enhances the accuracy of heart failure detection.
• This novel approach offers a more effective solution for identifying heart failure compared to current methods.
• The findings suggest the potential of this method for clinical application in improving HF diagnosis.