[Paper Review] Coronary Heart Disease Diagnosis Based on Improved Ensemble Learning

Kuntoro Adi Nugroho, Noor Akhmad Setiawan+1 more|arXiv (Cornell University)|Jul 6, 2020

Artificial Intelligence in Healthcare40 references4 citations

TL;DR

This study proposes a hybrid cascade ensemble learning method for coronary heart disease diagnosis, combining C4.5 and RIPPER as meta-learners with Naive Bayes as the base learner, using bagging and random subspace for ensemble construction. The cascade generalization with loose coupling strategy significantly improved diagnostic accuracy compared to standalone algorithms and Rotation Forest, though diversity among base models decreased.

ABSTRACT

Accurate diagnosis is required before performing proper treatments for coronary heart disease. Machine learning based approaches have been proposed by many researchers to improve the accuracy of coronary heart disease diagnosis. Ensemble learning and cascade generalization are among the methods which can be used to improve the generalization ability of learning algorithm. The objective of this study is to develop heart disease diagnosis method based on ensemble learning and cascade generalization. Cascade generalization method with loose coupling strategy is proposed in this study. C4. 5 and RIPPER algorithm were used as meta-level algorithm and Naive Bayes was used as baselevel algorithm. Bagging and Random Subspace were evaluated for constructing the ensemble. The hybrid cascade ensemble methods are compared with the learning algorithms in non-ensemble mode and non-cascade mode. The methods are also compared with Rotation Forest. Based on the evaluation result, the hybrid cascade ensemble method demonstrated the best result for the given heart disease diagnosis case. Accuracy and diversity evaluation was performed to analyze the impact of the cascade strategy. Based on the result, the accuracy of the classifiers in the ensemble is increased but the diversity is decreased.

Motivation & Objective

• To enhance the accuracy of coronary heart disease diagnosis using machine learning.
• To investigate the impact of cascade generalization and ensemble learning on diagnostic performance.
• To compare the proposed hybrid cascade ensemble method with non-ensemble and non-cascade baseline models.
• To evaluate the trade-off between accuracy and diversity in the ensemble classifier.
• To demonstrate the effectiveness of loose coupling strategy in cascade generalization for medical diagnosis.

Proposed method

• The study employs a cascade generalization framework with a loose coupling strategy to integrate multiple base learners.
• Naive Bayes is used as the base-level learning algorithm for individual classifiers.
• C4.5 and RIPPER are applied as meta-level algorithms to combine predictions from base models.
• Bagging and Random Subspace are used to construct diverse and robust base ensemble members.
• The hybrid cascade ensemble method integrates predictions through a two-level learning architecture to improve generalization.
• The method is evaluated against non-ensemble models and Rotation Forest for comparative performance analysis.

Experimental results

Research questions

• RQ1 Does the proposed cascade ensemble method improve diagnostic accuracy for coronary heart disease compared to individual learning algorithms?
• RQ2 How does the use of loose coupling in cascade generalization affect the performance of the ensemble?
• RQ3 What is the trade-off between accuracy and diversity in the proposed ensemble model?
• RQ4 How does the hybrid cascade ensemble compare to Rotation Forest in diagnosing coronary heart disease?
• RQ5 To what extent do bagging and random subspace contribute to the robustness of the ensemble?

Key findings

• The hybrid cascade ensemble method achieved the highest diagnostic accuracy among all evaluated methods for the given coronary heart disease dataset.
• The accuracy of individual classifiers in the ensemble increased due to the cascade generalization strategy.
• Despite improved accuracy, the diversity among base classifiers decreased, indicating a trade-off between performance and ensemble diversity.
• The proposed method outperformed both non-ensemble models and Rotation Forest in terms of classification accuracy.
• The use of bagging and random subspace contributed effectively to the construction of diverse and stable base learners.
• The results confirm that cascade generalization with loose coupling enhances generalization ability in medical diagnosis tasks.