Source-linked AI summary
Ensemble Framework for Cardiovascular Disease Prediction
Achyut Tiwari, Aryan Chugh, Aman Sharma
TL;DR
The paper addresses the need for early and accurate cardiovascular disease prediction, particularly where diagnostic resources are limited. It develops a stacked ensemble using multiple machine-learning algorithms and reports 92.34% accuracy, higher than existing literature.
Problem
The paper seeks an early and accurate cardiovascular disease prediction system because cardiovascular disease is a major cause of death and diagnosis may be constrained by limited resources.
Method
The framework trains diverse machine-learning models and combines selected high-performing models through stacking with a meta learner.
Results
92.34% accuracy was achieved by the stacked classifier.
Takeaways & Limitations
The authors report that the framework outperforms existing state-of-the-art literature and can contribute to earlier cardiovascular disease prediction.
Abstract
from arXiv · showhide
Heart disease is the major cause of non-communicable and silent death worldwide. Heart diseases or cardiovascular diseases are classified into four types: coronary heart disease, heart failure, congenital heart disease, and cardiomyopathy. It is vital to diagnose heart disease early and accurately in order to avoid further injury and save patients' lives. As a result, we need a system that can predict cardiovascular disease before it becomes a critical situation. Machine learning has piqued the interest of researchers in the field of medical sciences. For heart disease prediction, researchers implement a variety of machine learning methods and approaches. In this work, to the best of our knowledge, we have used the dataset from IEEE Data Port which is one of the online available largest datasets for cardiovascular diseases individuals. The dataset isa combination of Hungarian, Cleveland, Long Beach VA, Switzerland & Statlog datasets with important features such as Maximum Heart Rate Achieved, Serum Cholesterol, Chest Pain Type, Fasting blood sugar, and so on. To assess the efficacy and strength of the developed model, several performance measures are used, such as ROC, AUC curve, specificity, F1-score, sensitivity, MCC, and accuracy. In this study, we have proposed a framework with a stacked ensemble classifier using several machine learning algorithms including ExtraTrees Classifier, Random Forest, XGBoost, and so on. Our proposed framework attained an accuracy of 92.34% which is higher than the existing literature.
1. Introduction
Cardiovascular disease is a major global health burden with several forms, while limited diagnostic resources can hinder timely and accurate detection.
- 32% of total deaths worldwide were caused by cardiovascular diseases in 2019.
- Limited access to doctors and diagnostic infrastructure can make cardiovascular testing unavailable or inaccurate in poorer nations.
- Cardiovascular diseases include coronary heart disease, heart failure, congenital heart disease, and cardiomyopathy.
2. Literature Review
Prior studies apply diverse machine-learning and ensemble methods to cardiovascular disease prediction, but many use small datasets; this work emphasizes broader evaluation and a larger dataset.
- Prior research used algorithms including logistic regression, K-nearest neighbours, Naive Bayes, support vector machines, decision trees, and ensemble methods.
- 90% was the highest accuracy reported for a majority-voting ensemble in the reviewed studies.
- Many previous studies used small datasets, whereas this work applies algorithms to 1190 instances with 11 attributes.
- 2.1. Our Contribution:: The proposed framework uses stacking-based ensemble learning, a large training dataset, and hyperparameter tuning.
3. Background & Preliminaries
The framework evaluates multiple classification methods, selects strong models, and combines them through ensemble learning for cardiovascular disease prediction.
- Ten classifiers were trained, after which four models were selected according to accuracy for final ensembling.
- Random forest builds many decision trees from random training samples and uses majority voting for the final classification.
- A multilayer perceptron is an artificial neural network with input, output, and one or more hidden layers.
- K-nearest neighbor classifies an instance using the majority class among its nearest K training points; K was set to 9 after GridsearchCV.
- XGBoost is a scalable gradient-boosting system designed for speed and performance with tree penalization and additional randomization settings.
4. Proposed Framework
The proposed framework preprocesses a merged cardiovascular dataset, selects strong classifiers through cross-validation, and combines them with a second-level meta learner using stacking.
- Data visualisation & Correlation of data attributes: Preprocessing analyzes feature correlations, removes outliers, and splits the data into 80% training and 20% testing partitions.
- Model Selection: Ten classifiers are trained, and four are selected after initial accuracy-based evaluation and 10-fold cross-validation.
- Data set: The framework uses IEEE Data Port data compiled from five cardiovascular datasets, with 11 features and 1,190 instances.
- Proposed Methodology: Stacking trains multiple base classifiers and uses a meta classifier to estimate model outputs and weights before selecting top-performing models.
5. Results & Discussion
The study evaluates cardiovascular disease classifiers using multiple performance metrics and compares baseline algorithms with a stacked classifier. The stacked classifier achieves the strongest reported overall performance across several measures.
- Performance evaluation: The evaluation covers accuracy, precision, sensitivity, specificity, F1 score, ROC, and MCC across the proposed and existing models.The section defines these measures and uses them to compare classifier performance.
- Performance metrics: Sensitivity measures correct positive predictions when illness is present, whereas specificity measures correct negative predictions when illness is absent.These correspond to true positive and true negative rates, respectively.
- Model selection: Ten-fold cross-validation was used to select high-accuracy baseline models for the stacking approach.Random Forest, Gradient Boosting Classifier, XGBoost, and Extra Tree Classifier were selected for stacking.
- Classifier comparison: 92.34% accuracy was achieved by the stacked classifier, exceeding the accuracy of the other evaluated classifiers.The stacked classifier also attained the highest reported F1 score and MCC.
- Classifier comparison: 92.74% F1 score and 84.64% MCC were reported for the stacked classifier.The stacked classifier also had 91.07% specificity and 93.49% sensitivity, while Random Forest had the highest sensitivity at 95.12%.
- Interpretation: Comparing diverse algorithms can support medical research and diagnosis by showing how models perform across cardiovascular diseases.The paper connects this diversity to potential use of the stacked approach in medical practice.
6. Conclusion and Future Work
The proposed framework is intended to support early cardiovascular disease prediction, particularly in settings with limited resources and growing populations. The authors report robust evidence and identify larger datasets with deep learning as future work.
- Contribution: The framework is proposed to help with early diagnosis and early prediction of cardiovascular disease.The stated application is within the healthcare domain.
- Contribution: The authors describe the framework's evidence and results as robust because of its stacked machine-learning approach.They also report that it outperforms existing state-of-the-art literature.
- Future work: Future work will explore the approach on larger datasets using deep learning concepts.This is the paper's stated future research direction.