What is the clinical evidence from this study?

Study design: Other. Population: Heart disease (n=294). Intervention: XGBoost machine learning model vs. Logistic Regression, Naïve Bayes, and Support Vector Machines. Primary outcome: Prediction accuracy.

September 29, 2022Open Access

Machine learning-based heart attack prediction: A symptomatic heart attack prediction method and exploratory analysis

Q: What does this research mean for the field?

XGBoost provides the best prediction for heart disease among evaluated machine learning models, achieving an accuracy of 92% and an area under the curve of 0.94. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Key Result

XGBoost provided the best prediction for heart disease among the evaluated machine learning models, achieving an accuracy of 92% and an area under the curve of 0.94.

Structured PICO

Population

294 patient records from the Cleveland dataset (UCI ML Repository) containing 14 attributes related to heart disease risk factors.

Intervention

Machine learning-based heart attack prediction using XGBoost, Support Vector Machines, Logistic Regression, and Naïve Bayes.

Comparator

Comparison between the different machine learning algorithms.

Outcome

Model performance evaluated by accuracy, precision, recall, and area under the curve (AUC).

Machine learning models, particularly XGBoost, can efficiently predict heart attack risk based on clinical parameters, achieving an AUC of 0.94.

Limitations

Overfitting can occur while training the model
Requirement for normalization of the dataset
Minimal accuracy achieved during evaluation of real-world problem-based data
Small dataset size

Abstract

Background; Heart attack prediction is one of the serious causes of morbidity in the world’s population. The clinical data analysis includes a very crucial disease i.e., cardiovascular disease as one of the most important sections for the prediction. Data Science and machine learning (ML) can be very helpful in the prediction of heart attacks in which different risk factors like high blood pressure, high cholesterol, abnormal pulse rate, diabetes, etc... can be considered. The objective of this study is to optimize the prediction of heart disease using ML. Methods: In this paper, we are presenting a machine learning-based heart attack prediction (ML-HAP) method in which the analysis of different risk factors and prediction for heart attacks is done using ML approaches of Support Vector Machines, Logistic Regression, Naïve Bayes and XGBoost. The data of heart disease symptoms has been collected from the UCI ML Repository and analysis has been performed on the data using ML methods. The focus has been on optimizing the prediction on the basis of different parameters. Results: XGBoost provided the best prediction among the four. The Area under the curve achieved with XGBoost is .94 and Logistic Regression is .92. The prediction with ML models in identifying heart attack symptoms is highly efficient, especially with boosting algorithms. The prediction was done to evaluate accuracy, precision, recall, and area under the curve. ML models are being trained to perform optimized predictions. Conclusions : This prediction can help clinically in analyzing the risk factors of the disease and interpretation of the patient scenario. Boosting the algorithm provided promising results to predict symptoms of heart disease. It can further be optimized by working further on risk factors associated with this condition.

KI fragen

Bookmark

View Full Paper