Volume 09,Issue 01

Heart Disease Detection Employing Data Augmentations Using Machine Learning Algorithms with Model Tuning

Authors

Pardis Shahabinejad, Nazanin Hosseini

Abstract
Heart disease is a leading cause of death worldwide. Early detection of heart disease can be crucial in preventing serious health complications. Anomaly detection can help identify potential issues that may go unnoticed in large datasets. In this study, we employ data augmentations using machine learning algorithms for heart disease anomaly detection. We have defined a novel hybrid model which would apply various data augmentation techniques, such as oversampling, under sampling named as SMOTE, and feature scaling, to improve the performance of machine learning models. We compare the performance of different machine learning algorithms with 14 different methods, including decision tree, support vector machine, and random forest, for heart disease anomaly detection. The results show that data augmentations can effectively improve the accuracy of heart disease anomaly detection. Oversampling and feature scaling techniques are found to be effective in improving the performance of machine learning models. The logistic regression algorithm is found to be the most accurate in detecting heart disease anomalies with over 94% of accuracy with tuned model at last within 100 estimators’ selection. Our study demonstrates the potential of data augmentations using machine learning algorithms for heart disease anomaly detection, which can have significant implications for early detection and prevention of heart disease. Tuned proposed model showed the results with AUC score of 98.8% and an accuracy of 94.7%, which is higher than 88.8% accuracy and 94.4 AUC scores on non-tuned model.

Keyword: Heart Disease Prediction, Machine Learning, Data Augmentation, Classification, Anomaly Detection.

PDF [ 445.76 Kb ] | Endnote File | XML