International Conference on Computer and Knowledge Engineering

Home / 14th International Conference on Computer and Knowledge Engineering

Efficient Prediction of Cardiovascular Disease via Extra Tree Feature Selection

Authors :

Mina Abroodi¹ Mohammad Reza Keyvanpour² Ghazaleh Kakavand Teimoory³

1- Data Mining Laboratory, Department of Computer Engineering Faculty of Engineering, Alzahra University Tehran, Iran 2- Department of Computer Engineering Faculty of Engineering, Alzahra University Tehran, Iran 3- Data Mining Laboratory, Department of Computer Engineering Faculty of Engineering, Alzahra University Tehran, Iran

Keywords :

cardiovascular disease prediction،Logistic Regression algorithm،Extra Tree،Feature Selection،E-health،Cleveland dataset

Abstract :

All abnormalities related to the heart can be classified as heart diseases. These diseases are also known as cardiovascular diseases. Predicting heart disease is crucial due to its status as a leading global cause of mortality. According to predictions, if this disease can be diagnosed earlier, treatment cost can be reduced, prompt initiation of care for patients and millions of lives can be saved. Utilizing machine learning and artificial intelligence for this purpose is paramount. Previous research has explored various methodologies such as KNN and SVM. However, the preprocessing phase significantly influences model performance, especially in e-health domains. This work proposes a groundbreaking method that uses the Extra Tree Classification method for feature selection, optimizing the selection of key attributes from the Cleveland Heart Disease dataset. Our method is chosen for the low computational cost and interpretability of feature selection through randomized scoring. Additionally, we employ logistic regression, renowned for its effectiveness and efficiency in classification tasks and to ensure the robustness of the model, we used the k-fold cross validation method. Our findings demonstrate a notable accuracy of 87.09% when predicting heart disease, with a precision of 87.45% and an F-measure of 86.02%, surpassing previous benchmarks. Considering the high interpretability of models like logistic regression, this level of accuracy is particularly valuable in medical contexts where transparency and explainability are critical, making it a competitive alternative to more complex, less interpretable methods. These findings highlight the effectiveness of combining Extra Tree-based feature selection with logistic regression in improving the predictive accuracy of heart disease diagnostics. Furthermore, we address missing values and their impact on model performance, further strengthening the reliability of our approach.