Abstract:

Heart failure is a cardiovascular disease with significant morbidity and mortality, affecting an increasing number of people worldwide. It is very difficult to predict the survival of patients diagnosed with heart failure using clinical data. In this study, an intelligent machine learning (ML)-based artificial intelligence model is designed by identifying important features and an effective ML algorithm that will improve the accuracy of survival prediction of patients diagnosed with heart failure. For this purpose, the performance of seven different ML algorithms were compared regarding the extracted features from the dataset. Synthetic minority oversampling technique (SMOTE) and two different cross-validation techniques were used to remove the imbalance in the dataset and to obtain models with more realistic accuracy. Experimental results showed that SMOTE technique significantly improve the performance of classifiers in predicting the survival rate of patients diagnosed with heart failure. In the one-out cross-validation technique, the random forest algorithm achieved 90% accuracy, while in the 10-fold cross-validation technique, the extreme gradient boosting algorithm achieved the highest performance with 93% accuracy. When the results obtained are compared with the studies conducted in the literature used similar data sets, it is found that the proposed method exhibits higher performance. The proposed method has the potential to both improve the healthcare system and be an effective method for healthcare providers in predicting the survival of patients diagnosed with heart failure. In addition, it will guide specialist physicians in planning the treatment process of patients diagnosed with heart failure.

Keywords: