Abstract

Heart failure is one of the leading causes of death worldwide. Human survival, one of the functions the heart regulates, demands that the heart be safeguarded and aware of its danger. Therefore, the well-being of the heart determines whether humans will survive. A predictor tool is needed to address the issue of cardiovascular disease-related death events to reduce or prevent mortality and increase longevity. Data analytic techniques can be employed to identify patterns and relationships in patients' health records that are not currently recognised by clinical specialists. The predictive capabilities of Artificial Intelligence make it easy to accurately project future occurrences and results based on known records.

The aim of this study is to use machine learning to predict heart failure patient survival and to create a system that can predict heart failure patient survival using significant features The Heart Failure Clinical Records dataset, which includes data on 299 hospitalised patients, is used in this study to analyse heart failure survivors. Using the Random Forest algorithm to determine the importance of features, analysis of this data demonstrated that the most important features were Age, Ejection fraction, Serum creatinine, Serum sodium, and follow-up time. The ensemble learning approach was utilised to maximise the benefit of multiple model predictions. K-nearest Neighbor (KNN), Support Vector Machine (SVM), Logistic Regression (LR), and Decision Tree Classifier are the base models used (DTC).

The five most crucial features in the dataset were used to train the models. A hard voting ensemble approach was used, and the evaluation resulted in an accuracy score of 88%. Whereas the recall and f1 score for the deceased (Not alive) is lower than anticipated (62 and 74), this could be attributed to the relatively small number of instances in this category, but overall, the model's prediction is within the acceptable range. Simple software that predicts patient survival based on user input was created using the pre-trained voting classifier model.

Prediction of survival of heart failure patients using machine learning models