Predict the Outbreak of Sudden Heart Failure in Dialysis Patients using Machine Learning

Abstract

The technological advancements in IoT, Artificial Intelligence, and Machine Learning enable modern and convenient patient monitoring opportunities for paramedics. The integration of machine learning into contemporary medical diagnosis platforms enables healthcare practitioners to diagnose potential heart failure in dialysis patients at an early stage. The medical situation alongside treatment in dialysis patients leads to elevated heart attack probabilities. Accurate predictions about these events remain a difficult task for paramedical staff who provide care during dialysis sessions. The time period brings numerous complications to patients characterized by blood pressure changes and heartbeat abnormalities, as well as temperature fluctuations and psychological challenges. To handle this problem, in the research, we use machine learning techniques to predict sudden heart failure early during the dialysis period by collecting data from dialysis patients and passing it to the machine learning model. We use a dataset from dialysis patients and then predict sudden heart failure in dialysis patients. We utilize logistic regression, KNN, Naïve Bayes, Decision Tree, Support Vector Machine, Artificial Neural Network, and XGBoost models to predict sudden cardiac arrest. We measure each model's accuracy, precision, Recall, and F-score. The results indicate that we achieve 73.9% accuracy for Logistics Regression, 88.9% accuracy for KNN, 94.1% accuracy for Decision Tree, 70.9% accuracy for Naïve Bayes, 83.9% accuracy for SVM, 95.6% accuracy for XGBoost, and 89.4% accuracy for ANN. Therefore, according to the final analysis, the XGBoost model predicts a higher incidence of sudden heart failure in dialysis patients.