PREDICTION OF SURVIVAL OF HEART FAILURE PATIENTS USING RANDOM FOREST

Abstract

Human survival, one of the roles that is controlled by the heart, makes the heart need to be guarded and be aware of its damage. Heart failure is the final stage of all heart disease. The medical record tool can measure symptoms, body features, and clinical laboratory test values, which can be used to perform biostatistical analyzes but to highlight patterns and correlations not detected by medical doctors. So technology assistance is needed to do this in order to predict the survival of heart failure patients. With data mining techniques used in the available history data, namely the Heart Failure Clinical Records dataset of 299 instances on 13 features used the Random Forest algorithm, Decision Tree, KNN, Support Vector Machine, Artificial Neural Network and Naïve Bayes with resample and SMOTE sampling techniques. The highest accuracy with the resample sampling technique in the random forest is 94.31% and the SMOTE technique used in the random forest produces an accuracy of 85.82% higher than other algorithms.

Keywords

• smote
• random forest
• data mining
• heart failure
• resampling