Machine Learning Techniques for Heart Failure Prediction: An Exclusively Feature Selective Approach

The purpose of the work is to predict the risk of death due to a heart failure in the elderly. Even though, in today's age, when medical science has improved leaps and bounds to prevent death due to heart failure, there remains uncertainty among common people when it comes to this disease. Machine learning based models are used to predict the chances of a heart failure. A dataset with 12 different factors that determine the risk of having heart failure is used. SMOTE technique is applied to the imbalance dataset to create a balance and then feature engineering is applied to eliminate less important features and improve the performance of the model. The top seven features based Random Forest model is built to predict the chances of heart failure in a patient. Based on the selected features, there is a significant boost in the performance of the model with an accuracy of 90%. Also, there is an increase in the Precision, F1, Recall, ROC_AUC and Cohen kappa scores.