Early Prediction of Heart Failure by Using Stacking Classifier

The misdiagnosis of heart failure is a significant factor contributing to higher mortality rate. Accurate and timely diagnosis is crucial for effective management and treatment of heart failure. By using their clinical records and lab test results, this study intends to create an accurate decision-making system for highly accurate survival prediction of patients with heart failure. The existing variability and imprecision in heart failure prediction necessitate the identification of major causal factors and precise risk assessment. To address this, a machine learning model is proposed using a heart failure dataset. The methodology involves applying the SMOTE algorithm to tackle class imbalance and implementing a stacked classifier for improved classification performance. Also employing L1 regularization for feature selection, resulting in a concise and informative feature set. These combined techniques significantly enhanced the model's accuracy, achieving a maximum accuracy of 90%, G-mean of 87.53%, sensitivity of 93.02%, ROC Score of 87.69%, and F1 Score of 8.35%. The proposed model outperforms the existing researches highlighting the potential utility of the proposed model as an effective tool for identifying high-risk patients and guiding appropriate medical interventions.