Enhanced Feature Selection and Extraction for Ensemble Machine Learning-based Classification of Heart Disease based on ECG

Heart disease is still a significant global health concern that necessitates precise and efficient categorization techniques for early identification and therapy. This paper proposes a novel approach for optimizing feature selection and extraction in the context of electrocardiogram (ECG)-based cardiac disease classification, utilizing the potential of ensemble machine learning. The suggested method begins with the X2 optimal feature selection strategy, which finds the most insightful ECG features by combining statistical techniques and subject-matter knowledge. This reduces the dimensionality of the input space and strengthens the model's resilience to redundant and pointless inputs. A unique feature extraction method that successfully extracts intricate patterns and minute fluctuations present in the ECG signals is then used to turn the chosen ECG features into a higher-level representation. A machine learning ensemble strategy is used to optimize classification performance. A weighted combination of several classifiers, including Random Forest, Naive Bays, and Support Vector Machines, is used to combine their various strengths and generalization capabilities. A dependable and effective classification system for heart illness is produced as a result of the ensemble technique, which increases overall accuracy, sensitivity, and specificity. The maximum accuracy of 97.84% and 98.44% was attained by combining the grouping approach with data collecting using ensemble method with bagging and boosting respectively.