Early Heart Disease Diagnosis Through Feature Optimization with Ensemble Learning-Based Hybrid Bagging and Boosting Techniques

Cardiovascular disease, also known as heart disease, has emerged as a prominent global health concern over the past decade. Detecting heart complications at an early stage significantly enhances the effectiveness of treatment. In the pursuit of achieving precise early detection, various machine learning (ML) methods have been used. However, existing ML methods exhibit limitations in delivering efficient and accurate heart disease detection. In this work, we present a method for the early diagnosis of heart disease by employing a combination of feature optimization and ensemble learning using hybrid bagging and boosting techniques. Following data preprocessing, we harness an enhanced U-Net pre-trained architecture to extract features. This innovative approach utilizes both known and unknown features within the dataset to enhance feature extraction. To address data dimensionality challenges, we introduce a modified binary search (MBS) algorithm for feature optimization. This algorithm effectively identifies the most optimal set of features to address data dimensionality issues. Furthermore, we introduce an ensemble learning-based hybrid bagging and boosting technique for heart disease detection and classification, which leverages the strengths of both approaches. To gauge the performance of our proposed method, we conduct experiments using the Cleveland and Framingham's heart study datasets. This shows the effectiveness of our approach in terms of quality measures, thus validating its potential to advance the field of early heart disease diagnosis.