Executive Summary : | Chronic diseases, such as heart disease and diabetes, pose significant health challenges globally and have a profound impact on individuals, healthcare systems, and societies at large. Among chronic diseases, heart disease and diabetes stand out as major contributors to morbidity, mortality, and healthcare costs. Understanding the nature of these conditions, their risk factors, and the importance of prevention and effective management is crucial for promoting public health and improving the well-being of individuals affected by these diseases. Heart disease, also known as cardiovascular disease, encompasses a range of conditions affecting the heart and blood vessels. It is a leading cause of death worldwide and is closely connected to factors such as high blood pressure, high cholesterol levels, smoking, unhealthy diets, physical inactivity, and obesity. Diabetes, on the other hand, is a metabolic disorder characterized by high blood glucose levels. It occurs when the body either does not produce enough insulin or cannot effectively use the insulin it produces. Machine learning (ML) can aid in early CVD detection and mitigate mortality rates. Previous research studies have employed various ML approaches to detect CVD and determine severity levels. However, none of these studies have focused on optimizing the ML model's performance for CVD detection and severity-level classification. To resolve this issue, an effective method is proposed based on the Synthetic Minority Oversampling Technique (SMOTE) to address the issue of imbalanced distribution, six ML classifiers to detect patient status, and Hyperparameter Optimization (HPO) to find the best hyperparameter for ML classifiers along with SMOTE. Two public datasets were utilized to develop and test the model using all features. Results demonstrate that SMOTE and Extra Trees (ET), optimized using hyperband, outperformed other models and achieved a 99.2% and 98.52% accuracy in CVD detection, respectively. Additionally, the developed model achieved 95.73% severity classification using the Cleveland dataset. The proposed model can assist doctors in determining a patient's current heart disease status, potentially preventing heart disease-related mortality through early therapy. By training the models on large datasets with labelled instances of disease presence or absence, ML algorithms can learn to recognize complex relationships and identify risk factors that may not be apparent to traditional statistical methods. These predictive models can then be used to classify new individuals and estimate their likelihood of developing heart disease or diabetes. ML-based prediction systems have the potential to assist healthcare professionals in early detection, personalized risk assessment, and targeted interventions, ultimately improving patient outcomes and enabling more efficient allocation of healthcare resources. |