Cardioml: A Streamlined Machine Learning Pipeline for Cardiovascular Disease Detection and Classification

Cardiovascular diseases (CVDs) are amongst the prominent reason of death globally, emphasizing the requirement for early diagnosis and accurate classification models. This study leverages a cardiovascular disease dataset to develop an efficient Machine Learning framework for predicting the occurrence of CVD. The dataset, containing 70,000 samples with 13 features, was preprocessed to address outliers and remove redundant and unimpactful features. Exploratory Data Analysis (EDA) revealed a balanced dataset, with an equal allocation of the target attribute (cardio). Attribute selection was carried out by utilizing Variance Inflation Factor (VIF), and correlation analysis, reducing the dataset to nine impactful features. The Gradient Boosting Classifier and AdaBoost Classifier were evaluated for predictive performance. The Gradient Boosting Classifier outperformed AdaBoost, achieving a test set accuracy of 73 % with high precision, recall, and F1-scores. Further hyperparameter tuning confirmed the Gradient Boosting Classifier's superior performance, but improvements in accuracy were minimal. These results highlight the importance of effective data preprocessing and robust model selection for CVD prediction. This study provides a comprehensive pipeline, from data cleaning to model optimization, offering insights for developing ML-based CVD detection systems.