A robust CatBoost model for cardiovascular disease screening validated on external clinical data (Cardio Dataset and Z-Alizadeh Sani Data)

This dataset repository was compiled as part of an experimental effort to develop and rigorously validate a **highly sensitive machine learning model (CatBoost)** for **cardiovascular disease (CVD) screening**. The primary effort involved preprocessing, feature engineering, and applying advanced imbalance handling techniques (SMOTEENN) to optimize the model for maximum **Recall (sensitivity)**. The data provided here enables full reproduction of the model training and the crucial **external validation** phase, which forms the core finding of our associated manuscript. ### **Included Files and Descriptions** This submission contains three key files necessary for the full reproducibility of the CatBoost model: 1. **`cardio_train.csv` (Primary Training and Internal Validation Data):** This is the main cleaned and processed dataset used for training and internal validation of the CatBoost machine learning model. It contains the demographic and clinical features extracted from the main dataset, used to predict the presence of cardiovascular disease (CVD). 2. **`Z-Alizadeh_sani_dataset.csv` (External Validation Data):** This dataset serves as the independent, external validation set for the trained CatBoost model. It was used to rigorously test the model's generalization capability specifically for Coronary Artery Disease (CAD) diagnosis in a completely separate clinical context. 3. **`Ghalb_AI.ipynb` (Python Code/Software):** This Jupyter Notebook contains the complete workflow, from data loading, full preprocessing (including SMOTEENN application), model training, optimization, and final evaluation on both the internal test set and the external validation data. ### **Key Variables Description** The main clinical variables included in the two datasets are defined as follows: **1. `cardio_train.csv` Variables:*** **age:** Age of the patient (in days).* **gender:** Patient's sex (1: Female; 2: Male).* **ap_hi:** Systolic blood pressure (in mm Hg).* **ap_lo:** Diastolic blood pressure (in mm Hg).* **cholesterol:** Cholesterol level (1: normal, 2: above normal, 3: well above normal).* **cardio:** **Target Variable** (0: Healthy, 1: CVD present). **2. `Z-Alizadeh_sani_dataset.csv` Variables (External Data):*** **Age:** Age of the patient (in years).* **Sex:** Patient's sex (0: Female, 1: Male).* **HTN:** Hypertension status (0: No, 1: Yes).* **DM:** Diabetes Mellitus status (0: No, 1: Yes).* **CAD:** **Target Variable** (Coronary Artery Disease diagnosis) (0: No, 1: Yes).