This cardiovascular risk predictor was built using a fully custom Decision Tree algorithm implemented from scratch in Python without using any external ML libraries like Scikit-learn. It is trained on the Cardiovascular Disease dataset repository, which contains anonymized medical records of over 70,000 patients, including features such as age, BMI, blood pressure, cholesterol, and lifestyle factors. The dataset was preprocessed with thoughtful transformations: age was converted from days to years, BMI was calculated and categorized using WHO standards, and a custom blood pressure classification system was designed using both systolic and diastolic readings. Features like health_risk and lifestyle_risk were engineered to combine multiple indicators into a single interpretable metric. The final model was trained using an information gain-based ID3 tree builder with a maximum depth of 12 and a minimum gain threshold to prevent overfitting. Performance was evaluated using a manually implemented accuracy, confusion matrix, and classification report delivering high interpretability and transparency. The model was also serialized using pickle for easy deployment within the Flask-based web application.
NOTE: This model is a demonstration prototype built entirely without external machine learning libraries, so its performance metrics may not match production-grade standards. It is not intended for actual clinical or medical decision-making. Always consult qualified healthcare professionals for any medical concerns.