Stroke Prediction
Medical Dicom
ClassificationDeveloped an end-to-end Machine Learning pipeline to predict the likelihood of a patient experiencing a stroke based on clinical, lifestyle, and demographic risk factors. This project addresses a critical healthcare challenge by leveraging predictive analytics to identify high-risk patients early, enabling preventative clinical intervention and reducing global stroke mortality rates. Technical Workflow & Implementation Data Preprocessing & Cleaning: Managed missing data imputation for critical health metrics (e.g., BMI, average glucose levels) and handled highly imbalanced target data using techniques like SMOTE (Synthetic Minority Over-sampling Technique) to ensure unbiased model training. Feature Engineering: Engineered and encoded categorical variables (e.g., smoking status, work type, residence type) and normalized continuous clinical data (e.g., age, hypertension, and heart disease history) for optimal algorithmic performance. Model Selection & Training: Supervised and evaluated multiple classification algorithms—including Logistic Regression, Random Forest, and XGBoost—to find the optimal balance between high sensitivity (recall) and precision. Performance Metrics: Evaluated model success using ROC-AUC score, F1-score, and Confusion Matrices, prioritizing high recall to minimize dangerous false negatives in patient risk assessment.