{"success":true,"meta":{"author":"Rajat Surana","portfolio":"https://rajatsurana.is-a.dev","github":"https://github.com/rajatsurana19","self":"https://rajatsurana.is-a.dev/api/projects/diabetes-cdss","allProjectsEndpoint":"https://rajatsurana.is-a.dev/api/projects"},"project":{"id":"diabetes-cdss","slug":"diabetes-cdss","title":"Diabetes Clinical Decision Support","subtitle":"Machine Learning Early Screening Dashboard","category":"AI & ML","tech":["Python","Streamlit","Scikit-Learn","Random Forest","Pandas","NumPy"],"summary":"Interactive diagnostic support dashboard predicting diabetes risk using a Random Forest classifier trained on the Pima Indians dataset.","description":"A healthcare machine learning decision support application designed for clinical early screening. Trained on patient physiological data (glucose levels, insulin, BMI, blood pressure, age), the Random Forest model outputs risk probability distributions alongside confidence intervals, visual metric comparisons against safe medical thresholds, and individual risk factor explanations.","features":["Real-time patient risk classification based on clinical biometric inputs","Probability distribution visualization with dynamic risk category gauges","Feature importance attribution showing high-impact physiological drivers","Outlier detection and missing value imputation across clinical attributes","Deployed and accessible as a live interactive Streamlit cloud application"],"contributions":["Cleaned and normalized clinical data addressing zero-value physiological anomalies","Tuned Random Forest hyperparameters with cross-validation for optimal recall","Designed intuitive medical dashboard controls suitable for clinical screeners"],"metrics":["Interactive risk probability","Real-time clinical inference","Live Streamlit deployment"],"links":{"github":"https://github.com/rajatsurana19/Diabetes-CDSS","live":"https://diabetes-cdss.streamlit.app/","apiEndpoint":"https://rajatsurana.is-a.dev/api/projects/diabetes-cdss"},"image":"/project/diabetes.webp","highlight":false,"featured":false}}