machine learning apps
Hospital Readmission Risk Prediction
An interactive healthcare analytics demo that estimates the probability of 30-day hospital readmission using a tuned XGBoost model and patient-level risk factors.
PythonStreamlitXGBoostScikit-learnPandasMachine LearningHealthcare Analytics
Project Overview
This project turns a hospital readmission risk prediction workflow into an interactive Streamlit application. Users can adjust patient-level clinical and administrative features, then receive a predicted readmission risk score with model-driven feedback.
Project Goal
- Estimate whether a patient may be at risk of 30-day hospital readmission.
- Demonstrate how a tabular machine learning model can be packaged into a user-facing healthcare analytics prototype.
- Use confidence, risk level, and feature-based explanations to make the model output easier to understand.
Model Architecture
1.Clean and prepare tabular patient readmission data.
2.Train and tune an XGBoost classification model.
3.Evaluate the model with recall, precision, F1-score, ROC-AUC, and threshold-based analysis.
4.Save the trained model and preprocessing pipeline.
5.Build a Streamlit interface for patient risk input and prediction.
6.Present the prediction as a risk score with readable healthcare-oriented explanations.