ReAdmitAI

• 

  This how the dashboard looks like.

Inspiration

Hospital readmissions are costly and often preventable. I was inspired by the potential of machine learning to help healthcare providers identify at risk patients early, enabling timely interventions and improving patient outcomes. The availability of a rich, real-world dataset from the UCI repository motivated us to build a practical, data-driven solution.

What it does

ReAdmitAI predicts whether a patient is likely to be readmitted to the hospital within 30 days of discharge. It provides a user-friendly Streamlit web app where clinicians or analysts can input patient details and instantly receive a readmission risk prediction, supporting better decision-making in clinical settings.

How we built it

We started by exploring and cleaning the Diabetes 130-US hospitals dataset, handling missing values and encoding categorical features. Using Python, pandas, and scikit-learn, we performed feature engineering and trained a logistic regression model. The model was then integrated into a Streamlit app for real-time predictions. All data analysis and model development were documented in a Jupyter notebook for transparency and reproducibility.

Challenges we ran into

Data Quality: The dataset contained many missing values and inconsistencies, especially in medication and diagnosis columns, which required careful preprocessing. Feature Selection: Deciding which features to keep, drop, or engineer was challenging due to the dataset's complexity and high dimensionality.

Accomplishments that we're proud of

Successfully built an end-to-end pipeline from raw data . -Created a clean, interactive interface that makes advanced analytics accessible to non-technical users.

What we learned

• The importance of thorough data cleaning and preprocessing in healthcare datasets.
• Practical experience integrating machine learning models into web applications for real-world use. ## What's next for ReAdmitAI
• Model Improvements: Experiment with advanced algorithms (e.g., XGBoost, Random Forest) and hyperparameter tuning for better performance.
• Deployment: Deploy the app to a cloud platform for broader accessibility.
• Explainability: Add more detailed explanations and visualizations to help users understand individual predictions.

Built With

• python