README.md

metadata

title: SHAP_NLP_TKA
app_file: app.py
sdk: gradio
sdk_version: 5.47.2

🧠 How This Model Works

📌 Model • TF-IDF converts clinical notes → numeric features • Logistic Regression predicts 30-day readmission • Produces probability + class label

📌 SHAP • SHAP KernelExplainer works for text models • Provides a force plot explaining: • What words increased risk • What words decreased risk • Outputs SHAP HTML (works in Hugging Face)

📌 Example Words That Increase Risk • “infection” • “fever” • “drainage” • “swelling” • “diabetes” • “slow healing”

📌 Words That Lower Risk • “no complications” • “independent ambulation” • “normal vitals”