To Grant or Not to Grant — Machine Learning 2024/2025 (Group 9)

📚 MSc in Data Science and Advanced Analytics — Nova IMS

This project aims to build a Machine Learning model capable of predicting the Claim Injury Type for each case from the New York Workers’ Compensation Board (WCB), using real data from 2020–2022.

---

🎯 Objectives

• Automate the WCB’s decision on injury compensation type.
• Benchmark multiple supervised classification algorithms.
• Select the most generalizable and explainable model.
• Deploy a Streamlit web app for real-time predictions.

---

⚙️ Approach

1. Exploratory Data Analysis (EDA) – outlier and missing value treatment, variable inspection.
2. Feature Engineering – creation of temporal, categorical, and combined variables.
3. Model Benchmarking – Decision Tree, Random Forest, Gradient Boosting, XGBoost, and CatBoost.
4. Model Selection – GridSearchCV with Stratified Cross Validation (F1-macro).
5. Deployment – Streamlit app (app.py) for interactive use.

---

🧠 Final Model

The Random Forest model was selected for its robustness, interpretability, and best F1 performance on the test set.

---

🚀 Run Locally

pip install -r requirements.txt
streamlit run app.py