Insurance Claims Modeling Project

This project analyzes and models real-life insurance claims data using various machine learning approaches such as XGBoost (Classification and Regression) and Tweedie Regression.

---

Project Structure

.
├── datasets/
│   ├── balanced_dataset.csv
│   └── freMTPL2freq.csv
├── models&results/
│   ├── TweedieReg
│   ├── XGBclass
│   └── XGBreg
├── utils/
│   ├── data_utils.py
│   ├── tweedie_utils.py
│   ├── XGB_class_utils.py
│   └── XGB_reg_utils.py
├── 1_data_analysis.ipynb
├── 2_XGBoost_class.ipynb
├── 2_XGBoost_reg.ipynb
├── 3_Tweedie.ipynb
├── 4_Interpretation.ipynb
├── LICENSE
├── README.md
└── requirements.txt

---

Setup Instructions

These instructions will help you set up the project locally.

1. Clone the Repository

git clone https://github.com/EdinZiga/InsuranceModeling
cd InsuranceModeling

2. Create a Virtual Environment

python -m venv .venv

3. Activate the Environment

• On Windows:

.venv\Scripts\activate

• On macOS/Linux:

source .venv/bin/activate

4. Install Dependencies

pip install --upgrade pip
pip install -r requirements.txt

5. Launch Jupyter Lab

jupyter lab

6. Run with Docker (Alternative Setup)

If you prefer using Docker instead of setting up a local Python environment, follow these steps:

Build the Docker Image

docker build -t insurance-modeling .

Run the Container

docker run -p 8888:8888 -v $(pwd):/app insurance-modeling

On Windows (CMD/PowerShell), use:

docker run -p 8888:8888 -v %cd%:/app insurance-modeling

Access Jupyter Lab

Once the container starts, it will output a URL containing a token. Open that URL in your browser to access Jupyter Lab.

---

Usage

Once Jupyter Lab is running in your browser:

1. Open the file: notebooks/1_data_analysis.ipynb
2. Run through each cell step by step to explore the data and run models.
3. Follow the notebooks in order:
   • 1_data_analysis.ipynb
   • 2_XGBoost_class.ipynb
   • 2_XGBoost_reg.ipynb
   • 3_Tweedie.ipynb
   • 4_Interpretation.ipynb

---

Dependencies

Main dependencies include:

• numpy
• pandas
• scikit-learn
• xgboost
• matplotlib
• seaborn
• statsmodels
• jupyterlab
• scipy
• shap
• scikit-optimize
• tqdm

All required packages are listed in requirements.txt.

---

Notes

• Ensure Python 3.8+ is installed.
• The dataset is availalbe in its folder
• Everything is 'pre-ran', so there are representative outputs saved
• The utils folder and the .py files inside contain all the functions used in the notebooks.
• All models and results are saved under models&results folder, depending on the model chosen
• All best runs achieved by me are presented at the start of each notebook.
• Enjoy!

---

License

This project is licensed under the MIT License.