Allstate Claims Severity - Machine Learning

This project is part of NTU CE4041 Course Project.

Get started

Requirements

• This project works on Python 3.6.x 64-bit version.

If you want to train neural network and XGBoost tree model, you need these 2 requirements below. Otherwise, you do not need these requirements.

1. Tensorflow with GPU support - follow this instruction
2. XGBoost (with python package) - follow this instruction

Create virtual environment

1. Install virtualenv package:

   pip install virtualenv

2. Create virtual environment

   virtualenv venv

3. Activate the created virtual environment

   • Using Git Bash (reccommended):

   . venv/Scripts/activate

   • Using Command Prompt:

   venv\Scripts\activate.bat

4. Install requirements:

   • If Tensorflow is installed with GPU suport:

   pip install -r requirements.txt

   • else:

   pip install -r light-requirements.txt

Run Jupyter Notebook

Windows:

1. Double click start.bat. Default browser should be opened with Jupyter Notebook.
2. Open ./src/AllState.ipynb

Other OS:

1. Activate virtualenv

2. Use command:

   jupyter notebook

   Default browser should be opened with Jupyter Notebook.

3. Open ./src/AllState.ipynb

View Notebook in HTML without running code:

• html files are available in ./html folder
• There are 2 print out versions: with GPU and without GPU versions

Usefil links

One-Hot-Encoding

• https://machinelearningmastery.com/why-one-hot-encode-data-in-machine-learning/