Credit Card Fraud Detection

---

For this project, I used the datasets from the kaggle competition called IEEE-CIS Fraud Detection. The competition aims to improve fraud prevention system by building fraud detection models based on Vesta Corporation's real-world e-commerce transactional data, which contains information from device type to product features. My personal goal for this project is to not only explore the data and build models, but to also build an API server with retrainable model. To achieve this goal, I used fastapi, lightgbm and ray tune.

Also, I decided to develop this project to be the same as how data-related projects are developed in real-world scenarios, wherein the end goal of development is a project that is feasible for production. Therefore, I have put efforts on creating:

1. Exploratory Data Analysis (EDA) in the notebooks/ folder;
2. An API Server inside the api/ folder;
3. Files for deployment such as Dockerfile and docker-compose.yml;
4. Documentations in the docs/ folder; and
5. Some necessary scripts in scripts/ folder.

Services

I have two services: app and script.

App

The app service is a machine learning API that is open on port 8000. I used fastapi for the API server, so you can check it on http://localhost:8000/docs after you run the app service.

Scripts

The script sends the request for the predictions on new sets of data, such as the kaggle testing data. After the script get all the responses, files will be written on /tmp/submission.csv (on host and container), but this part can take a lot of time. It is suggested to use docker logs -f lightgbm-project-demo_script_1 to check the progress of the process.

How to run this demo

1. Install the requirements

• docker
• docker-compose
• make

2. Download the datasets

Option 1.

Setup kaggle API and use

make init-data

Option 2.

1. Create a data folder: i.e.

mkdir data

2. Download the data from kaggle IEEE-CIS Fraud Detection.

3. Put the ieee-fraud-detection.zip inside the data/ folder.

4. Unzip ieee-fraud-detection.zip.

3. Build the image

make build

4. Start the services

Start both the two services

docker-compose up

or only start the app service using

docker-compose up app

Here are some documentations

How to set up the working environment for this project

API example

Note

If you want to change the hyperparameters search space, you can go to config.py. Or you even want to use other framework the build the model, I think this demo is detail enough as a reference for your project.

About the hyperparameter search algorithm, I am using the random search for this demo but if you want to try other searching algorithm, you can change train.py.

I have another project use Tensorflow as the back bone model. Take a look about the project lightgbm-project-demo