IEEE-CIS Fraud Detection challenge was first hosted by Kaggle in 2019. The idea was for competitors to develop a model to detect fraud from customer transactions. While IEEE-CIS already have a fraud prevention system in place, researchers were looking for ways to improve the current figure being saved by the system, and improve the customer experience.
Clone this repository to your computer.
To view explorations navigate to the project directory cd IEEE-CIS Fraud Detection from
your terminal then cd into the notebooks
directory. This directory contains data analysis
and the pipeline we converted into a package. To run the notebooks, you'll have
to install the data into a directory
called data. The directory must live at the same level as the notebooks
and packages
directory.
To use the sample the deployed model locally through the API, navigate to the project
directory from your terminal then cd into packages/fraud_detection_api
. From here,
run the following command:
py -m tox -e run
This will create a localhost link, simply click it or copy and paste it into your
browser. Then select the docs option and go to the predict
heading. There is already
an example instance there, but you may play around with the values.
Some ideas to extend this work:
- Replace the model
- Add monitoring