This MVP of a fraud detection learning machine model has been developed by 3 data scientist students - Selma Rochet, Sébastien Lartigue, Patrick Chardavoine - in the context of a bootcamp provided by Jedha training center.
The MVP is hosted at "https://patrickcharda-detectionfraude.hf.space"
The dataset that has been choosen to build the MVP comes from Kaggle : "https://www.kaggle.com/datasets/umuttuygurr/e-commerce-fraud-detection-dataset?select=transactions.csv"
The app has 2 modes :
- unit tests
- bulk tests
Unit tests can be done by 2 ways :
-
a) automatic generation of a fraud or a legitime transaction by clicking a button that triggers a selection in an existing test dataset
-
b) get a json file that contains a fraud or a legit
The user is able to pick a csv file containing a formatted dataset of online transactions. Try it with X_test_app.csv (look at Assets section)
In the assets directory you can find :
- X_test_app.csv to try bulk test
- json files with fraud or legitime transaction
- fraud_xgb_model.pkl (xgboost model)
- a user guide (.odt)
- jupyter notebooks