Fraud Detection

Ubisoft home assignment

Arnaud Baumann

This project contains different files:

• assemble_adaboost.ipynb Notebook with answers to the questions. You need to run this notebook in order to create a save of the trained assemble.Adaboost model as well as a user_id_counter dictionnary. These two pickled objects will be reused for the api.

• utils.py A python file with helper functions for pre-processing and plots

• assembleAdaboost.py A python file containing the Assemble Adaboost model implementation

• app.py A python file containing the api

• Dockerfile the dockerfile to build the api

• test_app.py tests for the api

API build and run instructions

To build the app in a container, simply run: docker build -t fraud_detection . Then to run it run: docker run -d -p 5000:5000 fraud_detection

You get then make requests on http://localhost:5000/score