This is an experiment to build a card fraud detection model using Spark MLlib.

• The data file can be found here
• The data file should be installed locally at the path ./data/card-fraud/fraud-test.csv

This repo assumes that you have a Spark cluster available. We use a standalone cluster in Docker for dev and test purposes.

• training.py: trains the model and saves the MLlib pipeline to local file system
• inference.py: performs inference, using part of the training data
• data.py: a support module to load the data file into Spark
• features.py: a support module to add new features to the dataset, for the model to train on