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 systeminference.py
: performs inference, using part of the training datadata.py
: a support module to load the data file into Sparkfeatures.py
: a support module to add new features to the dataset, for the model to train on