These notebooks accompany several talks and workshops developed by Will Benton and Sophie Watson.

• Our workshop "From Statistics to Serverless: Intelligent Applications on OpenShift 4" was delivered at Red Hat Summit and IBM Think in 2020. Slides are available as a PDF or as a movie. The rest of the instructions in this README will cover getting the basic application running.
• Our GTC 2021 talk "Fighting Fraud With One App In Many Ways: GPU-Accelerated End MLOps on Kubernetes" built up a similar fraud-detection application in two versions using RAPIDS.ai. Notebooks from that talk are on this branch.

In order to build and run a model service, you'll need an OpenShift cluster, but you can experiment with the notebooks on your own time. Here's how:

Use binder. (We don't recommend this if you'll be running the tutorial over conference wifi, but it requires almost no setup and can run from a computer that only has a browser.)

If you want to experiment with the data generator, you'll want to use your own computer.

1. Make sure you have Python 3.7 installed, installing it if necessary
   • If you have a favorite package manager, use that
   • if not, python.org has binaries for many platforms
2. Make sure you have git installed, installing it if necessary
   • If you have a favorite package manager, use that
   • if not, git-scm.com has binaries for many platforms (you won't need a GUI)
3. Install pipenv
   • on a Mac, the easiest way is probably brew install pipenv
   • on a Fedora Linux machine, the easiest way is probably dnf install pipenv
   • on Windows, if you have Python installed already, the easiest way is probably to use pip

1. Clone this repository: git clone https://github.com/willb/fraud-notebooks/
   • tip: if you don't have git installed, you can also download an archive of this repository
2. Change to this repository's directory: cd fraud-notebooks
3. Install the dependencies: pipenv install
4. Run the notebooks: pipenv run jupyter notebook