OpenML DEMO: WATCH US TRAIN A MODEL OUTSIDE FEEDZAI, AND BRING IT ON IN

 

The OpenML Engine from Feedzai, launched just last week, represents the first-ever service that allows data scientists to bring their preferred approaches to a third party’s fraud-specific innovations. In this video, we show you how it works.

With all the hype that surrounds artificial intelligence, it can be easy to forget that AI is not just a concept. It’s a practical tool that banks must use in this high stakes game of cat and mouse, with fraudsters leveraging every tool they have to take advantage of large organizations.

To fight the rise of global financial  crime, a crowd of AI-enabled fraud detection platforms has emerged. These platforms force your data scientists to work within their confines — to use their proprietary data science methods, models, tools, languages, and libraries.

OpenML represents the opposite approach. Feedzai is embracing the data science community. We want your teams to integrate existing approaches with Feedzai’s system. This video demo shows just one example of how this works. Watch as a data scientist trains a machine learning model outside of Feedzai (in this instance, using Python and scikit-learn), and brings it inside Feedzai’s real-time processing for decision-making.

This demo begins with an example data set from a transaction historical CSV file. The data scientist applies sampling, trains the model, generates a ROC curve to score the evaluation of the dataset, and then persists the model.

Importing this model into Feedzai is a matter of simply copying a file to a certain location. From there, Feedzai’s product contains workflows with input schemas that specify the steps necessary for the model to begin processing real-time events, scoring them, and forwarding them to the Case Manager.

As we wrote in our OpenML announcement, the OpenML service has an SDK for Python, R, and Java; it provides close integration with tools like H20, R Studio, and DataRobot, and it works with libraries from any open source, like Spark’s MLlib and TensorFlow.

It took a company like ours, founded by underdog data scientists and aerospace engineers, to bring the flexibility and power of OpenML to fraud detection for the first time.

That’s how we know this much is true: make your data scientists happy, and the rest will follow.

RELATED:
Meet Feedzai’s OpenML Engine: “Bring Your Own Machine Learning” to Fight Fraud
Increasing Fraud Detecting By Speeding Up Machine Learning
What’s Next for Machine Learning: Ethics and Explainability in Machine Learning for Fraud