This session will introduce an end-to-end Machine Learning solution to detect credit card fraud. We will demonstrate how to use Snowpark for Feature Engineering and Java UDF to run ML models within Snowflake for scoring.

It will be using Snowpark to first load the data set being generated for this lab. Snowpark will then be used to create new features from the original data set that will be feeded into the ML algorithm that will be trained. Data Frames in Scala will be used to create some transformations and take care of all the Feature Engineering phase.

Python will be used to train a model with the features previously created. Features like zero-copy cloning will be automatically used to keep a copy of the data used for training. The model generated will be stored in PMML format and used to create a Java UDF that will be loaded into Snowflake for scoring.

 

Agenda

  • Snowflake for Data Science Intro – 15 minutes
  • Lab Introduction – 15 minutes
  • Credit Card Fraud Detection using Snowpark and Java UDF Lab – 1h 15m
  • Q&A – 15 minutes
Speakers
  • Carlos Carrero

    Partner Solutions Director Snowflake