Databricks launches Feature Views real-time fraud detection tutorial
Understand this story
Databricks published a tutorial demonstrating real-time fraud detection with Feature Views in Feature Store: features are defined once and simultaneously generate point-in-time-correct training data and online inference pipelines, avoiding training-serving skew and data leakage. The tutorial follows a $900 payment from ingestion via a Kafka stream to a fraud score returned by Model Serving, with step-by-step code. Databricks notes that the Sawtooth guide recommends waiting a full two days before going to production. When testing candidate features, the Unity Catalog function must already exist, but the feature definition does not need to be registered; additionally, feature permissions do not replace SELECT permissions on source or materialized tables for training and inference.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Databricks BlogReal-time fraud detection with Feature Views
Databricks describes how Feature Views in its Feature Store let teams define a feature once and use the same definition for point-in-time-correct training data and managed online-serving pipelines, reducing training-serving skew and data leakage. The walkthrough follows a $900 payment from a Kafka stream through rolling and Sawtooth window aggregations, an on-demand deviation feature, model training and Model Serving, with Unity Catalog providing governance and lineage.
Activity for this story
Not enough continuous observations to show a trend yet.