New Releases, Enhancements, Changes + Arize in the News!
Model Performance Tab: Output Segmentation Analysis
Confusion Matrix - Given a model is set up with a default positive class (set this up on the config tab), we generate a matrix showcasing TF, FP, TN, FN counts of the dataset(s) to easily visualize model performance issues.
single dataset confusion matrix
compare two datasets with side-by-side confusion matrices
Calibration Line charts - A new metric - also known as “Average Predictions over Average Actuals”, calibration shows the ideal threshold (orange diagonal line) in addition to the current dataset(s) calibration!
Model Performance Tab: Cohort Filtering
Various filter combinations can now be applied to individual datasets (or both datasets). Previously, filters were applied across both datasets by default.
Best Practices for ML Monitoring and Observability of Demand Forecasting Models
This holiday season, having an ML observability strategy in place may be the difference between having enough inventory on-hand to meet holiday demand or losing out on millions of sales due to out-of-stock merchandise.
Ancestry CEO, Deb Liu, On Building Teams, Closing the Gender Gap, and Learning From Failure
Best Practices For ML Observability In Lending and Insurance with America First Credit Union
The lending and insurance industries are being transformed by AI, as financial services and insurance companies deploy ML models to inform everything from pre-eligibility checks to credit decisioning and premium pricing. In this webinar, Reah Miyara, Arize AI's Head of Product, and Richard Woolston of Americas First Credit Union give an overview of ML observability, demo the Arize AI platform, and have a fireside chat on best practices for observing lending models in production. Watch it.