MLPER-17: Review for updated data/features for retraining

PDFRSS

Establish a framework to run data exploration and feature engineering at pre-determined time intervals based on data volatility and availability. New features that have not been considered in the model training can affect the accuracy of model inferences.

Implementation plan

• Explore changing data with Amazon SageMaker AI Data Wrangler - Evaluate the rate of change of the business environment to set a schedule for validating and possibly changing model input data and features. Analyze the data using Amazon SageMaker AI Data Wrangler and explore new features. Establish a team who will periodically evaluate and possibly change features and retrain the model.

Documents

• Prepare ML Data with Amazon SageMaker AI Data Wrangler

• Amazon SageMaker AI Model Monitor

Blogs

• Exploratory data analysis, feature engineering, and operationalizing your data flow into your ML pipeline with Amazon SageMaker AI Data Wrangler 

Videos

• Introducing Amazon SageMaker AI Data Wrangler – AWS re:Invent 2020