Get started with Autopilot

• Direct Marketing with Amazon SageMaker Autopilot
  • Contents
  • Introduction
  • Prerequisites
  • Downloading the dataset
  • Upload the dataset to Amazon S3
    • Reserve some data for calling batch inference on the model
    • Upload the dataset to Amazon S3
  • Setting up the SageMaker Autopilot Job
  • Launching the SageMaker Autopilot Job
  • Tracking SageMaker Autopilot job progress
  • Results
    • Perform batch inference using the best candidate
    • View other candidates explored by SageMaker Autopilot
    • Candidate Generation Notebook
    • Data Exploration Notebook
  • Cleanup
• Regression with Amazon SageMaker Autopilot (Parquet input)

Feature selection

• Bringing your own data processing code to SageMaker Autopilot
  • Table of contents
  • Setup
  • Generate dataset
  • Upload the data for training
  • Feature Selection
  • Prepare Feature Selection Script
  • Create SageMaker Scikit Estimator for feature selection
  • Batch transform our training data
  • Autopilot
    • First we add column names to transferred data
    • Set up and kick off autopilot job
    • Tracking SageMaker Autopilot job progress
    • Results
  • Serial Inference Pipeline that combines feature selection and autopilot
  • Set up the inference pipeline
  • Make a request to our pipeline endpoint
  • Delete Endpoint

Model explainability

• Explaining Autopilot Models
  • Table of Contents
  • Introduction
    • SHAP
  • Setup
    • Create an inference endpoint
    • Wrap Autopilot’s endpoint with an estimator class.
    • Data
    • Background data
    • Setup KernelExplainer
  • Local explanation with KernelExplainer
  • KernelExplainer computation cost
  • Global explanation with KernelExplainer
  • Conclusion