RewardEstimation Documentation

RewardEstimation provides functionality for estimating rewards for use in evaluating prescriptive problems. This documentation includes:

• an overview of reward estimation, including an introduction to observational data and the difficulty of evaluating prescriptive problems, how reward estimation can assist with this task, and some best practices for practical application of reward estimation.
• guides to the reward estimation process and the relevant learners for each type of reward estimation:
  • reward estimation with categorical treatments
  • reward estimation with numeric treatments
• the RewardEstimation API reference

For more information and a demonstration of reward estimation applied to real problems, you can also refer to the following examples:

• Categorical treatment:
  • Quick Start Guide: Optimal Prescriptive Trees
  • Quick Start Guide: Optimal Policy Trees
• Single numeric treatment:
  • Quick Start Guide: Optimal Policy Trees
  • Case Study: Grocery Pricing
• Multiple numeric treatments:
  • Case Study: Diabetes Management Policy