The authors introduce a What-If Machine that facilitates data-driven decision making through the evaluation of hypothetical scenarios. This tool creates alternative realities by resampling data distribution and comparing it to a baseline to measure impact on a target metric. The machine can confirm/reject manually developed intuitions and provide insights on the target metric. The method is applicable to any tabular data and supports data-informed decision making by utilizing historical data to infer future possibilities. The What-If Machine stands out for offering real-time analysis and automatic insights into high-impact areas on the target metric.
Publication date: 29 Sep 2023
Project Page: https://arxiv.org/abs/2309.17364v1
Paper: https://arxiv.org/pdf/2309.17364
