Reproduce “Decision-Value Attribution in Predict-then-Optimize Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Konstantinos Ziliaskopoulos, Alexander Vinel, Alice E. Smith
DVA applies Shapley attribution to decision value in a fixed prediction-and-optimization pipeline, rather than to forecasts alone. It measures feature contributions relative to a baseline, depending on the pipeline and background. In a CAISO battery-arbitrage test, masking features improved held-out decision value in 21 of 23 comparisons, averaging $1.63/MWh; this does not establish causality or guarantee gains elsewhere.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.