Reproducibility Package for “Causal-Constrained Offline Reinforcement Learning for Auditable Dynamic Pricing”
Wat erin zit
This dataset contains the reproducibility package associated with the manuscript “Causal-Constrained Offline Reinforcement Learning for Auditable Dynamic Pricing”. The manuscript introduces CCO-ADP, a finite-grid causal constrained offline reinforcement-learning framework for auditable dynamic pricing. The method estimates interventional reward, cost, and transition tables from historical pricing logs, optimizes reward lower confidence bounds subject to cost upper bounds and lower-confidence support restrictions, and emits policy-level certificates with decision-level reason logs.The package supports verification and reproduction of the manuscript’s main computational results. It includes Python source code, experiment configurations, synthetic structural-causal market generators, random seeds, processed experiment outputs, manuscript table exports, figure-generation scripts, policy certificates, decision-log summaries, and artifact manifests. The empirical workflow covers seven structural-causal stress scenarios: base, demand shock, group imbalance, hidden confounding, inventory censoring, low overlap, and proxy leakage. These experiments are used to evaluate causal elasticity …
Bestanden en diensten
- DataCite REST API (JSON)api.datacite.orgOpenen