Open Bandit Dataset
收藏资源简介:
Open Bandit Dataset是由康奈尔大学和ZOZO研究团队共同创建的公开数据集,专门用于支持现实且可复现的策略评估研究。该数据集包含在大型时尚电子商务平台ZOZOTOWN上收集的多个日志数据集,这些数据集是通过运行不同策略收集的。这使得首次能够实验性地比较不同的策略评估估计器。数据集的应用领域包括个性化推荐系统、在线广告和电子商务,旨在通过策略评估改进这些系统的性能。
Open Bandit Dataset is a public dataset jointly created by the research teams from Cornell University and ZOZO, specifically designed to support realistic and reproducible policy evaluation research. It contains multiple logged datasets collected on the large fashion e-commerce platform ZOZOTOWN through the deployment of various policies. This enables, for the first time, the experimental comparison of different policy evaluation estimators. Its application domains include personalized recommendation systems, online advertising and e-commerce, with the goal of improving the performance of such systems via policy evaluation.

- 1Open Bandit Dataset and Pipeline: Towards Realistic and Reproducible Off-Policy Evaluation康奈尔大学 · 2021年



