KuaiRand
收藏资源简介:
KuaiRand是由中国科学技术大学和快手科技联合开发的无偏序贯推荐数据集,包含数百万次随机曝光视频的干预交互。该数据集通过在快手视频分享移动应用中随机插入推荐视频,记录了12种用户反馈信号,如点击、喜欢和观看时间,并收集了用户和视频的丰富特征及用户行为历史。KuaiRand旨在解决推荐系统中的曝光偏差问题,支持无偏线下评估,适用于交互推荐、长序列行为建模和多任务学习等多个研究领域。
KuaiRand is an unbiased sequential recommendation dataset jointly developed by the University of Science and Technology of China and Kuaishou Technology, containing millions of interventional interactions from randomly exposed videos. This dataset records 12 types of user feedback signals such as clicks, likes, and watch durations by randomly inserting recommended videos into the Kuaishou video-sharing mobile application, and collects rich features of users and videos as well as user behavior histories. KuaiRand aims to address the exposure bias problem in recommendation systems, supports unbiased offline evaluation, and is applicable to multiple research fields including interactive recommendation, long-sequence behavior modeling, and multi-task learning.

- 1KuaiRand: An Unbiased Sequential Recommendation Dataset with Randomly Exposed Videos中国科学技术大学 · 2022年



