WC_FULL
收藏arXiv2025-09-30 收录
下载链接:
https://github.com/qwerfdsaplking/F2R-HMT
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了连续两天的日志数据(第一天和第二天),涉及1700万用户和50万个视频的2000万次展示/点击记录。其中,第一天的日志数据用于训练,第二天的日志数据用于测试。该数据集将图视为无向异构图,规模宏大,包含超过8亿个节点和460亿条边。任务是对点击通过率(Ctr)进行预测。
This dataset contains two consecutive days of log data (Day 1 and Day 2), involving 20 million display/click records from 17 million users and 500,000 videos. The log data from Day 1 is used for model training, while the data from Day 2 is reserved for testing. This dataset adopts undirected heterogeneous graphs as its graph structure, with a massive scale encompassing over 800 million nodes and 46 billion edges. The downstream task is Click-Through Rate (CTR) prediction.
提供机构:
WeChat



