CTAR
收藏资源简介:
CTAR数据集是由华为诺亚方舟实验室创建的半合成数据集,用于推荐系统中的因果推断研究。该数据集基于一个著名的电影评分网站收集的电影及其描述性标签和评分信息构建,包含9,715部电影、10,273个独特标签和75,460个评分。数据集通过引入因果图形模型和缺失机制,自动生成用户-物品评分及其对应的用户-物品标签,旨在提供用户为何评价物品的原因。CTAR数据集不仅限于推荐系统研究,还可用于其他研究任务,如因果推断和解释性推荐模型的开发。
The CTAR dataset is a semi-synthetic dataset created by Huawei Noah's Ark Lab for causal inference research in recommender systems. Constructed using movie information, descriptive tags and ratings collected from a well-known movie rating website, it contains 9,715 films, 10,273 unique tags and 75,460 ratings. By introducing causal graphical models and missing data mechanisms, the dataset automatically generates user-item ratings and their corresponding user-item tags, with the goal of uncovering the reasons behind users' rating behaviors. The CTAR dataset is not limited to recommender system research, and can also be applied to other research tasks such as causal inference and the development of explainable recommendation models.

- 1A Semi-Synthetic Dataset Generation Framework for Causal Inference in Recommender Systems华为诺亚方舟实验室 · 2022年



