AntM2C
收藏资源简介:
AntM2C是由蚂蚁集团基于支付宝平台真实工业数据创建的大型多场景多模态点击率预测数据集。该数据集覆盖了支付宝平台上的五种不同类型的商品点击率数据,包括广告、优惠券、小程序、内容和视频。除了基于ID的特征外,AntM2C还提供了原始文本和图像两种多模态特征,有效连接不同ID的商品。数据集包含2亿用户和600万商品,总计10亿样本,是目前最大规模的点击率数据集,用于可靠和全面的点击率模型评估。AntM2C的应用领域包括多场景建模、商品和用户冷启动建模以及多模态建模,旨在解决推荐系统中的点击率预测问题。
AntM2C is a large-scale multi-scenario and multi-modal click-through rate (CTR) prediction dataset created by Ant Group using real industrial data from the Alipay platform. This dataset covers click-through rate data of five distinct product types on the Alipay platform, including advertisements, coupons, mini-programs, content, and videos. Beyond ID-based features, AntM2C also provides two multimodal features: raw text and images, which effectively link products with different IDs. The dataset contains 200 million users and 6 million products, with a total of 1 billion samples, making it the largest click-through rate dataset to date for reliable and comprehensive CTR model evaluation. Application scenarios of AntM2C include multi-scenario modeling, product and user cold-start modeling, and multimodal modeling, aiming to solve the click-through rate prediction problem in recommendation systems.




