2020 Criteo Predictive Search Conversion Log Dataset

该数据集记录了2020年8月3日至10月16日Criteo预测搜索的日志信息。每条记录代表用户对一则产品相关广告的点击，包含了产品信息以及用户特征等各项属性。为了保护用户隐私，数据集中包含了经过哈希处理的分类特征，并且定义了转化行为为点击后30天内的购买行为。该数据集规模庞大，包含了多个子模型，这些子模型覆盖了不同时间窗口的转化情况。任务目标是预测转化率（CVR）以及每次点击的价值（VPC）。

This dataset contains log records of Criteo's predictive search from August 3 to October 16, 2020. Each record represents a user's click on a product-related advertisement, and includes various attributes such as product information and user characteristics. To protect user privacy, the dataset includes hashed categorical features, and defines a conversion behavior as a purchase made within 30 days after a click. This dataset has a large scale and includes multiple sub-models that cover conversion scenarios across different time windows. The task objectives are to predict the Conversion Rate (CVR) and Value Per Click (VPC).