CTR预估数据集汇总

点击率预估 (CTR estimation) 是在线信息系统的核心模块之一，是推荐系统、付费广告、搜索引擎重要的组成部分，广泛的应用于商品购物、短视频、本地生活等领域中，与人们的生活息息相关，具有重要的业务价值。随着深度学习的广泛应用，深度点击率预估模型被广泛用于工业界的线上系统中。
<br />本文整理了学术界/业界公用的CTR预估数据集，并在协议许可的提供下为算法研发人员提供了天池站点存储。

Click-Through Rate (CTR) estimation is one of the core modules of online information systems, and a vital component of recommender systems, paid advertising, and search engines. It is widely deployed in scenarios including e-commerce shopping, short-video platforms, and local life services, closely intertwined with people's daily lives and holding significant business value. With the widespread adoption of deep learning, deep CTR estimation models have been extensively deployed in online industrial systems.
This paper collates publicly available CTR estimation datasets from both academic and industrial communities, and provides hosted storage for algorithm researchers and developers on the Tianchi platform under the terms of permitted licenses.

搜集汇总

背景与挑战

背景概述

该数据集是一个点击率预估（CTR estimation）相关数据集的汇总资源，主要整理了学术界和业界公用的多个CTR预估数据集，包括Criteo、Amazon、淘宝和饿了么等来源的数据。其特点是提供了一个集中的PDF文档（74.52KB），方便算法研发人员快速访问和下载这些数据集，用于学习和模型开发，并由阿里云PAI团队和天池数据科学团队维护更新。