Taobao Public Dataset

该数据集是一个开源数据集，由淘宝上基于云的主题场景用户点击日志构建而成。在数据集中，测试场景在训练集中未曾出现，确保了评估的公正性。为了适应冷启动多场景推荐案例，该数据集经过了聚类和进一步的加工处理。数据集的规模包括250个训练场景和105个测试场景。该数据集的任务是优化冷启动场景下的推荐系统。

This open-source dataset is constructed from cloud-based user click logs of thematic scenarios collected from Taobao. Test scenarios are entirely absent from the training set, thus guaranteeing the fairness of model evaluation. To cater to multi-scenario recommendation cases under cold start settings, this dataset has undergone clustering and further processing. It comprises 250 training scenarios and 105 test scenarios. The core task of this dataset is to optimize recommendation systems in cold start scenarios.

搜集汇总

背景与挑战

背景概述

该数据集是淘宝APP云主题推荐场景的用户点击日志，包含超过140万条点击记录，覆盖355个不同购物场景。数据集中还提供了用户购买历史、商品与主题匹配关系以及用户和商品的嵌入向量，适用于推荐系统优化和验证研究。