在线购物用户行为数据集
收藏资源简介:
该数据集由亚马逊公司收集,包含了一个电商平台上的3526名用户的31865个购物会话,总计230965个用户行为。每个会话由一系列时间对齐的<context, action>对组成,context反映了用户观察到的网页(如产品视图、筛选状态),action代表用户的输入(如点击、搜索或会话终止动作)。数据集的最终结果包括4432个购买动作和27433个会话终止动作。此数据集用于严格评估各种LLM在动作级别的模拟人类行为的准确性。
This dataset was collected by Amazon, containing 31,865 shopping sessions from 3,526 users on an e-commerce platform, with a total of 230,965 user behaviors. Each session consists of a series of time-aligned <context, action> pairs, where context reflects the web pages observed by the user (e.g., product views, filtering states), and action represents user inputs such as clicks, searches, or session termination actions. The final dataset includes 4,432 purchase actions and 27,433 session termination actions. This dataset is used to rigorously evaluate the accuracy of various LLMs in simulating human behavior at the action level.




