用户行为冷热数据存储数据集

为了完成对冷热数据判定策略的精度测试与整体存储策略的功能、性能测试，所采用的数据库数据集是天池实验室所提供的“淘宝用户行为数据集”，该数据集记录了2017年11 月末一周的共计上亿条淘宝用户行为数据，其涉及约百万位不同用户的4种行为 （点击、购买、加购、喜欢）以及四百多万件不同商品。为将数据用于测试，实验对数据进行了清洗，首先将4种用户行为简化为了对数据的查询和修改操作，将'pv'和'fav'定义为GET操作，将'cart'和'buy'定义为 SET操作。最终数据保留了一条用户行为记录中的商品ID、行为类型以及时间戳信息，由于商品具体信息在数据中并未提供，因此本测试为商品加入了固定长度的商品信息，以此形成冷热数据存储机制的数据库数据集。To test the accuracy of the hot/cold data discrimination strategy as well as the functionality and performance of the overall storage strategy, the database dataset adopted is the Taobao User Behavior Dataset provided by Tianchi Laboratory. This dataset contains hundreds of millions of user behavior records spanning one week at the end of November 2017, involving four types of behaviors (click, purchase, add-to-cart, and favorite) from approximately one million unique users and over four million distinct commodities. For applying this dataset to the test, the experimental data was cleaned: first, the four types of user behaviors were simplified to query and modification operations, where 'pv' (page view) and 'fav' (favorite) were defined as GET operations, while 'cart' (add-to-cart) and 'buy' (purchase) were defined as SET operations. The final cleaned records retain only the commodity ID, behavior type, and timestamp information. Since no specific commodity information was included in the original dataset, fixed-length commodity metadata was added to each commodity to construct the database dataset for testing the hot/cold data storage mechanism.

搜集汇总

背景与挑战

背景概述

该数据集旨在测试冷热数据判定与存储策略的精度和性能，基于淘宝用户行为数据构建。数据经过清洗和简化，将用户行为归类为查询和修改操作，并保留了商品ID、行为类型及时间戳等关键信息。为适应测试需求，还添加了固定长度的商品信息以完善数据内容。