特征数据的量化方法和服务器
By performing feature dimension splitting, timestamp correction, and data cleaning on the Joiner data stream, and calculating the frequency of feature occurrences and total data volume in each time window, the problem of incorrect loading and use of feature data during the ad recommendation process was solved, thus improving the accuracy of ad recommendation.
CN120746647BActive Publication Date: 2026-07-17HONOR DEVICE CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HONOR DEVICE CO LTD
- Filing Date
- 2024-07-30
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
In existing technologies, servers cannot effectively measure whether feature data is correctly loaded and used during the ad recommendation process, which affects the accuracy of ad recommendations.
Method used
By acquiring Joiner data streams during the ad recommendation process, performing feature dimension splitting, timestamp correction, and data cleaning, calculating the frequency of feature occurrences and total data volume in each time window, and determining the quantification results of features, we can evaluate the loading and usage of features.
Benefits of technology
It enables accurate quantification of feature data during the ad recommendation process, monitors the loading and usage of features, and improves the accuracy of ad recommendations.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN120746647B_ABST
Abstract
本申请实施例提供一种特征数据的量化方法和服务器，方法包括：在根据用户的第一操作进行广告推荐的过程中，若产生用户转化行为，获取第一数据流，第一数据流为对归因数据流和多个第一特征进行拼接得到的数据流，归因数据流为根据用户转化行为生成的数据流，第一特征为广告推荐过程中使用的广告特征；对第一数据流进行第一处理过程，得到第一结果，第一结果表征每个时间窗口下各个第一特征在第一数据流中出现的次数；对第一数据流进行第二处理过程，得到第二结果，第二结果表征每个时间窗口下第一数据流的总数据量；根据第一结果和第二结果，确定对第一特征的量化结果。由此可以衡量广告推荐过程中特征的加载使用情况，监控广告推荐的预估过程。
Need to check novelty before this filing date? Find Prior Art