机器学习模型的训练方法、装置和数据的预测方法、装置
By segmenting HTTP request data and calculating path depth vectors, and combining fragment information of masked words with weighted training using a loss function, the problem of low processing efficiency of the Transformer pre-trained model is solved, and the prediction performance and sensitivity are improved.
CN116304666BActive Publication Date: 2026-07-17CHINA TELECOM CORP LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA TELECOM CORP LTD
- Filing Date
- 2022-12-12
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing Transformer-based pre-trained language models are inefficient when processing HTTP request data, leading to a decline in prediction performance.
Method used
By segmenting HTTP request data, calculating the path depth vector of words, and processing it using a machine learning model, the model is trained by combining fragment information of masked words and the weighted mean of the loss function, thereby reducing the learning of unnecessary information.
Benefits of technology
It improves processing efficiency, enhances prediction performance, and increases the model's sensitivity to HTTP request data and its ability to capture structural information.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116304666B_ABST
Abstract
本公开涉及一种机器学习模型的训练方法、装置和数据的预测方法、装置，涉及计算机技术领域。该训练方法，包括：对训练数据集合中的HTTP请求数据，进行分词处理，确定HTTP请求数据包括的多个词；计算多个词中的每一个的路径深度，以确定路径深度向量；利用机器学习模型，对路径深度向量进行处理，获取第一预测结果；利用第一预测结果和HTTP请求数据的标注，训练机器学习模型。本公开的技术方案能够提高处理效率，从而提高预测性能。
Need to check novelty before this filing date? Find Prior Art