基于深度学习的流量预测数据集

基于深度学习的流量感知数据集主要面向网络内生智能研究，将算力资源、感知数据和智能算法进行编排和管理，在5G网络切片WiFi切片平台引入基于深度学习的流量感知算法。首先使用安装在虚拟分组数据网网关所在的Docker容器中的网络数据采集工具Tcpdump和timeout工具完成数据的采集工作。每个业务的数据流通过vPGW与eNB之间的S1接口最终传输到用户端，通过timeout工具控制捕获不同时间粒度的数据流，然后使用Tcpdump工具捕获数据流，完成原始数据的采集。接着通过python脚本语言实现数据包长度解析、求取包络、数据截取和归一化四个过程，对流量预测数据预处理，将数据保存，得到用户的流量数据。最后将处理后的业务流数据输入流量预测模型中进行预测，将预测后的结果保存到相应的txt文档中，得到基于深度学习的流量预测感知数据，数据量18.3KB。

The deep learning-based traffic awareness dataset is primarily developed for network endogenous intelligence research, which orchestrates and manages computing resources, perception data and intelligent algorithms, and introduces deep learning-based traffic awareness algorithms onto the 5G network slicing and WiFi slicing platform. First, data collection is completed using the network data acquisition tools Tcpdump and timeout installed in the Docker container where the virtual Packet Data Network Gateway (vPGW) is located. The data flow of each service is finally transmitted to the user terminal through the S1 interface between vPGW and the evolved Node B (eNB). The timeout tool is used to control the capture of data flows with different time granularities, and then the Tcpdump tool is employed to capture the data flows, thus completing the collection of raw data. Next, four processes including packet length parsing, envelope calculation, data truncation and normalization are implemented via Python scripting language to preprocess the traffic prediction data, and the preprocessed data is saved to obtain the user's traffic data. Finally, the processed service flow data is input into the traffic prediction model for prediction, and the prediction results are saved into corresponding TXT documents to obtain the deep learning-based traffic prediction and perception data, with a total data size of 18.3 KB.

搜集汇总

背景与挑战

背景概述

该数据集是一个面向网络内生智能研究的流量预测数据集，专注于5G网络切片WiFi切片平台，通过采集和处理网络流量数据（使用Tcpdump和timeout工具，并经过Python脚本预处理），用于基于深度学习的流量感知算法验证。数据集由西安电子科技大学发布，数据量较小（18.3KB，总文件653.62KB），包含11个文件，源自国家重点研发计划项目，适用于通信技术领域的流量预测模型训练和分析。