电信科学 ›› 2020, Vol. 36 ›› Issue (3): 156-165.doi: 10.11959/j.issn.1000-0801.2020021

• 运营技术广角 • 上一篇    

基于机器学习分析VoLTE视频通话质量的研究及应用

钟其柱   

  1. 中国移动通信集团广东有限公司中山分公司,广东 中山 528400
  • 修回日期:2020-01-08 出版日期:2020-03-20 发布日期:2020-03-26
  • 作者简介:钟其柱(1985– ),男,中国移动通信集团广东有限公司中山分公司网络管理中心高级工程师,主要研究方向为移动通信网络的分析优化

Research and application of VoLTE video call quality based on machine learning

Qizhu ZHONG   

  1. Zhongshan Branch of China Mobile Group Guangdong Corporation,Zhongshan 528400,China
  • Revised:2020-01-08 Online:2020-03-20 Published:2020-03-26

摘要:

针对目前评估VoLTE视频通话质量的方法的缺点,提出了一种基于机器学习和网络指标参数的VoLTE视频通话质量评估方法。首先,采集解码核心网的网络参数指标数据,并进行预处理;然后,选取用于VoLTE视频通话质量评估的关键特征,并通过对比选取合适的机器学习算法,构建VoLTE视频质量评估的评估模型,从而实现不依赖于测试环境和原始视频的VoLTE视频通话质量实时评估。通过对从XDR(用户话单)数据中提取的特征指标数据进行预处理研究,解决了特征指标的标准化问题,便于指标特征输入评估模型;通过特征工程选出评估VoLTE视频通话的关键特征,减少特征维数,从而降低了算法的复杂度;同时采用先进的机器学习技术保证和提升算法的评估准确性。

关键词: 机器学习, 梯度提升树, VoLTE, 视频通话质量

Abstract:

To overcome the shortcomings of current methods for evaluating VoLTE video call quality,a method for evaluating VoLTE video call quality without reference based on machine learning and network index parameters was proposed.Firstly,the network parameters of the decoding core network were collected and preprocessed; then,the key features for VoLTE video call quality assessment were selected,and a reference-free evaluation model for VoLTE video quality assessment was constructed by comparing and selecting appropriate machine learning algorithms,so as to achieve real-time VoLTE video call quality assessment independent of the test environment and the original video.By researching the preprocessing of feature index data extracted from XDR data,the standardization of feature index was solved,and the evaluation model of feature input was convenient; the key features of VoLTE video call were selected and evaluated by feature engineering,which reduced the feature dimension and the complexity of the algorithm; at the same time,advanced machine learning technology was adopted to ensure and enhance the algorithm assessment accuracy.

Key words: machine learning, gradient boosting decison tree, voice over long-term evolution, video call quality

中图分类号: 

No Suggested Reading articles found!