网络侧估计终端用户视频体验建模后续研究  

A Follow-up Study of Network-side Estimation of Client User Video Experience Modeling

作　　者：吴天行 李化锋 王淑娟[2] WU Tianxing;LI Huafeng;WANG Shujuan(College of Intelligent Systems Science and Engineering,Harbin Engineering University,Harbin,Heilongjiang 150001,China;College of Mathematical Sciences,Harbin Engineering University,Harbin,Heilongjiang 150001,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150001 [2]哈尔滨工程大学数学科学学院,黑龙江哈尔滨150001

出　　处：《数学建模及其应用》2020年第3期61-66,74,F0003,共8页Mathematical Modeling and Its Applications

基　　金：黑龙江省教育教学改革项目(SJGY20190149)。

摘　　要：在2017年“深圳杯”数学建模挑战赛中,针对移动视频业务中网络质量与终端用户视频体验之间的关系进行了深入研究,分析了视频初始缓冲与发生卡顿的原理,分别基于数据驱动和物理背景,分析建立了初始缓冲时延和卡顿占比模型.本研究是针对初始缓冲时延模型的进一步改进.一方面,针对原模型在初始缓冲峰值速率较小情况下预测精度低的问题,引入了稳峰比的概念,修正了原模型中不合理的假设,显著提升了预测精度;另一方面,通过图示检验法分析了模型的异方差性,并建立了区间估计模型,从而更准确地反映数据的分布特征,扩大了模型的适用范围.In the 2017 Shenzhen Cup Mathematical Modeling Challenge,we conducted an in-depth study on the relationship between network quality and client user video experience in mobile video service.The principles of video initial buffering and occurrence of cadence are analyzed,and two models are established based on data-driven method and mechanism analysis,respectively.This paper is a further improvement of the original work.On the one hand,to address the problem that the prediction accuracy of the initial buffering latency model is particularly low in the case of small initial buffering peak rate,the concept of stable-to-peak ratio is introduced,which corrects the unreasonable assumptions in the original model and significantly improves the prediction accuracy in the case of poor network conditions.On the other hand,the heteroskedasticity of the prediction residual is analyzed through graphical test method,and the interval estimation model is established,which more accurately reflects the distribution characteristics of the data and expands the applicability of the model.

关 键 词：终端用户视频体验模型 初始缓冲时延 区间估计 稳峰比 

分 类 号：O29[理学—应用数学]