论文部分内容阅读
针对运营商难以对用户感知质量(QoE)进行主动评估和预警,且难以有效利用海量的网络日常运行数据的问题,在调研了运营商的IPTV网络运行数据和指标基础上,分析了现有的IPTV QoE评价方法和相关的数据挖掘算法,提出了一种基于数据挖掘的IPTV QoE评价方法.该方法包括特征指标相关性分析、指标选择、指标降维、QoE评分及QoE预警等,通过相关性分析、回归分析等算法实现了从原始指标数据到IPTV QoE评价模型建立的过程.基于真实数据集的验证结果表明:当该方法选择80作为QoE评分阈值时,能够达到66.35%的预警命中率.
It is difficult for operators to take the initiative to evaluate and warn users ’perceived quality (QoE), and it is difficult to effectively utilize the massive network daily operation data. Based on the investigation of operators’ IPTV network operation data and indexes, IPTV QoE evaluation method and related data mining algorithms, an IPTV QoE evaluation method based on data mining is proposed, which includes correlation analysis of feature index, index selection, dimension reduction of indicators, QoE score and QoE alerting, Analysis, regression analysis and other algorithms to achieve from the original index data to IPTV QoE evaluation model established based on the validation of the real data set results show that: when the method selects 80 as the QoE score threshold, it can achieve 66.35% of the early warning hit rate.