IPTV用户体验异常的自动化检测

IPTV系统架构复杂，涉及大量终端、网元和连接等。为此，运营商建立了较为完善的监控体系，特别是采集了海量的EPG体验数据，形成多维度的监控指标，旨在监控用户体验水平。然而，监控指标繁多，导致运维人员监察各项指标费时费力，无法及时发现异常，也难以确定异常原因。为解决上述人工运维的痛点，采用一种智能异常检测方法，并根据实际应用进行改进，高效地实现对海量数据的实时分析。实践表明，该方法计算成本较低，适应现网异常变化，快速准确地检测异常，从而减少人力成本，提高运维效率，推进智能运维。

Automated anomaly detection of IPTV user experience

Architecture of IPTV system is complex, involving a large number of terminals, network elements and connections. Therefore, a relatively complete monitoring system, which collected massive EPG experience data and formed multi-dimensional monitoring indicators, had been established to monitor user experience. Due to a large number of indicators, manual monitoring was time consuming and laborious. It was hard to detect anomalies in time and it was impossible to determine the cause of abnormality. To solve the pain points of the operation, an intelligent algorithm was implied and improved to analyze massive experience data. The practice indicates that the algorithm with low calculation cost adapts to abnormal changes in the network and detect anomalies accurately and quickly, which reduces labor costs, improves operation efficiency and promotes intelligent operation.