China Unicom was spending about €24 million per year ($24.44 million) and 251,000 working days on 4G and 5G field testing, but could only cover about 80% of the range. It took between 30 minutes and an hour for an experienced engineer to deal with a 5G network problem, from identification to discovering its root cause then deciding on the best course of action. A highly skilled person could only handle 12 problems a day at most.
Consequently, it took a long time to handle customers’ complaints and VIP customers’ complaints usually required engineers to visit sites for testing, which typically took at least four hours and reduced customers’ satisfaction.
Automating the network
China Unicom is rapidly automating network operations, and deploying AI and analytics. SWAT draws on the Hadoop platform’s distributed big data capabilities, which processes, correlate, and analyze some 55 billion records daily from the 31 provinces covered by the operator’s network. Then it automatically generates all the key performance indicators needed for field testing up to 100% of the network.
“SWAT provides closed-loop, virtual network testing. It carries out automated data collection, identifies problems, applies root-cause analysis, decides on the best course of action and evaluates the results. This closed-loop process is aligned with China Unicom’s Autonomous Network framework, which is based on TM Forum’s principles,” explains Xiansong Liu, Director of Shanghai Network AI Center. “Based on big data and AI, SWAT has been proven in practice to effectively replace labor-intensive field test and analysis work. It is saving us 167,000 working days a year.”
Intelligent analysis and decision-making are based on more than 10 mature AI models, which have reduced the time taken for root cause analysis and decision-making, by two-thirds. Using more than 50 RESTful open APIs, China Unicom exposes big data and AI capabilities to external systems such as those managing customer experience, churn, complaints and more. Those systems can now interpret customers’ perceptions in the wireless environment, measured against key performance indicators, without on-site testing.
