Data Production Capability
Data production is the beginning of operation and maintenance. End-pipe-cloud is a conceptual model that began in the mobile Internet era. Diverse access terminals, 5G network pipes, and Internet cloud are the most basic elements for connection in 5G networks and also have profound data value. The meaning of traffic analysis, translation and insight in the terminals, wireless networks, core networks, transport networks and Internet SP is the only way to accumulate raw data. 
At the network pipe side, ZTE VMAX intelligent big data platform performs in-depth analysis of traffic data of the control plane and user plane according to the operator's unified DPI collection and analysis specifications, and generates XDRs, which give the actual meanings to the traffic data. At the terminal side, its access to some terminal SDK data enriches the end-side data sources for joint analysis. 
In the first phase of China Mobile's 5G DPI tender, VMAX DPI occupied DPI projects in more than 20 provinces with a 70 per cent share. With over 15,000 services identification capability, the VMAX platform takes a leading position in the industry, supporting development and quality analysis of abundant service types in the 5G era, such as payment, instant messaging, short video, VR/AR and cloud game. The platform can identify encrypted services in the network with an accuracy of 95%. It identifies some HTTPS services and provides a more thorough insight into network traffic. Driven by diverse markets, the data consumption and production ecosystem is gradually expanding, which promotes powerful data production capability of the VMAX platform and brings infinite possibilities for the value of the data application layer. 

Data Connectivity Capability
Data connectivity can eliminate data islands, and is also a prerequisite for end-to-end service connectivity in the data dimension. End-to-end is a basic way of viewing problems from a global perspective, and end-to-end O&M is a key capability that operators pay close attention to for a long time. In general, the global analysis of the terminal-wireless-bearer-core network-Internet SP link can achieve major end-to-end O&M analysis for operators. In the terminal-wireless sublink, the data connectivity takes the terminal ID such as IMEI, IMSI, and UE IP as the connection point; in the wireless-bearer sublink, it takes the base station ID such as eNB IP and gNB IP as the connection point; in the bearer-core network sublink, it takes the core network device ID such as UPF ID, AMF ID and MME ID as the connection point; and in the core network-Internet SP sublink, it takes the Internet SP ID such as DNS IP and target server IP as the connection point. Monitoring the data of these connection points ensures end-to-end big data connectivity and prepares the data for the next step of global O&M analysis. The data involved in these connection points spans not only different professional fields, but even different manufacturers, thus promoting the key differentiated O&M capabilities of big data products that are different from those in other professional fields.
End-to-end connectivity is a rigid requirement in some typical application scenarios, such as EPS FB voice delay and quality analysis, and user complaint analysis. The EPS FB voice delay and quality analysis involves five domains: 5GC, EPC, IMS, NR and LTE. The connectivity of more than 20 interfaces is an end-to-end analysis request that cannot be fulfilled in a certain field. In the 5G SA network O&M task of an operator, the VMAX platform interconnects with the interface data of the other two vendors based on DPI data, and provides connectivity throughtout the whole EPS FB process. It demarcates the problems in various fields involved in the process and gives a preliminary root cause location, helping the operator increase the EPS fallback success rate to above 98% and shorten the call delay to less than 4 seconds. User complaint analysis is a typical capability that operators pay close attention to. The practice in an operator project shows that the "one-click complaint analysis" function of the VMAX platform helps to demarcate and locate the problem, and the processing efficiency is improved by more than three times. Compared with the actual results, the accuracy of automatic demarcation and location is up to 80%. 

Data Judgment Capability
Management master Peter Drucker said that if there is no measurement, there is no management, which illustrates the importance of measurement. Specifically in network O&M, user percetion is an abstract and comprehensive concept and topic that operators have been concerned about for a long time. Perception can be measured by big data in preparation for better perception. User perception evalution, which combines user experience of using different services, needs to complete the modeling process of KPI and KQI that affect user perception, such as classification and validity filtering, weight setting, and excellent/bad threshold definition, in order to objectively evaluate perception. This needs to be supported by indicator data and distribution data of different service types. Therefore, perception measurement and improvement are important scenarios to reflect data judgment capability. 
Based on VMAX DPI data, KPI and KQI selection and weight settings are made for the perception of web page browsing, videos, games, instant messaging and voice services. In addition, comprehensive perception scores are given based on the poor and excellent quality thresholds of existing network indicators. Specific service categories or APP perception scores can also be evaluated by service, user and region. Big data acts as a referee to provide methods for value-based O&M. The implementation results of an operator project show that more than 40% of the perception problems have been identified and resolved before KPI indicators deteriorate significantly. 

Data Monetization Capability
Data asset monetization is the ultimate form of business models in the digital economy era. With big data, users and networks can be fully presented. Its application scenarios include precise marketing for different users, high-value user mining, ARPU value increase, as well as precise network planning for high-value areas. This helps operators save Capex while guaranteeing network coverage. 
By analyzing and labeling the traffic, the VMAX platform makes a comprehensive analysis of user behaviors such as preference for watching videos, user attributes, and user traffic package, so as to customize the traffic and member packages and precisely push them and reach users. In the practice of an operator project, two months after the rights promotion function went online, the conversion rate of video rights users is 4.7 times that of the traditional method. 

Data Prediction Capability
The most important feature of big data is its ability to make appropriate predictions for the future based on historical data. The current capacity utilization and development predictability are the indicators of whether telecom network resources are sufficient or not, and are the content that operators pay close attention to.
In the practice of an operator project, based on the feature extraction and modeling of the historical development law of wireless data such as PRB, traffic and user base, the regression algorithm is used to predict peak traffic, number of peak users and peak PRB in a certain period in the future. The predicted mean absolute percentage error (MAPE) is lower than 12%, which is at a high level in the industry. The predicted results can help operators make capacity planning decisions. 

Big Data Capabilities in Evolution 

In the evolution of operators' different market segments, virtualized network O&M and autonomous networks, big data as the core support has penetrated into specific service scenarios, continuously looking for focus and releasing data value. 

All-Round Operation Support for CHBN

With the strong evolution of network digitization and intelligence, operators' planning for big data O&M system is gradually approaching to architecture decoupling, capability sharing, data unification and AI empowerment. Big data platform management covers consumer, home, business and new (CHBN) markets. The five major capabilities of big data including production, connectivity, judgment, monetization, and prediction will continue to unleash their unique value, whether in the areas closely related to daily life such as consumer and home markets, in the government and enterprise markets using 5G in vertical industries, or in newly emerging markets. 

O&M Support for Virtualized Networks 
Virtualization and cloudization are the characteristics of the 5G era. NFV/SDN brings the advantages of high resource utilization, high capability coordination and on-demand customization to 5G networks, and also increases the complexity of network O&M. With the big data technology, the service process can be connected from the data dimension, network indicators are evaluated quantitatively, and the AI technology can be used in virtualized networks for failure monitoring, cross-layer root cause analysis, and fault prevention and prediction. The capacity of the virtualized network resource pool can also be monitored and analyzed in real time, and data-based suggestions on the threshold for capacity expansion can be given. To sum up, big data needs to give full play to its differentiated capabilities and thus injects value into O&M of virtualized networks in terms of service continuity guarantee, perception guarantee, and capacity efficiency management. 

Evolution Support for Autonomous Networks
Operators and standardization organizations around the world have issued visions and goals for the evolution direction and steps of autonomous networks (AN). According to the visions of mainstream operators, the evolution requires strong data empowerment. China Mobile has put forward the concept of self-intelligence network, which plans to reach highly autonomous L4 level by 2025, and has made it clear in the goal that it will continue to improve big data capabilities of the data sharing platform under the NM "2+5+N" architecture and promote end-to-end cross-domain service coordination. With the basic idea of intelligence after data connectivity, Vodafone has proposed to add AI capabilities on existing O&M functions and change to unmanned zero-contact network O&M. China telecom has created a digital platform to support its overall cloud-to-digital transformation strategy, and has combined the strategy with its vision of automatic driving network to promote intelligent cloud-network operation and empower new development in the digital economy. China Unicom has built five core capabilities under the intelligent network platform to support the evolution of automatic driving networks, in which the collection of network data plays an important role. 
The mainstream operators have reached a consensus on the idea of planning and building intelligent network O&M capabilities based on big data. This fully reflects the supporting significance of big data's application value to the improvement of network intelligence, and also reflects operators’ rigid demands for the value injected into O&M by big data. 

Conclusion
Data empowerment maximizes value and helps operators lead the 5G era. ZTE VMAX intelligent big data platform will continue to rely on big data's irreplaceable advantage of "new energy". Facing the challenges of increasingly complex network structure, diverse services and high-quality experience, the platform will fully release its unique data value and support the automation and intelligence of the whole life cycle of the network, including planning, construction, maintenance, optimization and operation.