With network development moving from 2G to 3G to 4G and now shifting to 5G, there will be an explosion in traffic in the next few years, due in part to the operators’ zero-rating policy and the changing behavior of subscribers. Intelligent optimization and prediction for increasing network complexity and adaptive capacity will be a high priority. Done well, it will enable reduced cost and optimized investment.
AI technology accelerates the optimization initiatives
Some key questions include:
- How can network resources be assigned and automatically optimized in response to user density?
- How can close-loop performance diagnostics and self-healing be realized?
The answers will certainly have a direct bearing on labor costs and operators’ network operating expenditure (OPEX). Because of this opportunity, ZTE has been pursuing AI technology in order to realize automatic close-loop optimization. This helps Telecoms reduce the cost and labor to a great extent, as shown in figure 1-1.
ZTE developed their AI platform, which is integrated with several of typical machine learning algorithms. These can be easily used to improve the potential for optimization and resolve the tough issues.
It can be summarized in the following steps:
- Data preparation by DT, MR/CDT or OMC
- Data preprocessing
- Feature extraction based on the expert experience
- Iterative model exploration and evaluation by integrated algorithm
- Prediction model applied to the optimization process to enable automation
Figure 1-1: Close-loop auto optimization integration with AI technology
ZTE thinks that this optimization process, driven by AI algorithms is going to change and accelerate this optimization to a great extent in the coming years, without human intervention.
ZTE have developed advanced machine learning capabilities on optimization
To benefit from predictive and prescriptive analytics one requires a combination of relevant experience, skills and expertise. ZTE has this in abundance in the telecoms arena, and utilizing our experience and our people we are helping operators globally with our machine learning capabilities.
Some examples of how we help operators through machine learning include:
Root Cause Analysis of VoLTE – reduction of MTTR
In order to minimize the deployment cost, the key is to detect the wireless issues without probes deployed from a wireless node. To this end, our VoLTE solution has developed the outstanding correlation with a clustering algorithm, which enables the identification of coverage issues with only Gm and S1 interface data. The accuracy has now been tested to be as high as 85%. This feature will soon be applied in the telecoms market once further improvement and tests have been carried out.
Indoor and outdoor user identification – assuring the indoor user experience
MBB development drastically drives up data demands, with 90% of time spent and 80% of data traffic generated indoors. For better coverage quality and user experience in-building, indoor and outdoor user identification is the first requisite.
ZTE has collected thousands of data samples for indoor and outdoor users for machine learning and modeling. ZTE found that rsrp, rsrq, ta_calc index are the typical characters being used for training through the GBDT algorithm and the accuracy is as high as 92.5%. The module is already used in our application for close-loop planning and optimization, such as 3D planning, precise indoor coverage optimization and virtual drive test indoor etc.
Machine learning is applied to data to help operators with this growing challenge. By utilizing complex data models and algorithms to find patterns in data, operators are able to manage predictive analysis and auto optimization effectively.
This sponsored article was brought to you by ZTE