Artificial Intelligence and Machine Learning for the fundaments of 5G Network Monitoring

#DestinationAutomation is a six-part campaign featuring videos supported by in-depth white papers, examining how telcos and other businesses can maximize the benefits and opportunities of automation in digital transformation. We’ve already seen automation take off across many industry sectors – from smart cars to smart homes, telemedicine to finance, and retail to field service operations. #DestinationAutomation focuses on real-life use cases, so you can learn exactly how correctly implemented automation can drive your business.

Artificial Intelligence and Machine Learning for the fundaments of 5G Network Monitoring
Artificial Intelligence and Machine Learning for the fundaments of 5G Network Monitoring

Comarch new white paper, entitled “Artificial Intelligence and Machine Learning as the Foundations of 5G Network Monitoring”, looks in-depth at how telcos can automate management processes to achieve optimized results to meet the demands of 5G and the IoT. The paper demonstrates why automation is the only way forward, and explains the importance of high-quality, well-managed data. Read on this article to learn:

  • Challenges for network management arising from 5G and the IoT
  • Artificial intelligence as a solution for network management optimization
  • The importance of good quality data for high-end automated network management
  • Customer and human resource benefits of network management automation


Content Summary

What will you learn?
How network monitoring has developed
The rise of AI and Machine Learning (ML)
It’s time to organize your data
How to make it all happen – Comarch AI control desk
Summary

The 5G mobile network revolution is already happening. Together with the network infrastructure, network management centers have also evolved. They are now integrated with broader projects such as telecommunications networks and digital city initiatives. And, while they are not yet deployed by governments to oversee the smooth running of entire states, theoretical studies are being undertaken and wide-area tests are being carried out. The British city of Bristol is a good example of such a living experiment. The entire urban infrastructure and operations within Bristol’s private infrastructure components have been integrated into the digital city.

Proper management is essential for all systems. This is clearly visible in transport and logistics. We’re all too familiar with the frustrations of congestion in urban areas when certain events (perhaps roadworks or public gatherings) cause traffic to build up beyond expected levels. It’s not so clear in telecommunications, as the notion of a “jam” is rather abstract in this respect. Yet, smooth functioning of mobile networks is vital for our modern, connected, digital world.

What will you learn?

  • How 5G and the IoT are forcing telcos to rethink traditional network management methods
  • The role that artificial intelligence and machine learning can play in network management
  • How to implement and manage automated processes for network monitoring
  • Why it is so important to maintain good-quality data for automated network management
  • The role of technical staff in an automated network
  • How automation can help telcos succeed with customer-focused strategies

How network monitoring has developed

Whether we are considering physical networks, such as transport routes or the abstract connections involved in telecommunications, effective, automated management can help with long and short-term planning to prevent blockages. On one hand, specialists can drill deep down into the network in order to understand the causes of inconveniences and develop solutions to prevent them. On the other hand, continuous monitoring allows transport and telecommunications experts to stay on top of potential issues just before they happen, analyze their effects in real time, and deploy appropriate solutions to minimize damage.

How network monitoring has developed
How network monitoring has developed

Such dual monitoring isn’t new – but in recent years it has become far more sophisticated than it used to be. Technology has made it possible to raise the bar continuously, simultaneously increasing the complexity of challenges. The specialists at network management centers now need a greater depth of knowledge and expertise, they need to be able to share these effectively and efficiently, and they need to be constantly aware that a very short delay or minor error could hinder the operational processes of the network they are employed to oversee.

In some cases, network management specialists develop baselines in relation to historical data, for monitoring “everyday” occurrences. They can observe alarms generated by violations, and determine the necessary steps to be taken to prevent or mitigate network malfunction. Problems can arise, though, when a wholly singular issue emerges. That’s when the specialists’ own experience, contextual knowledge and accumulated skills really come into play – but there are still limitations to what they can do.

In some cases, network management specialists develop baselines in relation to historical data, for monitoring “everyday” occurrences.
In some cases, network management specialists develop baselines in relation to historical data, for monitoring “everyday” occurrences.

So today’s network management centers, and the experts who work there, turn to artificial intelligence and machine learning, particularly useful in the creation of baseline values.

The rise of AI and Machine Learning (ML)

AI enhances traditionally-defined baselines (established on the basis of historical data) by allowing automatic calculations of the probability that a given parameter will be reached or exceeded. In addition, it facilitates the application of different prediction models, any or all of which can feed into calculations.

Machine learning is also being deployed at network management centers for anomaly detection. Unlike traditional methods, which concentrated solely on the concrete values of baseline parameters to establish anomalies, modern techniques bring percentages into play. Combined with the above-mentioned probability factor, this makes today’s network management center systems far more versatile.

For example, adjusting percentages for low values (where, let’s say, an issue that irritates one customer from two can be said to affect 50 per cent of the total) is excellent for violation detection. With AI-based probability algorithms, the system can search for likely causes of the anomaly, select the most appropriate, and act upon it. Thus, the original anomaly event can be determined by allowing the system to work back from potentially multiple symptoms. And this doesn’t apply only to individual anomalies; common violations can be grouped together, which is important from the machine learning perspective because it means the system can learn what action to take in a given situation. Taking into account customer service, this is invaluable for telcos because the system monitors events directly from the network and connected with vital parameters such as KPIs and SLAs.

Four cycles of typical sample improvement
Four cycles of typical sample improvement

The industry has already made great leaps towards these goals, so how long does it take to establish this kind of “next generation” network management? This will depend primarily on the quality of data available.

It’s time to organize your data

As machine learning is based on historical data, such data must be comprehensive and of the highest quality and form. When this requirement is met (and the quality rule must apply to post-factum data too), it is possible to build an effective, automated network management center solution in just three months. However, poorquality data will significantly extend the time required for implementation, as far more human effort will be needed at the initial stages in order to properly collate and store data for machine analysis.

It follows, then, that even if artificial intelligence and machine learning are planned for some time in the future, the time to start organizing data is now. What’s more, telecoms operators are unlikely to get the best from 5G if they don’t adapt their network management processes to integrate AI and ML very quickly – which makes it even more urgent to address any issues with the availability, quality and form of data, right away.

How to make it all happen – Comarch AI control desk

Comarch Artificial Intelligence (AI) Control Desk puts the power of these AI and ML tools at the fingertips of telcos, supporting them in data collection, analysis and root-cause detection and resolution. As long as an anomaly has appeared once in an analyzed dataset, the system can identify it and take or recommend action automatically. This reduces the level of operator interaction required (and thus lowers the risk of human error in network management overall), freeing specialists from the need to carry out recurring tasks. The probability factor means that completely new events can also be identified, although in such cases human intervention is required – if only to confirm the root-cause of an incident and verify the solution for first and future occurrences.

This all feeds into Comarch’s vision of network management of the future. In this vision, networks will by and large be monitored and maintained by automatic – maybe even autonomous – systems, while the specialists will be free to concentrate on tactical, strategic and business decision-making. Of course (as noted above), there will be situations where the human touch is required, but automation is clearly the way forward, and advances in machine learning are paving the way for just such a future.

How to make it all happen - Comarch AI control desk
How to make it all happen – Comarch AI control desk

At Comarch, we have examined the effectiveness and efficiency of artificial intelligence in projects the company has implemented around the world. Our experts have compared process execution by averaging measured parameters before machine learning implementation and afterwards, under both real-life and laboratory conditions. It’s worth noting that there is still work to be done as they are derived from early-deployment situations in which ML is working intensively to refine and improve algorithms.

Summary

To return to our original concepts of networks as physical (transport) and abstract (telecommunications), there is one major difference to take into account. While transport networks can be managed automatically, the physical elements – vehicles and roads – do not change much in terms of their concrete reality; a traffic jam is a traffic jam, and no amount of automated anomaly detection and resolution is going to magically remove your car from the queue or create a new lane.

With 5G networks, especially given the potential for network slicing, this kind of “magic” is entirely possible; the route that data take between one piece of hardware (your phone, tablet or IoT device, for example) and another can be altered and optimized at any moment by a properly implemented network management solution, with little or no need for human intervention. This is why now is the best time to start using artificial intelligence in managing telecommunications networks, if you have not done so already.

Source: Comarch