What is the role of Machine Learning for an intelligent Network Investment Planning
“Capacity, not speed is the true 5G revolution”
Dave Dyson, 3 UK CEO
This thought-provoking statement has a tremendous resonance for me, not just from a 5G perspective, but for any technology that promises to offer a true broadband experience. It roused my curiosity to think about “Capacity” in detail and find answers to questions like: What does this mean to communication service providers (CSPs)? How is capacity being addressed from a technology perspective – today and in the future? And most importantly, how can capacity (or lack thereof) impact end user experience? Let’s look at these questions in isolation.
Technology Evolution: CSPs have seen significant evolution of technologies over the past few decades. While the original digital 2G networks were focused on providing improved voice capacity, 3G networks introduced basic mobile data connectivity with a substantial increase in voice and data capacity. Over the last decade, we witnessed LTE networks offering increased data speeds using IP architectures with significant improvement in spectral efficiency, resulting in an exponential increase in data consumption. From the network perspective, this fuelled even more capacity demand.
The other important aspect of technology evolution comes from the end user experience perspective.
User Experience: Rapid growth in usage of the network and heightened user expectations led 3GPP to come up with multiple releases (rel 8 to rel 14) for LTE. These helped in accelerating the evolution of LTE technology. Now with 3GPP release 15+, the subsequent move to 5G is mainly driven by the changing demands of consumers and enterprises. Mobile end users are demanding more network capacity, more data bandwidth and higher mobile data speeds with very low latency. They want to use their mobile devices in more places and have more devices connected to the networks. ITU predicts a huge upsurge in data traffic, mainly video traffic, which will require an immense capacity “blanket” to meet the end user expectations.
Figure 1: Estimation of mobile traffic (in Exabyte) by different service types globally by ITU
Such a tremendous technology evolution and corresponding end user expectations bring in a lot of challenges to CSPs. However, I see these challenges as an ocean of opportunity, if addressed in an efficient way. Let’s explore some key related questions that need to be asked:
2-Part Webinar Series on Intelligent Network Investment Planning
- What are the primary challenges and the associated opportunities that CSPs need to address?
Strategically, CSPs leverage capacity-driven investments to gain competitive advantage and enhance end user experience. Most of the time, such investments are highly cost intensive, both in terms of Capex and Opex. Any inaccuracies in the overall network investment planning can lead to unintended impacts on company financials and can create significant network inefficiencies. Such inaccuracies eventually impact end user experience and erode competitive advantage.
Conventionally, capacity investment plans require substantial amounts of historical data and consumes plenty of time in stitching the data for forecasting. Most of the time, it takes multiple iterations to test different scenarios. Moreover, inputs across multiple domains often tend to be overlooked, which could lead to further increase in iterations. Collectively it makes the overall process cumbersome and brings further inefficiencies as well.
So, the next obvious question is, how can we come up with a network investment plan which is accurate, has optimal cost, meets current & future requirements of users and gives competitive advantage over a long term?
- Is there a better approach to capacity management and planning?
In my view, cross domain analytics is quintessential for an accurate and effective investment plans. Establishing seamless alignment and strong correlation between data from different domains like network, user, business, marketing and operations will give CSPs a better understanding of how consumers and enterprises are using their mobile network. It also helps them to understand how capacity can be optimized and how the user base can be segmented based on behaviour (not simply demographics).
A strong data driven correlation between domains will help CSPs unlock hidden revenues from their networks, as well as understand which services are being used and valued (and those that are not). And most importantly, it can yield deep analytical insights for smart decision making towards future network investments.
Figure 2: Cross domain network view
- What are the key enablers to achieve an efficient network investment plan?
I believe there are 3 key pillars that can help CSPs overcome challenges in achieving an efficient network investment plan.
Cross domain data: Cross domain historical data (in years) is one of the basic requirements to achieve an efficient plan. With perhaps billions of records flowing, the plan will need seamless big data handling capabilities which could efficiently extract relevant and meaningful attributes for analysis. An intelligent handling of abnormal values can be leveraged to get desired aggregated data as per the requirement. I call it the most basic and important step, this will lay a strong foundation for later steps in the process.
Machine Learning: A good set of processed and aggregated data, taken as an input to train machine learning algorithms, will help in establishing the right forecasting model(s) for specific scenario(s). Such a fine-tuned model(s) can then be fed as a critical input for smart network investment plans, yielding optimized business decisions. Machine learning is a must to handle cross domain data efficiently.
Visualization: Complex analytical outcomes can be simplified with interactive visualization. It can enable and empower decision makers to identify relevant and meaningful patterns for decision making. A strong visualization capability can further help CSPs focus on areas most likely to influence their key business objectives.
I truly believe these enablers can bring significant efficiency in overall network investment plan process. It can help CSPs fulfil the demanding end user expectations proactively, while giving them a huge long-term competitive advantage.
To understand more on network investment planning, join our 2-part webinar series on “Intelligent Network Investment Planning”.
Alok Kumar Sinha, Product Director-Network Analytics, Subex, has around 16 years of experience in the telecommunications industry in roles spanning RF planning & optimization, pre-sales, consulting, and product management. He enjoys the challenge of simplifying network planning and optimization complexities and loves to explore new technologies to establish strategic thought-leadership, innovation, and ideation.