Rakuten Symphony has developed a prioritized list of the top 25 AI initiatives for mobile network operators (MNOs) and divided them into five categories. The Network Capacity Forecasting AI Model described in this blog post falls under level 3 automation of capacity forecasting and helps RF engineers to plan site capacity more accurately so that mobile network user experience is not impacted.
Capacity management for MNOs is an engineering function that ensures the network infrastructure has the capacity to handle data throughput for today’s users and has the scalability to handle expanded workloads in the future.
Without an AI model, network engineers use a rule-based approach where alarms are triggered when the network capacity needs adjustment. This approach is not very accurate and can lead to capacity issues that persist until an engineer can analyze the issue and respond.
Another legacy capacity planning approach uses simulation models that are dependent on network what-if scenarios. Each scenario has a limited scope of operations and can only monitor a small part of the network. Thus, multiple scenarios are required to provide the required coverage network wide.
In both of these methodologies, network data is stored in a database with KPIs executed using rule-based what-if analysis. The limitations of this reactive process include:
The primary function of the Rakuten Symphony Capacity Forecasting AI Model is to predict data traffic levels for cells deployed in the network using machine learning and deep learning algorithms. It currently predicts future traffic, users, physical resource blocks (PRB) and capacity based on historical data.
With the Capacity Forecasting AI Model, Rakuten Symphony is automating the capacity management function to improve service quality and efficiency. The capacity forecasting model also helps network engineers to predict the direction and quantity of network capacity, monitor network performance and optimize available network resources.
Using this AI model, network capacity planners get the following benefits:
Training data for this model comes from real network traffic that is built into Rakuten AI, our AI platform that is optimized for MNO network operations. The AI model has continuous pipelines into the KPI data and provides anomaly and forecasting responses at regular intervals.
One use case for the Network Capacity Forecasting AI Model is RAN capacity forecasting. This AI model allows engineers to seamlessly analyze historical data and user trends to predict future RAN usage KPIs. This data is automatically shared with Rakuten Symphony’s RAN Commander application allowing users to visualize the forecasts graphically to further simplify RAN capacity optimization.
In fact, automated RAN capacity forecasting is being successfully used by the several capacity planning teams in Rakuten Mobile.
Network capacity forecasting and management is an engineering function that is vital to monitoring and ensuring the network will not collapse under the weight of the telecommunication service demands of a growing number of users, considering this model is running and helping network operations and expansion planning when the network is already in production. This allows the network engineer to set accurate KPIs and avoid customer service issues related to network capacity.
The Rakuten Symphony Network Capacity Forecasting AI Model is a solution that accommodates today’s KPIs and is scalable for the future. It stands out because it is trained on real-world network data and integrates with other OSS tools.
