Scholarlink

The advent of 5G networks and the proliferation of Internet of Things (IoT) devices are driving an unprecedented surge in demand for high data throughput, ultra-low latency, and robust network reliability. These technological advancements present significant challenges to traditional mobile network capacity planning models, which often fail to adapt to the dynamic and nonlinear behaviors of modern networks. With the rapid increase in connected devices and data consumption, mobile network operators (MNOs) must adopt more innovative approaches to optimize resource allocation and minimize costs. This paper introduces a novel hybrid approach by integrating System Dynamics (SD) modeling with Predictive Analytics and an unsupervised learning-based clustering framework to tackle the complexities of 5G network capacity planning. The SD model provides a comprehensive mechanism to simulate intricate network interactions such as available bandwidth (stocks), data traffic (flows), latency (delays), and essential feedback loops. By complementing this with predictive analytics, the hybrid model enables both accurate forecasting of future network demand and real-time adaptation to changing network conditions. Additionally, we propose a base station (BS) clustering framework, assisted by real-world data, to identify high-traffic areas (HTC) critical for 5G deployments. Using the NetClustering algorithm, the appropriate number of clusters (k) is determined based on traffic density (MB/km \(^2\) ) and target deployment area (km \(^2\) ), fulfilling MNOs’ requirements for cost-effective network expansion. The algorithm demonstrates superior performance when compared to traditional methods such as the Elbow heuristic, providing higher network utilization and significant cost savings. Through a series of simulated scenarios, including the rapid growth of IoT devices usage, this hybrid approach shows an enhanced ability to manage changing network demands, ensuring optimized resource utilization and improved long-term capacity planning. The results underscore the value of integrating SD, predictive analytics, and unsupervised learning to create a resilient and efficient framework for future 5G network infrastructures.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Mobile Network Capacity Planning with Emerging Technologies: A System Dynamics and Machine Learning-Based Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jean-Claude Mudilu Kafunda,
  • Mohamed Alsisi,
  • Kelvin Egbine,
  • Witesyavwirwa Vianney Kambale,
  • Ange Taboria Baelongandi,
  • Nicodème Kabongo Lungenyi,
  • Marivaux Nzaji Bampende,
  • Kyandoghere Kyamakya

摘要

The advent of 5G networks and the proliferation of Internet of Things (IoT) devices are driving an unprecedented surge in demand for high data throughput, ultra-low latency, and robust network reliability. These technological advancements present significant challenges to traditional mobile network capacity planning models, which often fail to adapt to the dynamic and nonlinear behaviors of modern networks. With the rapid increase in connected devices and data consumption, mobile network operators (MNOs) must adopt more innovative approaches to optimize resource allocation and minimize costs. This paper introduces a novel hybrid approach by integrating System Dynamics (SD) modeling with Predictive Analytics and an unsupervised learning-based clustering framework to tackle the complexities of 5G network capacity planning. The SD model provides a comprehensive mechanism to simulate intricate network interactions such as available bandwidth (stocks), data traffic (flows), latency (delays), and essential feedback loops. By complementing this with predictive analytics, the hybrid model enables both accurate forecasting of future network demand and real-time adaptation to changing network conditions. Additionally, we propose a base station (BS) clustering framework, assisted by real-world data, to identify high-traffic areas (HTC) critical for 5G deployments. Using the NetClustering algorithm, the appropriate number of clusters (k) is determined based on traffic density (MB/km \(^2\) ) and target deployment area (km \(^2\) ), fulfilling MNOs’ requirements for cost-effective network expansion. The algorithm demonstrates superior performance when compared to traditional methods such as the Elbow heuristic, providing higher network utilization and significant cost savings. Through a series of simulated scenarios, including the rapid growth of IoT devices usage, this hybrid approach shows an enhanced ability to manage changing network demands, ensuring optimized resource utilization and improved long-term capacity planning. The results underscore the value of integrating SD, predictive analytics, and unsupervised learning to create a resilient and efficient framework for future 5G network infrastructures.