A Lightweight Framework for Dynamic Bandwidth Allocation Using Linear Regression in Latency-Sensitive Networks

Managing bandwidth dynamically has become increasingly important in today's communication networks, especially for services like video calls and streaming that are sensitive to delays. Traditional static methods often struggle to adapt to fluctuating network loads and evolving user demands. This paper introduces a lightweight, regression-based model for predicting bandwidth requirements and allocating resources dynamically based on user activity. The proposed solution was tested through a Python-based simulation, where user activities such as video calls, streaming, and downloads were prioritized based on their sensitivity to latency. The simulation demonstrated that high-priority tasks (e.g., video calls) consistently received maximum bandwidth (۱۰ Mbps), ensuring smooth real-time performance. Meanwhile, mid-priority (e.g., streaming) and low-priority (e.g., downloads) tasks utilized the remaining capacity efficiently. Compared to more complex models like LSTM and SVR, this approach achieved similar accuracy (RMSE = ۰.۰۵) while maintaining significantly lower computational costs. Thanks to its simplicity and efficiency, the proposed model is particularly suited for resource-constrained environments, such as IoT and mobile networks. Future improvements will focus on integrating real-time contextual data to further enhance prediction accuracy.

Keywords:

bandwidth allocation , latency-sensitive services , network resource management , user activity , real-time applications , lightweight prediction