Predicting Network Traffic Flow with a Multi-Layer Short-Term Memory Model

Abstract

In the dynamic realm of digital communication, managing and optimizing network traffic is critical. Efficient forecasting of network traffic volumes, crucial for maintaining network quality, poses significant challenges due to the volatile nature of data flow. This study introduces a novel approach using Long Short-Term Memory (LSTM) networks to enhance forecasting accuracy. We compare a multi-layer LSTM, designed to capture complex dependencies within the data, with a traditional single-layer LSTM, offering insights into their respective capabilities. The multi-layer LSTM’s architecture allows it to excel in understanding deeper temporal patterns, making it a significant advancement over simpler models. By employing these two architectures, our research aims to improve predictive performance in network traffic forecasting, crucial for optimized network management and planning. The findings reveal that the multi-layer model significantly outperforms its single-layer counterpart, demonstrating the potential of sophisticated LSTM networks in practical applications.