Optimizing Method for Network Traffic Classification Based on LSTM Network

With the rapid development of Internet technology, network traffic classification has become a key task in the field of network management and security. Accurate traffic classification is crucial for resource allocation, QoS assurance, network security, and other aspects. Although traditional machine learning and deep learning methods have achieved certain results in traffic classification, they have limitations in dealing with the long-term dependencies of time series data. To address this problem, this paper proposes a Multi-LSTM learning framework that incorporates LSTM networks through a multi-task learning framework, combining spatial feature extraction with CNN and temporal dynamic modeling with LSTM to optimize the performance of network traffic classification. This paper first reviews the relevant technologies and challenges of network traffic classification, and then provides a detailed introduction to the model framework. Through experiments on the public dataset QUIC, we demonstrated a significant reduction in the loss values of the model on bandwidth, duration, and traffic category tasks, while improving accuracy. The contribution of this paper is not only to propose a more efficient network traffic classification model, but also to provide new insights on how to use LSTM to process complex time series data. Future work will explore more optimization strategies and validate the effectiveness of the model in a wider range of application scenarios.