A Comparative Study of Traffic Classification Techniques for Smart City Networks

Area of Science:

• Computer Science
• Network Engineering
• Artificial Intelligence

Background:

• Smart city networks require robust Quality of Service (QoS) management.
• Conventional traffic classification methods struggle with dynamic ports and encryption.
• Machine learning offers an intelligent alternative for network traffic management.

Purpose of the Study:

• To evaluate machine learning algorithms for network traffic classification in smart cities.
• To compare the performance of machine learning against traditional port-based methods.
• To identify the most effective machine learning algorithm for this task.

Main Methods:

• Applied four supervised machine learning algorithms: Support Vector Machine, Random Forest, K-Nearest Neighbors, and Decision Tree.
• Utilized a port-based traffic classification method for comparison.
• Evaluated algorithm accuracy and performance based on classification results.

Main Results:

• The Decision Tree algorithm achieved the highest average accuracy at 99.18%.
• Machine learning methods significantly outperformed the conventional port-based classification.
• Intelligent network functions improved overall network management capabilities.

Conclusions:

• Machine learning, particularly the Decision Tree algorithm, is highly effective for smart city network traffic classification.
• This approach enhances network management and QoS support.
• Machine learning provides a more accurate and performant solution compared to port-based methods.