Network congestion control using a hybrid machine learning.

Author

Chamwaura, Fidel

Title

Network congestion control using a hybrid machine learning.

Abstract

This research presents the design and implementation of a hybrid machine learning system aimed at detecting and managing network congestion in real-time. The system combines both supervised and unsupervised learning techniques—specifically, the Random Forest Classifier and Isolation Forest algorithm—to leverage their complementary strengths. While Random Forest effectively classifies known congestion patterns based on historical data, Isolation Forest is employed to identify anomalies that may signal previously unseen or emerging congestion behaviors. This two-pronged approach improves the system’s ability to handle both routine and unpredictable traffic conditions.
Data for model training and evaluation was sourced from the UNIBS Campus Network Traffic Repository, offering realistic and diverse traffic scenarios. Following preprocessing and feature engineering, the models were integrated into a user-friendly web interface developed using Flask and Streamlit. The application supports manual input, real-time traffic monitoring, batch prediction, and automatic logging for model retraining, making it suitable for both technical and non-technical users.
The system was evaluated using standard performance metrics such as accuracy, precision, recall, and F1-score. Results demonstrated that the hybrid model outperformed individual algorithms in detecting congestion with higher accuracy and resilience. With its real-time monitoring capabilities, adaptive design, and practical interface, the proposed system offers a scalable and intelligent solution for modern network management.

Date

June 2025

Publisher

BUSE

Keywords

hybrid machine learning system

network congestion in real-time

Random Forest Classifier

Isolation Forest algorithm

Supervisor

Mr. J Musariwa