Network Traffic Prediction Incorporating Prior Knowledge for an Intelligent Network.

Area of Science:

• Computer Science
• Network Engineering
• Artificial Intelligence

Background:

• Accurate network traffic prediction is vital for managing Internet of Things (IoT) services.
• High burstiness in intelligent network traffic poses challenges for traditional prediction models.
• Existing models struggle with feature extraction from limited data and lack interpretability.

Purpose of the Study:

• To develop an interpretable network traffic prediction model that overcomes limitations of traditional and deep learning approaches.
• To enhance the accuracy and reliability of traffic prediction in IoT environments.

Main Methods:

• Proposes a fusion prior knowledge network incorporating self-similarity as a priori knowledge.
• Integrates self-similarity into the gating mechanism of a long short-term memory (LSTM) neural network.
• Combines a one-dimensional convolutional neural network (1D-CNN) with an attention mechanism for temporal feature extraction.

Main Results:

• The proposed model effectively recovers original data characteristics and better describes network traffic trends.
• Achieved an absolute correction factor of 76.4%, outperforming traditional statistical models by over 10%.

Conclusions:

• The fusion prior knowledge network offers superior performance and interpretability for network traffic prediction.
• This approach enhances the quality of service in IoT by providing more accurate traffic forecasts.