Hierarchical Network Traffic Prediction Model for Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network traffic prediction methods, such as those based on time series and linear programming, are inadequate for accurately predicting modern network traffic due to their sensitivity to historical periodicity and insufficient consideration of influencing factors, leading to over-fitting and inaccurate predictions.
Innovation Solution
A multiple layer network traffic prediction model is established, comprising a higher-level linear model and lower-level models, where independent variables are decomposed and substituted to create a complete prediction model, exploring the impact of service level variables on bandwidth values, and incorporating error compensation to account for periodic changes and service characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a time series model is used for network traffic prediction, then the model structure is simple, but the prediction accuracy deteriorates when network anomalies or strong human intervention occur due to sensitivity to historical periodicity
Solution Approach 1:
The patent segments the network traffic prediction into multiple hierarchical levels: global trend prediction and local periodic prediction. The global model captures long-term trends while the local model handles short-term periodic fluctuations, resolving the contradiction by dividing the prediction task rather than using a single monolithic model
Solution Approach 2:
The patent introduces a hierarchical dimension to the prediction model, adding global-local decomposition and multi-level aggregation. This transforms the single-dimensional time series prediction into a multi-dimensional hierarchical structure that can simultaneously capture both trends and periodicities without being overly sensitive to historical anomalies
2Device complexity
If a linear programming model is used for network traffic prediction, then the model is simple, but it fails to consider sufficient traffic influencing factors and periodicity fluctuations, leading to over-fitting
Solution Approach 1:
The patent segments the prediction model into multiple independent components operating at different hierarchical levels. Each level handles specific aspects of traffic prediction (global trends, local periodicities), allowing the system to consider sufficient influencing factors without requiring a single overly complex model
Solution Approach 2:
The patent makes the model dynamic by allowing different prediction strategies at different hierarchical levels and time scales. The global model adapts to long-term changes while local models adapt to short-term periodic fluctuations, preventing over-fitting through adaptive multi-level processing
3Adaptability or versatility
If existing traffic prediction methods are used, then they can handle basic prediction needs, but they cannot adapt to the diversification of network traffic properties caused by dramatic increase in network service categories
Solution Approach 1:
The patent creates a universal hierarchical prediction framework that can handle diverse network traffic properties through multi-functionality. The same hierarchical structure adapts to different service categories and traffic types by adjusting the global and local models, providing both adaptability and precision across varied traffic scenarios
Data Source
AI summary
The present disclosure provides a network traffic prediction method, an apparatus and an electronic device. In the present disclosure, the network traffic prediction method includes establishing a higher-level network traffic prediction model, where the higher-level network traffic prediction model includes a linear model; using independent variables of the linear model in the higher-level network traffic prediction model as dependent variables, establishing at least one layer of a lower-level network traffic prediction model, and decomposing independent variables in the lower-level network traffic prediction model till all independent variables of each lower-level network traffic prediction model are acquired naturally; and substituting a calculation model of the lower-level network traffic prediction model into the higher-level network traffic prediction model to obtain a complete network traffic prediction model.