Local Network Connectivity Demand Prediction via ML Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information technology infrastructure management systems struggle to accurately predict network local connectivity demands due to limited demand predicting capabilities and failure to consider the reasons behind infrastructure usage, leading to suboptimal infrastructure configurations.
Innovation Solution
The implementation of local network agents associated with network devices, utilizing machine learning models trained by a central authority for long-term predictions. These models are retrained and updated during the prediction term to enhance accuracy, and the predictions are adapted for resource-constrained devices through compilation into simpler algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are trained locally on network devices, then prediction accuracy improves, but device complexity and computing resource requirements worsen
Solution Approach 1:
The system divides the machine learning workflow into two segments: centralized model training performed by a cloud server, and local model execution performed by network devices. This segmentation allows complex training operations to be performed remotely while keeping local devices simple, resolving the contradiction between prediction accuracy and device complexity.
Solution Approach 2:
A centralized training server acts as an intermediary between data collection and local prediction execution. The server collects training data from multiple sources, performs complex model training, and distributes the trained models to network devices. This intermediary handles the computational complexity centrally while enabling accurate local predictions.
2Productivity
If monitoring agents report frequently according to determined operational parameters, then network demand analysis capability improves, but network bandwidth consumption and system complexity worsen
Solution Approach 1:
The system performs preliminary model training in advance using historical data collected during normal operations. The trained models are then deployed to network devices, enabling them to perform accurate demand predictions without requiring frequent data reporting. This preliminary action reduces ongoing bandwidth consumption while maintaining high analysis capability.
Solution Approach 2:
Network devices use the deployed machine learning models to perform self-service predictions of their own connectivity demands. Instead of requiring continuous external analysis through frequent agent reporting, each device autonomously predicts its demand using locally stored models, significantly reducing network bandwidth consumption while maintaining productivity.
Data Source
AI summary
The network local connectivity demand prediction method (100), comprises:a step (105) of long-term local connectivity demand prediction, by a machine learning model operated by a local agent,in a duration equivalent to the duration of the long-term prediction, more than one short-term iteration of:a step (110) of aggregating values representative of, at least:a connectivity demand of the network comprising at least a physical network device,a purpose of use of the network comprising at least a physical network device,a local long-term connectivity demand prediction for the physical network device,a step (115) of training a machine learning device to produce a machine learning model to provide a long-term local connectivity demand prediction, anda step (120) of deploying the machine learning model, anda step (125) of updating the long-term local connectivity demand prediction.