ML-Based Network Capacity Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless service providers face challenges in consistently measuring network performance across a network with components from various manufacturers, as different base stations report inconsistent and insufficient metrics, making it difficult to assess performance from a customer perspective and identify issues affecting specific service classes.
Innovation Solution
Implementing a user experience metrics service that generates additional metrics using machine learning models to characterize user experience, collects application-specific data usage information, and prioritizes resource allocation to improve network performance by identifying areas with lower user experience metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional metrics collection from base stations is used, then network performance measurement is simple, but the metrics are inconsistent and insufficient across different manufacturers
Solution Approach 1:
The patent introduces a machine learning model as an intermediary that processes metrics from multiple base station manufacturers. The model learns manufacturer-specific metric patterns and transforms them into a unified user experience assessment, enabling consistent performance measurement across diverse hardware sources without requiring changes to the base stations themselves.
Solution Approach 2:
The system transforms traditional network metrics into new derived metrics that better represent user experience. By changing the parameter representation from raw base station metrics to ML-generated user experience scores, the system achieves both manufacturer compatibility and improved measurement accuracy for customer-perceived performance.
2Reliability
If comprehensive metrics are collected from all base stations, then network performance can be thoroughly assessed, but the data complexity and processing burden increase
Solution Approach 1:
The patent extracts only the essential features needed for user experience assessment from the comprehensive set of base station metrics. The machine learning model identifies and extracts the most relevant parameters while discarding redundant information, reducing processing complexity while maintaining reliable performance assessment.
Solution Approach 2:
Instead of processing all raw metrics directly, the system creates simplified copies or representations of the data through machine learning models. These model-generated metrics capture the essential performance characteristics in a more manageable form, reducing computational burden while preserving assessment reliability.
3Productivity
If resources are allocated based on traditional metrics, then network management is straightforward, but user experience issues may be missed
Solution Approach 1:
The patent implements a feedback mechanism where machine learning models continuously analyze user experience metrics and provide insights back to the network management system. This feedback loop enables dynamic resource allocation decisions that directly address user experience issues, improving both allocation efficiency and measurement precision through iterative optimization.
Data Source
AI summary
Systems and methods are described for monitoring performance and allocating resources to improve the performance of a wireless telecommunications network. A wireless telecommunications network may be comprised of base stations and other infrastructure equipment, which may be sourced from various suppliers. Users may generate traffic on the wireless network, and performance metrics relating to the user experience may be collected from individual base stations. The set of available metrics for a particular base station may vary according to the supplier. A machine learning model is thus trained using metrics from multiple base stations, and used to estimate values for metrics that a particular base station does not provide. The metrics are then further characterized according to the service classes of the users, and resources for improving the performance of base stations are allocated according to the reported and estimated metrics for various service classes.