Network Quality Scoring Model for Real-Time Impairment Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
End users and Multiple System Operators (MSOs) face difficulties in determining the root cause of network issues, such as slow internet or video freezing, due to the lack of communication between the end user's application and the network, leading to frustration and difficulty in resolving impairments.
Innovation Solution
A scoring model using a machine learning algorithm that combines application Key Performance Indicators (KPIs) and network data to generate an end user score, which is compared to predetermined thresholds to identify impairments, determine their type and location, and suggest resolution actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If network quality measurements are collected from multiple sources (application KPIs and network data), then the accuracy of impairment detection is improved, but the system complexity increases
Solution Approach 1:
The system segments network quality assessment into two distinct measurement sources: application KPIs (collected via first API from end user applications) and network data (collected via second API from network infrastructure). Each source is independently collected, processed, and weighted, allowing the system to maintain high measurement precision while managing complexity through modular data collection architecture
Solution Approach 2:
The scoring model acts as an intermediary that receives measurements from multiple sources (application KPIs and network data), processes them through machine learning algorithms, and generates a unified end user score. This intermediary component integrates diverse data sources without requiring direct complex interactions between them, thereby improving detection accuracy while containing system complexity
2Productivity
If a machine learning scoring model is implemented to generate end user scores, then the ability to identify and resolve impairments is improved, but the computational resources required increase
Solution Approach 1:
The scoring model is trained in advance using labeled training data that includes historical measurements and known impairment conditions. This preliminary training phase allows the model to learn optimal weighting and patterns offline, so that during runtime, the model can quickly generate end user scores with minimal computational overhead, improving impairment resolution efficiency while managing online computational resources
Solution Approach 2:
The system changes the parameter representation by transforming multiple raw measurements (application KPIs and network data) into a single aggregated end user score through the scoring model. This parameter transformation consolidates complex multi-dimensional data into a simplified metric that is easier to interpret and act upon, improving productivity while reducing the computational burden of analyzing multiple individual parameters
3Reliability
If real-time measurements are processed through the scoring model, then the responsiveness to network issues is improved, but the processing time increases
Solution Approach 1:
The system continuously collects measurements from application KPIs and network data sources, continuously processes them through the scoring model, and continuously generates updated end user scores. This continuous operation ensures that the system remains responsive to network issues in real-time, improving reliability while optimizing processing throughput to minimize time loss through efficient batch or stream processing
Data Source
AI summary
Systems and methods for quantifying network quality using a scoring model are provided. At least one measurement is received by an analysis agent and inputted into a scoring model. The scoring model comprises a machine learning model configured to receive the at least one measurement, input the at least one measurement into a plurality of functions, weight the output of each function by a corresponding weight of a plurality of weights, and combine the weighted output to generate an end user score. The end user score is compared to a predetermined threshold.