Network Data Fill Selection Using Missing-Data Error Curves

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The presence of missing or incomplete network data poses challenges in accurately assessing network performance and applying advanced analytics techniques, such as machine learning, due to factors like system outages or data collection issues.

Innovation Solution

A method is described that involves identifying network elements with less than a specified percentage of missing data, applying various data fill methods, calculating the percentage of missing data successfully filled and error, generating error curves, and selecting the most appropriate data fill method based on these curves to train machine learning models effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If data fill methods are applied to missing network data, then data completeness is improved, but measurement precision deteriorates due to error introduction

Engineering Contradiction:

Improvedata completenessVSAvoiddata accuracy

Core Design Contradiction:

Loss of informationVSMeasurement precision

Solution Approach 1:

The system calculates error metrics for each data fill method and uses this feedback to select the optimal method. Error curves are generated by incrementally removing data and measuring fill accuracy, then this feedback is used to choose methods that minimize error while maximizing data completeness for machine learning training.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system varies the percentage of missing data (parameter) to generate error curves that show how different fill methods perform under different missing data conditions. This allows selection of methods whose error characteristics are acceptable for the specific application's tolerance levels.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If multiple data fill methods are evaluated, then selection accuracy is improved, but device complexity increases

Engineering Contradiction:

Improvemethod selection accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The evaluation process is segmented into discrete steps: remove a percentage of data, apply fill methods, calculate error metrics, and generate curves. This segmentation makes the complex evaluation manageable and automated through systematic iteration over different missing data percentages.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Error curves are pre-calculated for multiple fill methods across different missing data percentages before actual model training. This preliminary evaluation allows the system to select the best method in advance, avoiding the need for complex real-time comparisons during deployment.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12587457B2Systems and methods for calculating a synthetic data error factor to simulate missing network element data

Publication Date: 2026.03.24 VERIZON PATENT & LICENSING INC

AI summary

The techniques described herein relate to a method that involves identifying a set of network elements with less than a specified percentage of missing data, removing a percentage of data to create a test data set for each network element, applying various data fill methods to the test data set to generate filled data sets, calculating the percentage of missing data successfully filled and the error between the filled data and the removed data for each data fill method, analyzing the relationships between the percentage of missing data, the percentage of successfully filled data, and the error to generate error curves for each data fill method, and selecting one or more data fill methods to apply to missing data in a network element based on the corresponding error curves.