Network Fault Identification via Customized Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional network technologies cannot automatically diagnose and remediate network faults, relying on manual identification of impacted flows, primary switches, and underlying events, which is inefficient and prone to overfitting with large datasets.
Innovation Solution
A customized set of feature vectors is created using historical data to target likely network components, reducing overfitting by selecting features based on probability thresholds and root cause analysis, which are then input into a machine learning algorithm to predict network faults and their components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional networks send large amounts of data for fault identification, then more information is available for diagnosis, but the system cannot extract relevant intelligent information and manual identification becomes inefficient
Solution Approach 1:
The patent extracts relevant intelligent information from large amounts of network data by creating customized feature vectors that selectively capture only the most pertinent characteristics. This extraction process filters out irrelevant data while preserving the essential information needed for fault identification, thereby improving both information quality and processing efficiency.
Solution Approach 2:
The patent transforms raw network data into a different parameter representation through feature vector creation. By changing the parameters from raw data format to structured feature vectors with specific characteristics, the system enables more effective fault identification while reducing the computational burden of processing large datasets.
2Measurement precision
If manual identification of flows, switches, and events is used, then detailed analysis is possible, but the process is inefficient and time-consuming
Solution Approach 1:
The patent performs preliminary action by pre-processing network data into customized feature vectors that are ready for immediate fault analysis. This preliminary structuring of data includes identifying relevant characteristics and organizing them in a format optimized for fault identification, thereby reducing the time required for actual fault diagnosis while maintaining high accuracy.
Solution Approach 2:
The patent introduces an intermediary layer between raw network data and fault identification - the customized feature vectors. This intermediary structure processes and organizes data in a way that bridges the gap between raw information and diagnostic requirements, enabling faster and more accurate fault location without direct manual analysis.
3Reliability
If large datasets are used for fault analysis, then more comprehensive coverage is achieved, but overfitting occurs and reduces prediction accuracy
Solution Approach 1:
The patent applies local quality by creating feature vectors that emphasize locally relevant characteristics specific to each fault scenario. Instead of treating all data uniformly, the system identifies and weights the most pertinent features for each specific fault type, thereby improving prediction accuracy while working with comprehensive datasets without suffering from overfitting.
Data Source
AI summary
Customized feature vectors are used to train a machine learning algorithm to automatically identify a network component where a network fault has occurred. A database comprising network components and associated network faults is analyzed to select a set of network components associated with the largest quantity of network faults. Customized features associated with the network faults are identified and selected for use in a feature vector as input to a machine learning algorithm. The features are selected based upon analysis of consistency checks, component configuration limits, and network wide configurations.