Network Node Performance Monitoring via Machine Learning Thresholds

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional performance monitoring systems for network nodes are reactive, failing to detect impending abnormal behavior and lacking granularity in understanding correlations between performance metrics across interconnected nodes.

Innovation Solution

A computer-implemented method and apparatus that use machine-learning algorithms to derive node performance assessment thresholds, enabling proactive monitoring of network nodes by analyzing correlations between performance metrics across interconnected nodes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional performance monitoring using fixed thresholds is used, then the monitoring system is simple to implement, but it fails to detect impending abnormal behavior and lacks granularity in understanding correlations between performance metrics

Engineering Contradiction:

Improvedetection accuracy of abnormal behaviorVSAvoidmonitoring system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transforms fixed performance thresholds into dynamic, node-specific thresholds derived through machine learning analysis of historical performance data. The system learns optimal threshold values for each node based on its unique performance patterns and correlations with interconnected nodes, enabling precise detection of abnormal behavior while adapting to changing network conditions

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system performs preliminary machine learning analysis during a training phase to establish baseline performance characteristics and correlations between nodes before actual monitoring begins. This preliminary action enables the system to detect impending abnormalities by comparing real-time metrics against pre-established patterns, allowing proactive intervention before failures occur

Inventive Principle:

Principle #10Preliminary action

2Reliability

If reactive performance monitoring is used, then the monitoring approach is straightforward, but abnormal behavior is detected only after it has already occurred, reducing the ability to prevent performance degradation

Engineering Contradiction:

Improvenode performance reliabilityVSAvoiddetection time for abnormal behavior

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent implements a feedback mechanism where the machine learning model continuously learns from performance data and updates its understanding of normal versus abnormal patterns. The system provides feedback loops that adjust thresholds and predictions based on observed performance trends, enabling the network to adapt to changing conditions and improve detection accuracy over time

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent replaces conventional mechanical threshold-based monitoring with an intelligent machine learning system that analyzes performance metrics and predicts abnormalities. This substitution enables proactive detection by identifying patterns and correlations that fixed thresholds cannot detect, transforming reactive monitoring into predictive monitoring

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If machine learning algorithms are used to derive node performance assessment thresholds, then early detection of abnormal behavior is enabled, but the processing load and computational resources increase

Engineering Contradiction:

Improveperformance assessment accuracyVSAvoidcomputational energy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the machine learning analysis by creating node-specific performance models that focus on individual node characteristics and their specific correlations with interconnected nodes. This segmentation allows the system to process data more efficiently by avoiding redundant computations across the entire network, reducing overall computational energy consumption while maintaining high detection accuracy

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12273748B2Monitoring the performance of a plurality of network nodes

Publication Date: 2025.04.08 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)

AI summary

A computer-implemented method and apparatus obtain a plurality of data sets. A data set comprises a respective value of a performance metric for each of a plurality of network nodes interconnected in a multi-hop arrangement. Each of the plurality of data sets is classified as normal or abnormal by comparing the respective values of the performance metric of each of the plurality of network nodes to a corresponding normality threshold, thus providing a plurality of classified data sets. The plurality of classified data sets is processed using a machine-learning algorithm in order to derive, for at least one network node of the plurality of network nodes, a node performance assessment threshold indicative of a value of the performance metric of the at least one node at which the plurality of network nodes has a predetermined likelihood of being classified as normal.