Method and System for Rating Measured Values Taken from a System

Simple SummaryContent extracted from patent full text and abstract with AI.

This patent describes a method and system for evaluating measured values (data points) collected from a system, such as a communication network, which can be in either a normal or faulty state. The method uses advanced machine learning techniques that do not require pre-labeled (marked) data and eliminates the need for arbitrary threshold values. Instead, it processes unmarked data by probabilistically removing or weighting outlier values using a random-based method, which helps to build a more accurate model for detecting anomalies or faults in the system.

Use CasesContent extracted from patent full text and abstract with AI.

• Monitoring and diagnosing faults in computer networks or IT infrastructure
• Detecting anomalies in industrial automation systems or manufacturing plants
• Quality control in production lines through unattended sensor data analysis
• Network intrusion and security breach detection in communication systems
• Operational health monitoring of cloud-based or virtualized services
• Predictive maintenance by analyzing equipment sensor data for early failure detection
• Telecommunication service performance monitoring and troubleshooting
• Automated system health reporting for IoT or smart devices

BenefitsContent extracted from patent full text and abstract with AI.

• Does not require manually labeled or classified data for training the model, reducing setup effort and cost.
• Eliminates dependence on fixed thresholds, reducing the risk of misclassification and need for manual tuning.
• Can cope with complex, multi-modal, or unknown data distributions as it does not need detailed prior knowledge of the system’s normal state.
• Enhanced robustness in anomaly or outlier detection without discarding rare but normal data points.
• Applicable to a wide variety of data types and sources, such as time series, event logs, or resource usage metrics.
• Reduces the risk of false positives and negatives in fault or anomaly detection.
• Allows for iterative and probabilistic handling of ambiguous or borderline data, improving learning outcomes.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The invention relates to a method for rating measured values taken from a system (S) that may be in an error-free or erroneous state, wherein the system (S) has at least one communication network, a network component of a communication system and/or a service of a communication network, having the following steps, preferably in the following order: formation of a set (V) of unmarked measured values (v) from the system (S); formation of a modified learning set (V') containing measured values (v') for a learning system (L) by removal and/or weighting of measured values from the set (V) using a random-based method; formation of a model (M) for the rating of measured values from the system (S) by the learning system (L) from the modified learning set (V'); and rating of measured values from the system (S) by a rating system (B) using the model (M). Furthermore, the invention relates to a system for rating measured values taken from a system (S) that may be in an error-free or erroneous state, wherein the system (S) has at least one communication network, a network component of a communication system and/or a service of a communication network, having: a device for forming a set (V) of unmarked measured values (v) from the system (S); a device for forming a modified learning set (V) containing measured values (v') for a learning system (L) by removal and/or weighting of measured values from the set (V) using a random-based method; learning system (L) suited to forming a model (M) for the rating of measured values from the system (S) from the modified learning set (V'); and rating system (B) suited to rating measured values from the system (S) using the model (M).