Telecom Node KPI Forecasting With Online Polynomial Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection models in telecommunications networks fail to adequately address complexities such as memory footprint, processing speed, and data-drift, necessitating complex model lifecycle management through periodic retraining and re-deployment.
Innovation Solution
An online learning algorithm that continuously updates polynomial function coefficients to adapt to data-drift in telecommunications networks, eliminating the need for retraining and re-deployment by using incremental polynomial fitting and z-score anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are deployed to process and analyze network data in real-time, then anomaly detection accuracy and predictive capability are improved, but device complexity and computational resource requirements increase
Solution Approach 1:
The system divides the network monitoring function into distributed network nodes, each running lightweight machine learning models locally. This segmentation allows anomaly detection to be performed at the edge rather than requiring a centralized complex system, improving detection accuracy while distributing computational complexity across multiple simple nodes.
Solution Approach 2:
The patent introduces an intermediary layer that collects data from multiple sources, processes it through machine learning models, and generates standardized anomaly reports. This intermediary approach simplifies the overall system architecture by creating a dedicated processing layer that handles the complexity of machine learning operations.
2Measurement precision
If comprehensive network data collection and analysis are implemented, then forecasting accuracy for key performance indicators is improved, but loss of time for data processing and system overhead increases
Solution Approach 1:
The system performs preliminary data collection and preprocessing at network nodes before anomalies occur. Historical data is stored and pre-processed locally, allowing the machine learning models to quickly analyze current conditions without requiring extensive real-time data gathering, thus reducing processing time while maintaining forecasting accuracy.
Solution Approach 2:
The patent replaces traditional mechanical data collection methods with automated machine learning-based data gathering and analysis. This substitution enables parallel processing of multiple data streams and automated pattern recognition, significantly reducing the time required to analyze comprehensive network data compared to manual or rule-based systems.
3Reliability
If real-time anomaly detection and forecasting are implemented across the network, then network reliability and proactive maintenance capability are improved, but energy consumption and computational resources increase
Solution Approach 1:
The system segments the network into autonomous nodes that each perform local anomaly detection using lightweight machine learning models. This segmentation distributes computational energy consumption across many nodes rather than requiring high energy consumption at centralized points, enabling real-time monitoring while managing overall energy usage through parallel distributed processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosure relates to a method, for reporting an anomaly in a telecommunications node. The method comprises obtaining a measurement of a key performance indicator (KPI) of the telecommunication node. The method comprises, upon receiving the measurement of the KPI, updating coefficients of a polynomial function. The method comprises, based on the updated coefficients of the polynomial function, computing an expected measurement of the KPI coefficient. The method comprises computing a confidence band for the expected measurement of the KPI. The method comprises reporting the anomaly when the measurement of the KPI is outside of the confidence band. The disclosure also relates to a method for forecasting a plurality of expected measurements for a key performance indicator (KPI) in a telecommunications node, as well as to apparatuses for executing these methods.