Network Performance Change Point Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network upgrades often result in unintended changes in network performance, which are difficult to detect due to the complexity of large-scale networks and the overlapping impacts of multiple triggers, leading to challenges in identifying significant behavior changes and correlating them with specific triggers.
Innovation Solution
A method and system for identifying change points in network performance data by ranking and calculating cumulative sums, determining change scores, and correlating these with triggers, allowing for the detection of statistically significant changes and filtering out insignificant ones, thereby identifying trigger/change point pairs and attributing behavior changes to specific network upgrades.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network upgrades are implemented to improve performance and support new functions, then network capabilities and reliability are improved, but unintended changes in network performance become difficult to detect due to system complexity and overlapping trigger impacts
Solution Approach 1:
The patent segments the complex network performance data into individual time series from multiple measurement points. Each time series is analyzed separately to identify change points, and then the results are aggregated. This segmentation approach breaks down the complexity of detecting changes in the entire network system into manageable individual analyses, making it possible to detect unintended performance changes even in large-scale networks with multiple overlapping upgrades.
2Adaptability or versatility
If multiple network upgrades are implemented simultaneously to improve network capabilities, then network functionality and performance are enhanced, but it becomes challenging to correlate specific behavior changes with specific triggers
Solution Approach 1:
The patent implements a feedback mechanism where detected change points in network performance are correlated with known upgrade triggers. The system aggregates change points from multiple measurement points and uses statistical methods to determine which triggers are most likely responsible for observed changes. This feedback loop allows the system to attribute behavior changes to specific upgrades even when multiple upgrades are implemented simultaneously, preventing loss of trigger-correlation information.
3Measurement precision
If comprehensive network monitoring is performed to detect all performance changes, then detection coverage is improved, but false positives increase due to natural performance variations and insignificant changes
Solution Approach 1:
The patent changes the parameter of analysis by using statistical methods to evaluate the significance of detected change points. Instead of simply detecting all changes, the system calculates whether observed changes exceed natural performance variations and meet predefined significance thresholds. This parameter change approach allows comprehensive monitoring coverage while filtering out false positives caused by insignificant natural variations in network performance.
Data Source
AI summary
A system and method are provided for identifying a change point in a set of data. The system performs the method by receiving a set of data. The data indicates a plurality of performance measurements from a measurement point in a network. Each of the plurality of measurements represents a single type of performance measurement made at the measurement point at each of a corresponding plurality of points in time. The method also includes dividing the set of data into a plurality of data points in a chronological order. Each data point has a value corresponding to the performance measurements. The method also includes ranking the data points in an ascending order, calculating a cumulative sum for each of the data points, calculating a change score for the set of data points. A change point is identified in the data set if the change score exceeds a predetermined confidence level.