Burst Detection Module for Activity Pattern Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional algorithms for detecting changes in systems or organizations based on monitored activities are often inefficient and inaccurate, failing to reliably identify non-momentary increased or decreased bursts of activity that indicate underlying changes or anomalies.
Innovation Solution
A burst detection module processes time series data to identify change points by modeling inter-arrival durations using Poisson processes and applying regression cumulative sum (CUSUM) detection techniques, distinguishing between momentary and non-momentary bursts, and providing alerts for actionable changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional algorithms are used for detecting changes in systems, then the detection process can be performed, but the detection efficiency and accuracy are insufficient
Solution Approach 1:
The patent transforms the detection problem by changing parameters from raw activity counts to inter-arrival time distributions. By modeling the temporal patterns of events using Poisson processes and analyzing deviations in inter-arrival times, the system achieves more accurate and efficient change detection compared to conventional algorithms that directly analyze activity counts.
Solution Approach 2:
The patent replaces conventional mechanical counting methods with statistical modeling techniques. Instead of simply counting events, the system uses probability theory and statistical distributions to model and detect changes, substituting straightforward mechanical counting with more sophisticated analytical methods.
2Reliability
If conventional algorithms detect bursts of activity, then detection can be performed, but they fail to reliably distinguish between momentary and non-momentary bursts
Solution Approach 1:
The patent introduces dynamic analysis by continuously monitoring the distribution of inter-arrival times and adapting the detection thresholds based on the evolving patterns. This dynamic approach allows the system to reliably distinguish between transient momentary bursts and sustained non-momentary bursts by analyzing how the statistical parameters change over time.
Solution Approach 2:
The patent performs preliminary modeling of the expected inter-arrival time distribution using Poisson processes before actual change detection. By establishing the baseline statistical characteristics in advance, the system can reliably identify deviations that indicate genuine bursts, filtering out false positives from momentary fluctuations.
Data Source
AI summary
A time series of data values representing occurrences of events at plural time points is received. Durations between successive events are computed, and a burst of activity based on the computed durations is detected. It is determined that a change has occurred in response to detecting the burst of activity.