Time-Series Anomaly Detection Using Dynamic Time Windows

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The vast amount of time-series data collected in IT networks makes it difficult to manage and synthesize for identifying anomalies, leading to challenges in detecting issues and potential false anomaly reporting.

Innovation Solution

An application server analyzes time-series data using a relational time-series query request, classifies the data based on one or more classifications, generates a statistical model, and calculates an anomalous score by adjusting the time window dynamically to identify and report anomalies accurately.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a fixed time window is used for anomaly detection, then the detection process is simple, but the accuracy decreases when data probability varies over time

Engineering Contradiction:
Improveanomaly detection accuracyVSAvoidtime window adjustment mechanism
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements a dynamic time window adjustment mechanism where the time window size is automatically modified based on the probability characteristics of the time-series data. The system calculates data probability and adjusts the time window accordingly - expanding it when probability is low and contracting it when probability is high - thereby adapting the detection sensitivity to varying data conditions and improving anomaly detection accuracy.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If more time-series data is collected, then the statistical model becomes more accurate, but the difficulty of synthesizing and managing the data increases

Engineering Contradiction:
Improvestatistical model accuracyVSAvoiddata synthesis difficulty
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent extracts and utilizes only the necessary historical data points required for constructing the statistical model, rather than processing all available time-series data. By selectively extracting relevant data based on the dynamic time window and probability thresholds, the system achieves accurate statistical modeling while avoiding the complexity of synthesizing and managing vast amounts of raw data.

Inventive Principle:
Principle #2Taking out (Extraction)

3Reliability

If the time window is expanded to capture more data patterns, then the anomaly detection coverage improves, but the response time to detect anomalies increases

Engineering Contradiction:
Improveanomaly detection coverageVSAvoidanomaly detection delay
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system dynamically adjusts the time window size based on real-time probability calculations. When data probability is low (indicating potential anomaly), the time window is expanded to improve detection coverage and reliability. When data probability is high (normal conditions), the time window is contracted to enable faster response time. This dynamic adjustment resolves the contradiction between coverage and response time by adapting the window size to current data conditions.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3451232B1Systems and methods for anomaly detection
Publication Date: 2025.07.09 SERVICENOW INC
  • EP3451232B1 patent drawingFigure 1~2
  • EP3451232B1 patent drawingFigure 3
  • EP3451232B1 patent drawingFigure 4

AI summary

An anomaly detection module may include a time-series analyzer that classifies current time-series data into at least one of a plurality of classifications based upon historical data and may construct a statistical model representative of the current time-series data based upon the at least one of the plurality of classifications. An anomaly detector monitors a stream of the current time-series data and identifies statistical outliers of the stream of the current time-series data, based upon the statistical model and may determine an anomalous score for the statistical outliers by tracking a history of the statistical outliers; wherein the anomalous score comprises a representation of a magnitude of deviation between the current time-series data and the statistical model over multiple measurements of the current time-series data, over a particular time interval, or both.