Anomaly Detection in Media Event Sequences via Cosine Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods for media event sequences in communication systems fail to account for relationships between key performance indicators (KPIs), leading to undetected anomalies and false alarms, as they monitor KPIs individually rather than considering their interdependencies.
Innovation Solution
A method using a cosine similarity function to compare profile vectors of media event sequences over different time periods, detecting anomalies by computing the scalar product of these vectors and determining the presence of anomalies when the cosine similarity drops below a threshold, while also correlating with KPIs to identify the cause of anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If individual KPIs are monitored separately, then the monitoring system is simple and easy to implement, but anomalies involving relationships between KPIs remain undetected
Solution Approach 1:
The patent combines multiple individual KPI monitoring functions into a unified anomaly detection system that analyzes KPIs collectively. The system merges separate monitoring streams into a comprehensive analysis framework that evaluates relationships between KPIs, thereby detecting anomalies that individual monitoring would miss while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The patent creates a universal anomaly detection system that can handle multiple types of KPIs and anomaly patterns simultaneously. The system is designed to be multi-functional, capable of detecting various anomaly types (sudden changes, gradual drifts, relational anomalies) across different KPIs using a single unified approach, rather than requiring separate specialized monitors for each KPI relationship.
2Ease of manufacture
If individual KPI thresholds are monitored, then implementation is straightforward, but false alarms occur due to user behavior changes rather than system issues
Solution Approach 1:
The patent dynamically adjusts monitoring parameters and thresholds based on historical data and contextual information. Instead of using fixed thresholds, the system adapts parameters according to observed patterns, user behavior changes, and system conditions, thereby reducing false alarms while maintaining ease of implementation through automated parameter adjustment rather than complex manual configuration.
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously learns from detected patterns and adjusts its monitoring behavior. The system uses feedback from historical anomaly data and normal operation patterns to refine its detection criteria, reducing false alarms caused by user behavior changes while maintaining straightforward implementation through automated learning rather than complex rule sets.
3Reliability
If comprehensive KPI relationships are analyzed, then anomaly detection accuracy improves, but system complexity and computational requirements increase
Solution Approach 1:
The patent segments the comprehensive KPI analysis into manageable modules that can be processed independently and then integrated. The system divides complex relationship analysis into smaller analytical units, each handling specific KPI relationships or anomaly types, thereby maintaining high detection accuracy while reducing overall system complexity through modular, organized processing stages.
Solution Approach 2:
The patent transforms the complexity of multi-KPI relationship analysis by introducing a new dimensional approach to data representation and processing. The system reorganizes KPI relationships into a different analytical dimension or framework that simplifies the computation of complex interactions, enabling accurate anomaly detection without proportionally increasing system complexity through efficient mathematical transformations.
Data Source
AI summary
A method and detector for detecting anomalies among media event sequences are disclosed. One method includes obtaining a first profile of media event data over a first time period, where the first profile is a first distribution of different media event sequences. Each sequence is associated with a number of occurrences of the sequence and the first profile having a first profile vector. The method further includes obtaining a second profile of media event data over a second time period less than the first time period. The second profile is a second distribution of the different media sequences and the second profile having a second profile vector. The method also includes comparing the first profile vector and the second profile vector, and determining one of a presence and absence of at least one anomaly in the second profile vector of media event data based on the comparison.