Semantic Trajectory Modeling for Context-Aware Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for raw trajectory mining lack context and fail to account for underlying causes influencing movement, leading to unreliable and high false positive rates in anomaly detection from multi-source time series data.
Innovation Solution
A machine learning approach that transforms raw time series data into a hierarchical abstraction, incorporating semantic concepts and contextual information to generate enriched trajectories, which are then processed to identify anomalies through interdependencies between sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If raw trajectory mining is used to detect anomalies in time series data, then the detection process is simple and fast, but the accuracy is low and false positives are high due to lack of context
Solution Approach 1:
The patent segments the time series data into multiple sources and dimensions, processing each separately before integrating them through cross-correlation analysis. This segmentation allows complex multi-source data to be handled systematically, improving detection accuracy while managing computational complexity through structured processing stages.
Solution Approach 2:
The patent introduces a contextual dimension by incorporating external knowledge bases and semantic annotations alongside the temporal and spatial dimensions of trajectory data. This additional dimensional layer enables the system to understand context-aware patterns, reducing false positives while maintaining processing efficiency through hierarchical organization of data at multiple abstraction levels.
2Reliability
If multiple sources of time series data are combined to improve detection reliability, then the opportunity to detect anomalies increases, but the difficulty of obtaining correlations among dimensions increases
Solution Approach 1:
The patent introduces cross-correlation analysis as an intermediary mechanism that systematically computes relationships between different data sources and dimensions. This intermediary approach transforms the complex task of finding correlations among multiple sources into a structured process, enabling reliable multi-source anomaly detection while managing computational complexity through efficient correlation algorithms.
3Measurement precision
If contextual information is added to trajectory data to reduce false positives, then the interpretation accuracy improves, but the data processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing and organizing contextual information from external knowledge bases before the main anomaly detection process. Contextual data is structured and indexed in advance, allowing rapid retrieval and integration during analysis. This preliminary organization significantly reduces processing time while maintaining high precision in contextual pattern recognition.
Solution Approach 2:
The patent applies partial action by selectively incorporating contextual information only where needed and relevant to the specific anomaly detection task. Rather than processing all available contextual data uniformly, the system identifies and processes only the necessary contextual dimensions, improving measurement precision while minimizing processing time through targeted contextual enrichment.
Data Source
AI summary
A method for time series anomaly detection includes: generating, based on multi-source time series data and contextual data, geometric trajectories representing movement of an entity; processing the geometric trajectories and the contextual data to extract a plurality of features, wherein the plurality of features include temporal features, spatial features and contextual features; generating a data structure representing semantic trajectories, wherein each of the semantic trajectories includes the temporal features, the spatial features and the contextual features; generating, using the data structure, based on the contextual features, contextual encodings corresponding to the semantic trajectories and generating, based on the temporal features, temporal encodings corresponding to the semantic trajectories; processing, with a machine learning model, the contextual encodings and the temporal encodings to generate source embeddings representing interdependencies between the semantic trajectories; and outputting, based on the source embeddings, an indication of whether one of the semantic trajectories is anomalous.