Graph-Based Event Forecasting Using Matrix Profile Motif Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multivariate time series prediction techniques fail to effectively analyze multiple events within a time interval and provide clear explanations for anomaly events, as they do not adequately account for dependencies between time-dependent variables.
Innovation Solution
A graph-based machine-learning model that uses a matrix profile-based motif detection technique to identify events in a multivariate time series, generating a bipartite graph representation to predict event occurrences and calculate anomaly scores, thereby improving event forecasting and anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional multivariate time series prediction techniques are used, then the system is simple to implement, but it cannot effectively analyze multiple events within a time interval and cannot provide explanations for anomaly events
Solution Approach 1:
The patent segments the time series data into distinct events using motif detection, and further segments the analysis into graph-based relationships between events. This allows multiple events within a time interval to be analyzed independently and systematically, improving measurement precision while managing complexity through structured decomposition
Solution Approach 2:
The patent introduces graph representations as an intermediary layer between raw time series data and prediction outcomes. The graph structure mediates the analysis by capturing temporal and causal relationships between events, enabling both multiple event analysis and explanatory capabilities without directly complicating the core prediction mechanism
2Loss of information
If graph-based machine learning models with motif detection are used, then multiple events can be analyzed and explanations provided, but computational complexity increases
Solution Approach 1:
The patent performs preliminary motif detection and graph construction before the actual prediction process. By pre-processing the time series data into structured graph representations that capture event dependencies, the system preserves information about temporal relationships while reducing the computational burden during prediction, as the complex pattern recognition is done upfront
Solution Approach 2:
The patent creates simplified graph-based copies or representations of the complex time series data. Instead of directly processing raw multivariate time series during prediction, the system uses pre-extracted graph structures that replicate the essential temporal and causal relationships, reducing computational complexity while preserving dependency information
Data Source
AI summary
Provided is a system for event forecasting using a graph-based machine-learning model that includes at least one processor programmed or configured to receive a dataset of data instances, where each data instance comprises a time series of data points, detect a plurality of motifs representing a plurality of events in the dataset of data instances using a matrix profile-based motif detection technique, generate a bipartite graph representation of the plurality of motifs in a time sequence, and generate a machine-learning model based on the bipartite graph representation of the plurality of motifs in the time sequence, where the machine-learning model is configured to provide an output and the output includes a prediction of whether an event will occur during a specified time interval. Methods and computer program products are also provided.