Multivariate Time-Series Prediction with Motif Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data analysis methods fail to effectively identify and visualize frequently occurring patterns, or motifs, within large datasets from real-world settings, such as chiller data in data centers, which are crucial for system performance verification and future event prediction.
Innovation Solution
The combination of multivariate time-series prediction and motif pattern discovery techniques, using pattern-preserving prediction and time-distance optimization, to generate predicted data points and locate frequently occurring patterns, which are then displayed with associated performance characteristics for user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multivariate time-series prediction is performed on large datasets, then predicted data points can be generated for future events, but the complexity of data processing and pattern identification increases
Solution Approach 1:
The patent segments the data processing into distinct modules: multivariate time-series prediction module for generating predicted data points, motif discovery module for identifying patterns, and visualization module for displaying results. This segmentation reduces overall complexity by making each module independently manageable and optimizable.
Solution Approach 2:
The patent introduces motifs as intermediary representations that bridge the raw data and the analysis results. Motifs serve as condensed pattern representations that simplify the relationship between complex multivariate time-series data and the final visualization, making the system more manageable.
2Reliability
If motif discovery is performed on large datasets, then frequently occurring patterns can be identified, but the time required for analysis increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing the data to identify and extract significant patterns before the main analysis. The motif discovery algorithm pre-organizes data structures and pre-computes statistical properties, enabling faster pattern identification during the actual analysis phase.
Solution Approach 2:
The patent dynamically adjusts analysis parameters such as motif length, minimum occurrence threshold, and search depth based on dataset characteristics and user requirements. This allows the system to optimize analysis time by adapting parameters to the specific problem instance rather than using fixed parameters.
3Ease of operation
If data visualization is enhanced to show motifs and performance characteristics, then user understanding improves, but the interface complexity increases
Solution Approach 1:
The patent uses color coding to represent different motif types and performance characteristics in the visualization. Different colors indicate different patterns, their occurrences, and performance levels, allowing users to quickly interpret complex data through intuitive color semantics without increasing interface complexity.
Solution Approach 2:
The patent creates simplified visual copies or representations of the complex data patterns in the form of motif icons and performance indicators. These visual copies distill the essence of complex multivariate time-series data into easily interpretable graphical elements that maintain information integrity while improving usability.
Data Source
AI summary
Multivariate time-series prediction is combined with motif discovery. Multivariate time-series prediction is performed on measured (past) data points to generate or determine predicted data points. One or more motifs are discovered or determined within the combined measured data points and the predicted data points. The motifs within the measured data points are associated with the motifs within the predicted data points to permit inferences of the motifs within the measured data points to be applied to the motifs within the predicted data points. The measured data points, the predicted data points, and the motifs are displayed. User interaction with this display of the measured data points, the predicted data points, and the motifs is permitted.