Time-Series Feature Extraction Using Dynamic Segment Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting features from time-series data face challenges in determining whether to extract global or local features, especially when data is noisy, and clustering similar data results in ineffective local feature extraction.
Innovation Solution
A time-series feature extraction apparatus that classifies data into segments using regression models, clusters them, and calculates representative elements to capture both global and local features, with noise removal and segment position adjustment mechanisms to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time series data is divided into segments to extract local features, then local feature extraction is enabled, but similar data in clusters results in similar extracted partial data making local feature extraction ineffective
Solution Approach 1:
The patent implements dynamic adjustment of segment boundaries based on the representative elements of each cluster. Instead of using fixed segmentation, the system adapts the segmentation to the actual data distribution by calculating representative elements (such as mean or median values) for each cluster and using these to dynamically determine optimal segment boundaries. This ensures that segments are positioned to maximize feature distinguishability even when clusters contain similar data.
Solution Approach 2:
The patent changes the parameter of segment boundary positions based on cluster characteristics. By calculating representative elements for each cluster and adjusting segment boundaries accordingly, the system transforms static segmentation into adaptive segmentation that responds to data distribution. This parameter change enables the extraction of distinguishable local features even from similar clustered data.
2Measurement precision
If global feature extraction is used for time series data, then overall data characteristics are captured, but local distinctive features may be lost
Solution Approach 1:
The patent divides time series data into multiple segments before extracting features from each segment. This segmentation allows the system to capture both global characteristics (through the overall structure of segments) and local distinctive features (through individual segment analysis). The segmented approach enables parallel processing of global and local features without information loss.
Solution Approach 2:
The patent adds a segment dimension to the feature extraction process. Instead of extracting features only at the global level or only at the local level, it creates an intermediate dimensional structure where data is organized into segments. This dimensional transformation enables simultaneous capture of global patterns (across segments) and local characteristics (within segments).
3Ease of operation
If clustering is performed to group similar time series data, then data organization is improved, but extracted partial data become similar reducing feature effectiveness
Solution Approach 1:
The patent implements dynamic segment boundary adjustment based on cluster representative elements. Instead of using static segmentation that would produce similar results for similar clusters, the system dynamically calculates optimal boundaries using representative elements (mean/median) of each cluster. This dynamic approach ensures that even similar clusters produce distinguishable local features by adapting segmentation to each cluster's specific characteristics.
4Reliability
If noise is present in time series data, then data quality deteriorates, but both global and local features may not be correctly extracted
Solution Approach 1:
The patent extracts representative elements (such as mean or median values) from each cluster as a separate entity. This extraction process effectively separates the essential feature information from noise in the original data. By working with these extracted representative elements rather than raw noisy data, the system achieves more reliable and accurate feature extraction even when noise is present in the source data.
Data Source
AI summary
A time-series feature extraction apparatus has a coefficient outputter to output a coefficient to be used in calculation for classifying time series data into a plurality of segments, a segment position outputter to perform calculation for classifying the time series data into the plurality of segments based on the coefficient to output information on boundary positions of the plurality of segments, a cluster classifier to classify the plurality of segments into a certain number of plurality of clusters equal to or smaller than a certain number of the plurality of segments, a representative element outputter to output a representative element which represents a local feature of each of the plurality of clusters and is set for each of the plurality of segments, a feature degree calculator to calculate a feature degree of the representative element, and a representative element updater to update the representative element based on the feature degree.