Time Series Relevance Discovery via State Transition Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for monitoring multiple time series data lack effective means to identify correlations between series, leading to resource-intensive monitoring and inefficient analysis, as traditional correlation measurements like Pearson's coefficients fail to capture meaningful relationships, especially in the presence of outliers or non-linear changes.
Innovation Solution
A method using state transition points to measure time series relevance through mutual information maximization, involving the conversion of time series values to ranks, identification of valid bifurcation points, and application of a hill climbing algorithm to estimate true relevance, while leveraging triangular inequalities for efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional correlation measurements like Pearson's coefficients are used to monitor multiple time series, then the monitoring process is resource-intensive and computationally expensive, but the ability to capture meaningful relationships and identify correlations is insufficient
Solution Approach 1:
The patent segments time series data at state transition points to create discrete segments that can be analyzed independently. This segmentation allows the algorithm to focus computational resources on identifying relevant correlations only where state changes occur, rather than continuously analyzing all data points, thus improving both correlation detection accuracy and monitoring efficiency
Solution Approach 2:
The patent transforms the correlation measurement approach by changing from traditional continuous correlation coefficients to a discrete state-based relevance measurement. By converting time series values to ranks and identifying state transition points, the method changes the parameter space from continuous correlation values to discrete state sequences, enabling more efficient computation while capturing non-linear relationships
2Reliability
If all time series are monitored continuously to ensure complete coverage, then resource consumption increases, but the ability to derive relationships among time series remains insufficient
Solution Approach 1:
The patent extracts only the essential information from time series data by identifying state transition points and creating discrete state segments. Instead of continuously monitoring all time series values, the method extracts key state change events and uses these to infer relationships, thereby maintaining reliable monitoring coverage while significantly reducing resource consumption
Solution Approach 2:
The patent performs preliminary segmentation of time series data at state transition points before conducting correlation analysis. By pre-processing the data to identify and segment state changes, the method prepares the data in a form that enables efficient relevance calculation, avoiding the need for continuous resource-intensive monitoring while ensuring complete coverage of meaningful events
3Ease of operation
If traditional correlation methods are applied to detect relationships among time series, then the analysis is simplified, but the ability to capture non-linear correlations and handle outliers is inadequate
Solution Approach 1:
The patent inverts the traditional approach by not directly measuring correlation between time series, but instead measuring the relevance of their discrete state sequences. By converting continuous time series to discrete state segments and then calculating relevance between these segments, the method captures non-linear relationships and handles outliers more effectively while maintaining analytical simplicity through the use of rank-based transformation
Data Source
AI summary
A method for measuring time series relevance using state transition points, including inputting time series data and relevance threshold data. Then convert all time series values to ranks within [0,1] interval. Calculate the valid range of the transition point in [0,1]. Afterwards, a verification occurs that a time series Z exists for each pair of time series Z and Y, such that the relevances between X and Z, and between Y and Z are known. Then deduce the relevance of X and Y. The relevance of X and Y must be at least one of, (i) higher, and (ii) lower than, the given threshold. Provided Z is found terminate all remaining calculations for X and Y. Otherwise, segment the time series if no Z time series exists, use the segmented time series to estimate the relevance. Apply a hill climbing algorithm in the valid range to find the true relevance.