Time-Series Compression Using Piecewise Regression Intervals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The proliferation of data from sources like smart meters and sensors leads to large volumes of data that become impractical to store in their entirety, necessitating efficient compression methods that maintain accuracy within a tolerable error margin.
Innovation Solution
A computing system that uses piecewise regression to approximate data intervals by determining the lowest-order polynomial functions that fit data segments within a maximum allowable error, incrementally adding data points and switching to different function types as needed, while storing key parameters and intervals for efficient storage and retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all data points from smart meters and sensors are stored in their entirety, then measurement precision and data completeness are maintained, but storage requirements become excessive and impractical
Solution Approach 1:
The patent segments the time series data into multiple intervals and fits different polynomial functions to each interval. This segmentation allows the system to store only the essential parameters (polynomial coefficients and interval boundaries) rather than all raw data points, significantly reducing storage requirements while maintaining measurement precision within each segment.
Solution Approach 2:
The patent transforms the stored representation from raw data points to polynomial function parameters (coefficients). By changing the storage format from storing actual measurements to storing mathematical model parameters that generate the measurements, the system achieves compact storage while preserving the ability to reconstruct data with controlled precision.
2Quantity of substance
If data is compressed into intervals with polynomial approximations, then storage requirements are reduced, but data accuracy and fidelity to original points may be compromised
Solution Approach 1:
The patent incorporates a feedback mechanism where the system evaluates the fit quality of polynomial approximations against the original data points. Based on this evaluation, it determines whether to use lower-order or higher-order polynomials to achieve the desired accuracy threshold, ensuring that compression does not compromise data fidelity beyond acceptable limits.
Solution Approach 2:
The patent dynamically adjusts the polynomial order and interval boundaries based on the characteristics of the data being compressed. Rather than using a fixed compression scheme, the system adapts the approximation complexity to match the local behavior of the data, using simpler models where appropriate and more complex models where higher fidelity is needed.
3Measurement precision
If higher-order polynomial functions are used to approximate data intervals, then data fidelity is improved, but computational complexity and processing requirements increase
Solution Approach 1:
The patent applies partial action by using the minimum necessary polynomial order for each data interval rather than uniformly applying high-order polynomials throughout. This approach achieves sufficient approximation accuracy for each segment without the excessive computational complexity that would result from using high-order polynomials everywhere, optimizing the balance between fidelity and processing requirements.
Data Source
AI summary
According to an example implementation, at least one non-transitory computer-readable storage medium may include instructions stored thereon. When executed by at least one processor, the instructions may be configured to cause a computing system to select at least one function type to approximate an interval of data points within a time series, the at least one function type being less than all of a plurality of functions available to approximate the interval, while the first function type will approximate the interval of data points within a maximum allowable error, add data points from the time series to the interval, and when the first function type will no longer approximate the interval with the added data points within the maximum allowable error, select a function of a different type or a function of a different term from the plurality of functions to approximate the interval with the added data points.