Online Data Decomposition for Real-Time Seasonal Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data decomposition methods, such as Seasonal and Trend decomposition using LOESS (STL), are resource-intensive and perform batch processing, leading to inaccurate and time-consuming analysis, as they require extensive data points for decomposition and frequent recalculations, which can result in data inaccuracies and high resource utilization.
Innovation Solution
The implementation of online data decomposition that iteratively processes incoming data points using a limited set of previous data points and components to determine trend, seasonal, and residual components in real-time, allowing for accurate and efficient decomposition without awaiting subsequent data points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If batch processing with STL decomposition is used, then comprehensive data analysis is achieved, but processing time and resource utilization increase significantly
Solution Approach 1:
The patent segments the batch processing task into individual online processing steps, where each incoming data point is decomposed independently using a limited window of previous data points. This segmentation enables real-time decomposition without waiting for complete data sets, resolving the contradiction between comprehensive analysis and processing time.
Solution Approach 2:
The patent performs preliminary decomposition actions on each incoming data point as it arrives, rather than waiting for batch completion. By maintaining a limited history window and performing decomposition iteratively, the system achieves timely analysis while reducing overall processing time and resource utilization.
2Measurement precision
If frequent recalculations are performed in batch processing, then data accuracy is maintained, but resource utilization increases
Solution Approach 1:
The patent extracts only the necessary historical data points (limited window) required for decomposition, rather than recalculating with all available data. This extraction approach maintains decomposition accuracy by using sufficient recent data while eliminating the need to process entire historical data sets, thus reducing resource utilization.
Solution Approach 2:
The patent applies partial action by using a limited subset of historical data points rather than all available data. This partial approach is sufficient for accurate online decomposition while significantly reducing computational resources required compared to full batch recalculations.
3Measurement precision
If extensive data points are used for decomposition, then decomposition accuracy improves, but storage requirements increase
Solution Approach 1:
The patent segments the data storage requirement into a limited historical window rather than storing all incoming data points. By maintaining only the necessary number of previous data points and components for online decomposition, the system achieves accurate decomposition while minimizing storage requirements.
Data Source
AI summary
Embodiments described herein are directed to facilitating performing online data decomposition to identify multiple seasonal components. In accordance with aspects of the present disclosure, a first iterative process is performed to determine a first seasonal component associated with an incoming data point based on a set of previous data points of a time series data set and corresponding data components. In addition, a second iterative process is performed to determine a second seasonal component associated with the incoming data point based on previous data points of the time series data set and corresponding data components. The first seasonal component and the second seasonal component can then be provided for analysis of the incoming data point (e.g., for presentation, for use in determining trend and residual components, etc.).