Time-Series Sampling Period Optimization via Attractor Roundness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time-series data analysis methods using topological data analysis face challenges in accurately extracting features when the sampling period is excessively short or long, leading to oversampling, which obscures the difference between normal and abnormal data due to the generation of shattered or linear pseudo attractors.
Innovation Solution
A machine learning method that generates a sine wave from input data with a periodic property to determine an optimal sampling period based on the roundness of the attractor, thereby suppressing oversampling and improving feature extraction by using the determined sampling period to generate pseudo attractors and perform machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the sampling period is excessively short to capture phenomenon features thoroughly, then the data capture completeness is improved, but the pseudo attractors become shattered and linear, causing feature extraction accuracy to deteriorate
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the sampling period based on the detected periodicity of the time-series data. Instead of using a fixed sampling period, the system calculates the periodicity from the data itself and sets the sampling period to a multiple of this periodicity, thereby optimizing the balance between capturing phenomenon features and maintaining pseudo attractor shape quality.
Solution Approach 2:
The patent implements feedback by using the periodicity detection result from the time-series data to determine the sampling period. The system detects the periodicity of the input data and feeds this information back to set the sampling period accordingly, creating a closed-loop optimization that adapts to the actual data characteristics.
2Productivity
If the sampling period is excessively long, then the computational complexity is reduced, but the feature capture completeness deteriorates
Solution Approach 1:
The patent uses parameter changes by calculating the periodicity from the time-series data and setting the sampling period to a multiple of this periodicity. This dynamic parameter adjustment ensures that the sampling period is neither excessively short nor excessively long, but optimally configured for the specific data characteristics, thereby balancing processing efficiency with feature capture accuracy.
3Measurement precision
If manual search for appropriate sampling interval is performed by repetitive work, then the sampling optimization is achieved, but the time consumption and operational complexity increase
Solution Approach 1:
The patent applies self-service by enabling the system to automatically detect the periodicity of the time-series data and determine the optimal sampling period without requiring manual intervention. The system performs periodicity detection and sampling period determination autonomously based on the input data characteristics, eliminating the need for manual repetitive search while achieving optimized sampling intervals.
Solution Approach 2:
The patent replaces the mechanical manual search process with an automated computational approach. Instead of requiring manual repetitive work to find the appropriate sampling interval, the system uses algorithms to detect periodicity and calculate the optimal sampling period automatically, substituting human operational complexity with computational efficiency.
Data Source
AI summary
A machine learning method includes: generating, by a computer, a sine wave using a basic period of input data having a periodic property; determining a sampling period based on a degree of roundness of an attractor generated from the sine wave; sampling the input data at the determined sampling period to generate a pseudo attractor; and performing a machine learning by using the pseudo attractor.