Time-Series Window Width Determination via Frequency Domain Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for setting the window width in time-series data analysis require significant man-hours, especially when prior knowledge of the data and domain knowledge are lacking, leading to inefficiencies in parameter adjustment.
Innovation Solution
An information processing apparatus and method that transforms time-series data into frequency domain data, compares frequency domain data across classes, and determines an optimal window width based on frequency differences, allowing for automatic adjustment of the window width during machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large window width is used to increase information amount for accurate prediction, then prediction accuracy is improved, but prediction speed decreases
Solution Approach 1:
The patent transforms the time-series data parameter from time domain to frequency domain through Fourier transform, changing the representation parameters while preserving the information content. This allows the system to analyze the same data with different frequency components without increasing the time window width, thus maintaining prediction speed while improving accuracy through better feature discrimination in the frequency domain
2Productivity
If prior knowledge and domain knowledge are used to set window width based on frequency peak, then parameter adjustment efficiency is improved, but applicability to unknown data domains decreases
Solution Approach 1:
The system performs self-adjustment of the window width parameter by automatically analyzing the frequency domain characteristics of the input data. The determining unit calculates the window width based on the frequency spectrum without requiring external prior knowledge or manual intervention, enabling the system to adapt to any data domain automatically while maintaining high parameter adjustment efficiency
Solution Approach 2:
The patent performs preliminary Fourier transform and frequency spectrum analysis before setting the window width parameter. By pre-calculating the frequency characteristics and determining the appropriate window width in advance based on the data's own spectral properties, the system eliminates the need for subsequent manual parameter tuning and ensures optimal performance for the specific data domain
3Measurement precision
If brute force comparison of huge number of window width candidates is performed, then optimal parameter selection is improved, but man-hours for parameter adjustment increase significantly
Solution Approach 1:
The patent replaces the mechanical brute-force search method with a mathematical transformation approach. Instead of manually or computationally testing numerous window width candidates through trial and error, the system uses Fourier transform to convert the optimization problem into a frequency domain analysis, where the optimal window width can be determined directly from spectral characteristics, dramatically reducing the time and computational resources required
Data Source
AI summary
An information processing apparatus according to the present invention includes: a transforming unit configured to transform time-series data that is learning data into frequency domain data; a comparing unit configured to perform comparison between the frequency domain data corresponding to the learning data belonging to different classes, respectively; and a determining unit configured to determine a time width of the time-series data that is the learning data based on a result of the comparison between the frequency domain data. The time width is set at a time of performing machine learning of the time-series data.