Time-Series Data Analysis Device for Explanatory Waveform Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for analyzing time-series data using machine learning often lack explanatory properties, making it difficult to appropriately identify and visualize the portions of data that serve as the basis for analysis.
Innovation Solution
An information processing device that designates the number of sections in time-series data to be analyzed, using an estimation model to update these sections based on both original and masked time-series data, thereby optimizing the objective function to minimize prediction error and section length.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black box analysis technology is used for high-performance waveform analysis, then analysis performance is improved, but explanatory property is degraded
Solution Approach 1:
The patent segments the time-series data into multiple sections and identifies important sections that contribute to the analysis result. By dividing the continuous data into discrete segments and ranking their importance, the system provides explanatory information about which parts of the data are most relevant, thus adding explanatory property to the black box analysis without sacrificing performance.
2Loss of information
If saliency map technology is used to highlight portions contributing to prediction, then explanatory property is improved, but the number of extracted portions becomes uncontrolled and sections become fragmented into many small pieces
Solution Approach 1:
The patent introduces a parameter N that specifies the number of important sections to extract. By controlling this parameter, the system can limit the number of extracted sections to a manageable amount, preventing fragmentation into excessive small pieces while maintaining explanatory property. This parameter control allows users to balance between detail and manageability.
3Loss of information
If mask processing is applied to obtain portions serving as basis of analysis, then explanatory property is improved, but prediction error increases due to loss of data information
Solution Approach 1:
The patent applies mask processing selectively to only the sections that are not identified as important, rather than masking the entire time-series data. By leaving the important sections unmasked and intact, the system maintains the data information necessary for accurate prediction while still providing explanatory information about which sections are most relevant. This partial application of masking reduces prediction error compared to full masking.
Data Source
AI summary
An information processing device according to one embodiment includes one or more hardware processors. The hardware processors executes update processing on a designated number of first sections on the basis of a first estimation result and a second estimation result. The first estimation result is obtained by inputting first time-series data to an estimation model. The second estimation result is obtained by inputting second time-series data to the estimation model. The second time-series data is obtained by applying mask processing to partial time-series data of a second section in the first time-series data. The second section is other than the designated number of the first sections.