Time-Series Data Label Allocation via Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for allocating labels to time-series data struggle to efficiently analyze details, particularly when features change over time, leading to difficulties in identifying patterns and accurately allocating labels to time-series data from sources like acceleration sensors, which can result in inefficient and inaccurate label allocation.

Innovation Solution

The proposed method involves dividing time-series data into segments, allocating labels based on segment features, and using a classification model to assign segment labels, allowing for detailed analysis and efficient allocation of labels to each pattern within the data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If time-series data is allocated with a single label based on overall features, then the allocation process is simple, but the ability to analyze detailed patterns and features changing over time is lost

Engineering Contradiction:

Improvelabel allocation accuracyVSAvoiddata processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides time-series data into multiple segments and allocates different labels to different segments based on their local features. This segmentation approach allows the system to capture detailed patterns within each segment while maintaining overall structure, thereby improving label allocation accuracy without requiring excessive complexity in the processing system.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different labeling strategies to different segments of time-series data based on their specific characteristics. Each segment can have its own label reflecting local features, while the overall data structure is preserved. This local quality approach enables accurate representation of features that change over time without complicating the entire processing system.

Inventive Principle:

Principle #3Local quality

2Loss of information

If time-series data is divided into segments and labeled individually, then detailed pattern analysis is improved, but the overall label allocation consistency may be compromised

Engineering Contradiction:

Improvepattern information retentionVSAvoidlabel allocation consistency

Core Design Contradiction:

Loss of informationVSStability of the object's composition

Solution Approach 1:

The patent combines multiple labeling results from different segments to produce a final comprehensive label. By merging the local labels from segmented data with the overall data characteristics, the system retains detailed pattern information while ensuring consistent label allocation across the entire time-series dataset.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent employs feedback mechanisms where the labeling of segments is adjusted based on the overall data context and previously allocated labels. This feedback loop ensures that detailed pattern information is preserved in each segment while maintaining consistency with the overall label allocation strategy for the complete time-series data.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11438277B2Allocation method, allocating device, and computer-readable recording medium

Publication Date: 2022.09.06 FUJITSU LTD

AI summary

An allocation method executed by a computer includes dividing each of a plurality of pieces of time-series data into a plurality of segments, allocating a label to each of the pieces of time-series data based on features of each segment in the pieces of time-series data, and allocating a predetermined segment in time-series data, included in the pieces of time-series data, with a label allocated to the time-series data to which the predetermined segment belongs.