Classification Model Tree for Time-Series State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current analysis systems for estimating the state of an observation target can only perform estimation for the entire time-series data, failing to accurately determine the state of specific parts of the data, which limits their applicability in scenarios like rehabilitation or technical training where partial correct actions are performed.
Innovation Solution
An information processing apparatus that generates a classification model for each time section of the time-series data, using boundary times to determine whether the action is correct or incorrect, allowing for precise determination of action states at every time point by using a model tree and classification models with varying identification performance thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single classification model is generated for the entire time-series data, then the device complexity is reduced, but the measurement precision for specific time points deteriorates
Solution Approach 1:
The patent divides the time-series data into multiple time sections using boundary times, and generates a separate classification model for each time section. This segmentation allows the system to achieve high measurement precision for specific time points while managing model complexity through structured organization of multiple simpler models rather than one complex comprehensive model.
2Measurement precision
If classification models are generated for all time sections, then the measurement precision improves, but the calculation load increases
Solution Approach 1:
The patent applies local quality by generating classification models only for time sections where the identification performance exceeds a predetermined threshold. This selective approach maintains high measurement precision for critical time sections while avoiding unnecessary calculations in sections with low identification performance, thus improving overall calculation efficiency.
3Adaptability or versatility
If the identification performance threshold is lowered, then the coverage of time sections increases, but the accuracy of determination deteriorates
Solution Approach 1:
The patent uses parameter changes by dynamically adjusting the identification performance threshold based on the specific time section being evaluated. Different thresholds are applied to different time sections, allowing the system to adapt to varying requirements and maintain both high coverage and accurate determination across diverse temporal contexts.
Data Source
AI summary
According to one embodiment, an information processing apparatus includes an estimator configured to estimate a state of an observation target on a first time, based on data on the first time included in time-series data obtained from the observation target. Also, an information processing method for estimating a state of an observation target on a first time, based on data on the first time included in time-series data obtained from the observation target is provided.