State Classification Models With Influential Data Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems lack the ability to effectively identify and address the key measurement data influencing the state indication value of an object and recommend modifications to change its state from one condition to another based on complex data sets and operation data.
Innovation Solution
An apparatus and method that acquire and analyze multiple types of measurement data, using a model to classify the object's state and identify influential data, and generate simplified models to recommend modifications by calculating vectors and intersections, allowing for the display and execution of recommended changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complex model is used to classify object states from multiple measurement data types, then classification accuracy is improved, but model complexity and computational burden increase
Solution Approach 1:
The patent segments the complex model analysis into two distinct phases: (1) a classification model that outputs state indication values for categorizing object states, and (2) a separate explanation model that identifies influential measurement data. This segmentation allows each model to be optimized for its specific function, reducing overall system complexity while maintaining classification accuracy.
Solution Approach 2:
The patent introduces an intermediary explanation mechanism that acts as a bridge between the complex classification model and the user. This explanation model generates human-understandable outputs identifying which measurement data types most influenced the classification, making the complex model's decision-making process transparent without requiring simplification of the classification model itself.
2Measurement precision
If multiple types of measurement data are analyzed to identify influential data, then identification accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary action by pre-processing measurement data into standardized formats and pre-training the explanation model alongside the classification model. This allows the system to quickly process new data without performing complex analysis from scratch, reducing processing time for real-time applications while maintaining identification accuracy.
Solution Approach 2:
The patent changes parameters by adjusting the granularity and depth of analysis based on system requirements. The explanation model can operate at different levels of detail, and the system can selectively analyze only the most relevant measurement data types based on preliminary filtering, reducing computational burden while maintaining identification accuracy for critical parameters.
3Adaptability or versatility
If the model outputs detailed state indication values, then state classification capability is improved, but interpretability and ease of understanding decrease
Solution Approach 1:
The patent introduces an intermediary explanation layer that translates the model's detailed state indication values into human-understandable insights. The explanation model identifies and presents the key measurement data types that influenced the classification, acting as a mediator between the complex internal model representations and user comprehension, thereby maintaining both classification capability and interpretability.
Solution Approach 2:
The patent extracts the essential explanatory information from the complex model outputs by identifying and presenting only the most influential measurement data types. This extraction process separates the critical interpretive elements from the detailed state indication values, providing users with concise, actionable insights while preserving the full classification capability of the underlying model.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus is provided comprising a first acquisition unit for acquiring a data set including a plurality of types of measurement data indicating a state of an object, a supplying unit for supplying, in response to the data set being input, the data set acquired by the first acquisition unit to a model that outputs a state indication value indicating classification of a state of the object, a first identification unit for identifying, in a case where one of the state indication value is output from the model in response to one of the data set being supplied, at least one type of measurement data, among the plurality of types of measurement data, having a larger influence on the one state indication value than a reference, based on the one data set, and a display control unit for displaying the one state indication value and the at least one type of measurement data along with each other.