Prediction Error Contribution Analysis for Model Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies cannot quantitatively evaluate the factors contributing to prediction errors in prediction models, making it difficult to improve the precision of these models.
Innovation Solution
An information processing device that calculates prediction errors and evaluates the contribution of explanatory variables, objective variables, and prediction models to these errors using indices and contribution calculation units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only binary analysis (whether or not factors are attributable) is performed, then analysis simplicity is maintained, but quantitative evaluation capability is lost
Solution Approach 1:
The patent segments the prediction error into multiple components by introducing index calculation units that separately evaluate different factors (explanatory variable abnormalities, distribution distances, model performance). This segmentation allows quantitative measurement of each factor's contribution to the overall error, resolving the contradiction between simplicity and quantitative capability.
Solution Approach 2:
The patent introduces index values as intermediary elements between the binary analysis and quantitative evaluation. These indices serve as mediators that transform qualitative factors into quantifiable metrics, enabling precise measurement while maintaining analytical structure and clarity.
2Measurement precision
If quantitative evaluation of prediction error factors is implemented, then precision improvement capability is enhanced, but analysis complexity increases
Solution Approach 1:
The complex task of quantitative error analysis is segmented into multiple independent index calculation units, each responsible for evaluating specific factors. This modular segmentation reduces overall analysis complexity by breaking down the complex evaluation into manageable, specialized components while maintaining high measurement precision.
Solution Approach 2:
The patent creates a universal framework where multiple index calculation units work together within a single analysis system. Each unit performs a specialized function but contributes to the overall quantitative evaluation, making the complex system manageable through standardized, multi-functional components.
3Ease of operation
If only distribution distance evaluation is performed, then data similarity analysis is simplified, but comprehensive error factor analysis is insufficient
Solution Approach 1:
The patent segments the comprehensive error analysis into multiple independent evaluations, including distribution distance as one component among others (explanatory variable abnormalities, model performance indices). This segmentation maintains the simplicity of distribution distance evaluation while ensuring comprehensive error factor analysis through additional parallel evaluations.
Solution Approach 2:
The patent merges multiple evaluation methods (distribution distance, abnormality detection, model performance metrics) into a unified quantitative analysis framework. This combination preserves the ease of individual evaluations while achieving comprehensive and reliable error factor analysis through their integration.
Data Source
AI summary
An information processing device 100 of the present disclosure can assist decision-making by a user by including: an error calculation unit 121 that calculates a prediction error which is a difference between a prediction value which is output obtained when an explanatory variable of subject data is input to a prediction model and an objective variable of the subject data; an index calculation unit 122 that calculates, on a basis of data that can be used for calculating the prediction error, an index for evaluating an amount of contribution of at least one of the explanatory variable of the subject data, the objective variable of the subject data, and the prediction model to the prediction error; and a contribution calculation unit 123 that calculates the amount of contribution on a basis of the prediction error and the index.