Artificial Case Generation for Machine Learning Prediction Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for generating artificial cases for machine learning models do not effectively improve predictive performance, as they often produce redundant cases that do not contribute to accuracy and may distort the original data distribution.

Innovation Solution

An information processing device and method that selectively generates and adds artificial cases based on actual cases where the model's prediction is uncertain, focusing on cases close to the decision boundary and ensuring diversity to enhance prediction accuracy without over-generating redundant cases.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If artificial cases are generated based on actual cases, then the number of training cases is increased, but the predictive performance of the machine learning model does not necessarily improve

Engineering Contradiction:

Improvenumber of training casesVSAvoidpredictive performance

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The patent applies local quality by generating artificial cases with different degrees of similarity to actual cases. Specifically, it creates multiple artificial cases with varying similarity levels (high, medium, low) to different actual cases, rather than uniformly generating all artificial cases with the same similarity. This allows the system to selectively use artificial cases that locally match the characteristics of uncertain actual cases, thereby improving predictive performance in specific regions of the feature space where the model is uncertain.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent implements partial action by generating only the necessary number of artificial cases with appropriate similarity levels, rather than excessively generating all possible artificial cases. It selectively generates artificial cases based on the uncertainty of actual cases and their similarity relationships, avoiding redundant generation. This partial generation approach efficiently increases training data quantity while maintaining model reliability by focusing computational resources on the most beneficial artificial cases.

Inventive Principle:

Principle #16Partial or excessive action

2Quantity of substance

If multiple artificial cases are generated from each actual case, then more training data is available, but redundant cases are created that do not contribute to accuracy improvement

Engineering Contradiction:

Improvenumber of training casesVSAvoidredundant cases

Core Design Contradiction:

Quantity of substanceVSObject-generated harmful factors

Solution Approach 1:

The patent applies parameter changes by varying the similarity parameter when generating artificial cases. Instead of generating all artificial cases with a fixed similarity level to their source actual cases, it creates artificial cases with different similarity parameters (high, medium, low). This parameter variation ensures that the generated artificial cases have diverse characteristics and reduces redundancy by spreading the artificial cases across different similarity ranges, thereby improving the effectiveness of training data expansion.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If artificial cases are added to training data, then the model can handle more cases, but the original data distribution may be distorted

Engineering Contradiction:

Improvemodel handling capabilityVSAvoiddata distribution

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent applies local quality by generating artificial cases with different similarity levels to preserve local data distribution characteristics. High-similarity artificial cases maintain the local structure near actual cases, medium-similarity cases extend to intermediate regions, and low-similarity cases explore broader areas. This multi-level similarity approach allows the model to handle more diverse cases while preserving the original data distribution at different scales, preventing distortion of the underlying data structure.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20240403723A1Information processing device, information processing method, and recording medium

Publication Date: 2024.12.05 NEC CORP

AI summary

In an information processing device, an input means acquires each actual case formed by features. An artificial case generation means generates a plurality of artificial cases based on each acquired actual case. An artificial case selection means configured to select each artificial case in which a prediction of a machine learning model is to be uncertain, from the plurality of artificial cases. After that, an output means outputs each selected artificial case.