Prediction Evaluation Using Complementary Data and Perturbation Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies face difficulties in evaluating change attributes and amounts when a portion of attribution values in input data has defects, making it challenging to suggest effective actions for achieving desired prediction results.

Innovation Solution

An evaluation program generates complementary data patterns to address defects, determines perturbation information for changing labels, and evaluates this information based on determination results to suggest effective actions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If complementary data is generated to handle defective input data, then the ability to evaluate change attributes and amounts is improved, but the complexity of the processing system increases

Engineering Contradiction:
Improveevaluation accuracy of change attributesVSAvoidprocessing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the processing of defective data by generating multiple complementary data patterns separately, evaluating perturbation information for each pattern independently, and then synthesizing the results. This allows the complex task of handling defective data to be broken down into manageable steps that can be processed systematically.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by generating complementary data patterns before the actual evaluation of change attributes. By pre-processing the defective data into multiple valid patterns, the system prepares the data in advance, making the subsequent evaluation process more accurate and manageable.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If multiple complementary data patterns are generated and evaluated, then the reliability of action suggestions is improved, but the processing time increases

Engineering Contradiction:
Improvereliability of action suggestionsVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies partial action by generating a limited number of complementary data patterns rather than exhaustively processing all possible patterns. This allows the system to achieve sufficient reliability for action suggestions while avoiding excessive processing time that would result from evaluating every possible data variation.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system performs self-service by automatically generating complementary data patterns and evaluating perturbation information without requiring external intervention. This automated process improves reliability through consistent application of evaluation criteria while minimizing the time loss associated with manual processing.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If perturbation information is determined from multiple patterns, then the accuracy of predicted label changes is improved, but the computational resources required increase

Engineering Contradiction:
Improveaccuracy of label change predictionVSAvoidcomputational resource consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent merges the evaluation results from multiple complementary data patterns to determine the final perturbation information. By combining the findings from different patterns, the system achieves more accurate predictions of label changes while optimizing computational resource usage through shared processing components and methodologies.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4664363A1Evaluation program, evaluation method, and information processing device
Publication Date: 2025.12.17 FUJITSU LTD
  • EP4664363A1 patent drawingFigure 1
  • EP4664363A1 patent drawingFigure 2
  • EP4664363A1 patent drawingFigure 3

AI summary

An evaluation program causing a computer to execute a process. The process includes: generating, when a portion of values of a plurality of attributes included in input data has a defect, complementary data of a plurality of patterns obtained by complementing the defect in a plurality of ways; determining perturbation information including an attribute to be changed and a change amount from among the plurality of attributes of complementary data in order to change a label predicted by the complementary data of the plurality of patterns; and evaluating the perturbation information based on a determination result as to whether the perturbation information determined in the complementary data of one pattern among the plurality of patterns can change the label also with respect to the complementary data of another pattern among the plurality of patterns.