Decision Tree Scoring for Condition-Specific Causal Graph Generation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional causal discovery techniques struggle to determine appropriate conditions for classifying data and generating causal graphs, leading to decreased accuracy and increased workload, especially when users lack prior knowledge.

Innovation Solution

An information processing method that iteratively generates and updates decision trees to identify proper conditions for classifying data and generating causal graphs, using methods like LinGAM or No-tears, and evaluates the likelihood of generated causal graphs to optimize the decision tree process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If conventional causal discovery techniques are used to generate causal graphs, then the process can be automated, but the accuracy of the causal graphs decreases when users lack prior knowledge to determine appropriate classification conditions

Engineering Contradiction:
Improveautomation of causal graph generationVSAvoidaccuracy of causal graph
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The system automatically determines optimal classification conditions by having the computer itself evaluate multiple decision trees and their corresponding causal graphs, rather than requiring user input. The computer autonomously scores different classification approaches and selects the best one based on evaluation metrics.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system implements a feedback mechanism where causal graphs are evaluated using scoring functions, and this evaluation feedback is used to iteratively improve the classification conditions. The system continuously refines the decision trees based on the quality assessment of generated causal graphs.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If multiple decision trees are generated and evaluated to find optimal classification conditions, then the accuracy of causal graph generation is improved, but the computational complexity and processing time increase

Engineering Contradiction:
Improveaccuracy of causal graph generationVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by pre-generating multiple decision trees and their corresponding causal graphs before final selection. This allows the system to evaluate multiple classification approaches in advance and choose the optimal one, rather than iteratively searching through possibilities.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system segments the causal graph generation process into multiple independent decision trees, each representing a different classification condition. This allows parallel evaluation of different classification approaches and simplifies the overall complexity by breaking down the problem into manageable segments.

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If operators manually determine classification conditions for causal graph generation, then the accuracy may be improved, but the workload and time required increase significantly

Engineering Contradiction:
Improveaccuracy of causal graphVSAvoidtime required for condition determination
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system eliminates the need for operator intervention in determining classification conditions by implementing automatic evaluation and selection of decision trees. The computer autonomously performs the entire process of generating, evaluating, and selecting optimal classification conditions.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system changes the approach from manual parameter selection to automated parameter optimization by using scoring functions and evaluation metrics to automatically determine the optimal classification conditions, thereby reducing time loss while maintaining accuracy.

Inventive Principle:
Principle #35Parameter changes

4Device complexity

If a single causal graph is generated without classification, then the process is simple, but the ability to capture condition-specific relationships is lost

Engineering Contradiction:
Improvesimplicity of processVSAvoidability to capture condition-specific relationships
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The system segments the data into multiple groups based on different classification conditions represented by decision trees. Each segment is then analyzed to generate a specific causal graph, allowing the system to capture condition-specific relationships while maintaining a structured and manageable process.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system implements a dynamic approach where the causal graph generation process adapts to different classification conditions. Instead of a static single-graph approach, the system dynamically generates multiple causal graphs based on evaluated decision trees, enhancing versatility while keeping the process systematic.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4647973A1Recording medium, information processing method, and information processing device
Publication Date: 2025.11.12 FUJITSU LTD
  • EP4647973A1 patent drawingFigure 1
  • EP4647973A1 patent drawingFigure 2
  • EP4647973A1 patent drawingFigure 3

AI summary

A method for processing information for a first and second decision tree that are mutually different, the former representing, by a path from a root to a leaf, a condition for classifying multiple data from which a causal graph is generated, the latter being based on the former, the method being executed by a computer and including: generating a first causal graph, for each path from the root to a leaf of the first decision tree, based on data that among the multiple data, meets the condition; after generating the first causal graph, repeatedly performing until an end condition is satisfied: generating a second causal graph, for each path from the root to a leaf of the second decision tree, based on the data that meets the condition; and updating the first decision tree with the second decision tree when a second score evaluating a likelihood of the second causal graph is better than a first score evaluating a likelihood of the first causal graph.