Explainable Complex Model via Attribution Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex machine learning models in regulated industries face challenges in providing transparent and compliant explanations for their decisions, particularly in industries like finance, where regulations require users to understand the reasons behind outcomes, such as loan approvals or denials.
Innovation Solution
A method and system that generate human-readable explanations by identifying the features with the greatest impact on the outcome of a complex machine learning model, using attribution values like Shapley values, and mapping these to compliance regulations to provide reasons for decisions, ensuring compliance with regulations like the Fair Credit Reporting Act.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If complex machine learning models are used to improve decision accuracy and speed, then productivity and measurement precision are improved, but device complexity increases and explainability deteriorates
Solution Approach 1:
The patent introduces an explanation generation system as an intermediary component that sits between the complex machine learning model and the user. This intermediary translates the model's internal decisions into human-readable explanations, allowing users to understand complex decisions without reducing the model's complexity or sacrificing accuracy.
Solution Approach 2:
The system segments the explanation generation process into distinct components: feature importance calculation, attribute mapping, and explanation formulation. By dividing the complex model into manageable segments with specific functions, the system maintains model complexity while improving explainability through structured decomposition of the decision-making process.
2Measurement precision
If complex machine learning models are used to improve decision accuracy, then measurement precision is improved, but ease of operation deteriorates due to lack of explainability
Solution Approach 1:
The explanation generation system acts as a mediator that bridges the gap between the accurate but opaque machine learning model and the user who needs to understand the decision. It translates technical model outputs into accessible explanations without compromising the model's accuracy.
Solution Approach 2:
The system provides feedback to users about why specific decisions were made by identifying and explaining the most important features that influenced the outcome. This feedback mechanism improves ease of operation by giving users insight into the decision-making process while maintaining high measurement precision.
3Ease of operation
If features with highest attribution values are identified to provide explanations, then ease of operation is improved through better understandability, but loss of information occurs by focusing on only the top feature
Solution Approach 1:
The system segments the feature importance analysis by ranking features based on their attribution values and presenting them in order of importance. This segmentation allows the system to focus on the most impactful features for explanation while maintaining the hierarchical structure that preserves information about less important features.
Solution Approach 2:
The explanation generation applies local quality by tailoring the level of detail to each feature's importance. Top-rated features receive detailed explanations, while less important features are summarized or omitted, optimizing understandability without completely losing information about the decision-making process.
Data Source
AI summary
Certain aspects of the present disclosure provide techniques for generating a human readable summary explanation to a user for an outcome generated by a complex machine learning model. In one embodiment, a risk assessment service can receive a request from a user in which a risk model of the risk assessment service performs a specific task (e.g., determining the level of risk associated with the user). Once the risk model determines the risk associated with the user, in order to comply with regulations from a compliance system, the risk model can provide a user with an explanation as to the outcome for transparency purposes.