SHAP Decomposition for Credit Decision Interpretability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive data analysis solutions lack interpretability, making it difficult for financial institutions to provide explanations for adverse actions, such as credit denials, due to the complex and dynamic nature of machine learning models used.
Innovation Solution
A predictive analysis machine learning model that generates entity scores and predictive contribution reports using Baseline Shapley techniques to determine per-candidate feature contribution scores, allowing for the explanation of adverse actions by identifying the impact of each feature on the entity score.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex machine learning models are used for predictive data analysis, then prediction accuracy is improved, but interpretability deteriorates
Solution Approach 1:
The patent introduces SHAP (SHapley Additive exPlanations) as an intermediary layer between the complex machine learning model and the user. SHAP values act as a mediator that translates the model's internal decision-making process into interpretable feature contribution scores, allowing users to understand which features influenced predictions without simplifying the underlying complex model architecture.
Solution Approach 2:
The patent segments the model's prediction output into individual feature contribution components. By decomposing the overall prediction into separate SHAP values for each feature, the system maintains the accuracy of the complex model while providing granular interpretability of how each input feature contributed to the final prediction.
2Device complexity
If traditional explanation methods are used, then simplicity is maintained, but accuracy of explanation deteriorates
Solution Approach 1:
The patent replaces traditional mechanical explanation methods (such as LIME or simple rule-based explanations) with a game-theoretic approach based on Shapley values. This substitution provides mathematically rigorous and accurate explanations by calculating the marginal contribution of each feature across all possible feature combinations, rather than relying on approximate or heuristic methods.
3Reliability
If model complexity increases, then predictive power is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent pre-computes and stores SHAP values for training data during the model training phase. This preliminary action creates a reference framework that accelerates explanation generation for new predictions, avoiding the need to perform computationally intensive SHAP calculations in real-time for every prediction while maintaining high predictive power.
Data Source
AI summary
Systems, apparatuses, methods, and computer program products are disclosed for generating a predictive contribution report for an attribute using machine learning techniques. An example method includes generating an entity score for an entity using a predictive analysis machine learning model. The method further includes, in an instance the entity score fails to satisfy a determination decision threshold, selecting a reference entity from a plurality of candidate reference entities and determining a plurality of per-candidate feature contribution scores using a predictive analysis machine learning model. The method further includes generating a predictive contribution report, where the predictive contribution report includes an indication that the entity does not satisfy the determination decision threshold, and an indication of one or more candidate features associated with largest contributions to the entity score.