Boolean Friction Segmentation for Actionable ML Explainability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning solutions provide either global explainability that is too general or local explainability that is too granular and unactionable, making it difficult for businesses to effectively mitigate undesirable predictions.
Innovation Solution
A method that segments data entities based on Boolean friction points using Shapley values to generate granular, actionable explanations by clustering data and highlighting the most significant friction points for each segment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If global explainability is provided, then statistical overview is given, but granularity is insufficient for practical mitigation
Solution Approach 1:
The patent segments the dataset into multiple segments based on Shapley values, transforming the single global explanation into multiple segment-level explanations. This segmentation provides finer granularity while maintaining manageability, as each segment represents a subset of instances with similar characteristics and friction points.
2Loss of information
If local explainability per instance is provided, then detailed explanation is given, but quantity of instances is too large to address separately
Solution Approach 1:
The patent merges individual instance explanations into segment-level explanations by grouping instances with similar Shapley value profiles and friction points. This combining reduces the number of explanations from instance-level to segment-level, making mitigation planning feasible while preserving essential explanatory details.
3Loss of information
If local explainability attributes are provided, then detailed reasons are given, but actionable attributes are masked making mitigation difficult
Solution Approach 1:
The patent extracts and highlights the most significant friction points and actionable attributes from the full set of explainability attributes for each segment. By focusing on top friction points rather than presenting all attributes, the solution makes mitigation actionable while preserving the most critical explanatory information.
Data Source
AI summary
As described herein, a system, method, and computer program provide explainability of entity data segmentation based on Boolean friction points. A dataset is processed, using a machine learning model, to calculate a plurality of Shapley values for the dataset, wherein the dataset includes friction points and explanatory variables. The dataset is clustered to generate a plurality of segments, based on the Shapley values. For each segment of the plurality of segments, a global explanation is generated for the segment using a predefined list of Boolean friction columns and the Shapley values.