Multi-stage Machine Learning for Explainable Risk Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced machine-learning techniques provide accurate but unexplainable risk assessments, while simpler models are accurate but lack explainability, making it difficult to understand the impact of input variables on predictions.
Innovation Solution
A multi-stage machine-learning approach using an explainable risk assessment model to generate initial risk indicators and provide explanatory data, followed by a second-stage model for more accurate but unexplainable risk assessments, combining high predictive capability with explainability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advanced machine-learning techniques (deep neural networks) are used for risk assessment, then prediction accuracy is improved, but explainability of the relationship between input variables and output deteriorates
Solution Approach 1:
The patent divides the risk assessment process into two distinct stages: a first stage using an explainable machine-learning model that provides both risk indicator and explanatory data, and a second stage using a more accurate but unexplainable deep neural network model. This segmentation allows each model to operate in its optimal domain - the first model provides interpretability when needed, while the second model delivers superior accuracy when the first model's risk indicator exceeds a threshold value.
2Loss of information
If simpler risk assessment models (linear or logistic regression) are used, then explainability is improved, but prediction accuracy deteriorates
Solution Approach 1:
The patent merges two different types of machine-learning models with complementary strengths - an explainable model (linear or logistic regression) and an unexplainable but more accurate model (deep neural network). The system dynamically switches between or combines the outputs of these models based on the risk level, thereby achieving both explainability and high prediction accuracy that neither model could achieve alone.
Data Source
AI summary
Certain embodiments involve providing explainable risk assessment via multi-stage machine-learning techniques. A risk assessment server can determine, in response to a risk assessment query for a target entity, a first risk indicator for the target entity by applying a first risk assessment model to predictor variables associated with the target entity. Responsive to determining that the first risk indicator indicates a risk higher than a threshold value, the risk assessment server can generate explanatory data for the predictor variables and determine a second risk indicator for the target entity by applying a second risk assessment model to the predictor variables associated with the target entity. A response message can be generated and transmitted to include the first risk indicator, the explanatory data, and the second risk indicator, for use in controlling access to one or more interactive computing environments by the target entity.