Prediction Target Clustering to Isolate Underperforming Borrower Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated prediction mechanisms struggle with accuracy issues, particularly when modifying them to improve performance in specific categories, as they fail to distinguish between individual entities within a category that performed well and those that did not, leading to suboptimal adjustments.
Innovation Solution
The technique involves identifying clusters of individual prediction targets (IPTs) into 'UE' and 'PE' groups, determining the 'DIFF-SET' variables and critical ranges that differentiate them, and using these to fine-tune the prediction mechanism, focusing on the differences between UEs and PEs within underperforming categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If automated prediction mechanisms are modified to improve accuracy in underperforming categories, then prediction accuracy for those categories improves, but the mechanisms fail to distinguish between individual entities that performed well and those that did not, leading to suboptimal adjustments that may penalize well-performing entities
Solution Approach 1:
The patent segments the population of individual prediction targets into distinct clusters based on their performance characteristics and feature similarities. By dividing the population into clusters of similarly-situated entities, the system can apply targeted modifications to improve predictions for underperforming clusters without affecting well-performing individual entities, thus resolving the contradiction between improving prediction accuracy and avoiding penalty to well-performing entities.
Solution Approach 2:
The patent applies local quality by making different parts of the prediction mechanism have different functions. Specifically, it identifies clusters of underperforming entities and applies targeted adjustments only to those specific clusters rather than uniformly across all entities. This allows the system to improve prediction accuracy for problematic groups while preserving the effectiveness for well-performing entities, thereby avoiding the penalty issue.
2Measurement precision
If manual adjustments are made to improve predictions for underperforming categories, then overall category performance improves, but the adjustments are complicated and error prone
Solution Approach 1:
The patent implements self-service by enabling the automated prediction mechanism to automatically identify underperforming clusters and generate appropriate adjustments without requiring manual intervention. The system autonomously analyzes performance data, segments entities into clusters, determines which clusters need improvement, and applies targeted modifications, thereby eliminating the complexity and error-proneness associated with manual adjustments while maintaining improved prediction accuracy.
Solution Approach 2:
The patent employs feedback mechanisms by continuously monitoring the performance of prediction targets, identifying underperforming clusters based on actual outcomes, and using this feedback information to automatically adjust the prediction mechanism. This closed-loop feedback system enables the mechanism to self-correct and improve accuracy for specific clusters without manual intervention, reducing modification complexity while maintaining high prediction accuracy.
3Measurement precision
If the prediction mechanism is adjusted to penalize all entities in an underperforming category, then category-level prediction accuracy improves, but well-performing individual entities within that category are incorrectly penalized
Solution Approach 1:
The patent segments the population into clusters of similarly-situated entities based on feature similarity and performance characteristics. By identifying and working with clusters rather than treating all entities in a category uniformly, the system can target adjustments specifically to underperforming clusters while excluding well-performing individual entities, thus improving category-level accuracy without compromising individual entity prediction reliability.
Solution Approach 2:
The patent applies local quality by making the prediction mechanism's adjustments localized to specific underperforming clusters rather than applying uniform penalties across entire categories. This allows the system to improve predictions for problematic local groups while preserving accurate predictions for well-performing individuals within the same category, thereby maintaining both category-level and individual-level prediction reliability.
Data Source
AI summary
Techniques are provided for forming clusters of individual prediction targets (IPTs). An initial prediction target is a target for which an automated prediction has been generated. IPTs may be, for example, borrowers to which a lending entity has extended loans based on predictions generated by a credit policy. Each cluster includes (a) a “core” of underperforming entities (UEs), and (b) a set of boundary performant entities (PEs). The UEs that belong to the UE core of a cluster are “similarly situated” relative to the values of their features. For example, in the context where the IPTs are borrowers, the UEs at the core of a cluster may correspond to defaulting borrowers that had similar bureau data, lending entity data, and borrower data. The boundary performant entities of the cluster may be borrowers that have not defaulted, but had similar credit qualifications as the UEs of the cluster. Having formed these clusters, the clusters may be used in a variety of ways, including but not limited to improving the accuracy of the credit model, identifying potentially problematic future borrowers, generating visualizations that illustrate the relative importance of clusters of defaulting borrowers, etc.