Statistical Approximation for Real-Time Credit Underwriting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current underwriting processes for financial service accounts rely on trial and error, assigning the same underwriting criteria to all clients within a segment, leading to suboptimal decisions for individual clients, as the optimization problem is complex and time-consuming, resulting in best guesses rather than optimal values.
Innovation Solution
A computer-implemented method and system that uses historical account data and an optimization model to determine optimal credit limits for individual clients, generating a statistical approximation based on account data criteria, enabling real-time, close-to-optimal underwriting decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual account-level optimization is performed using the LVM for each client, then underwriting precision is improved, but computational time and complexity increase significantly
Solution Approach 1:
The patent pre-computes and stores optimal underwriting values for different client segments and scenarios before actual underwriting decisions are needed. This allows the system to quickly retrieve pre-calculated optimal values during real-time underwriting without performing complex LVM computations on-the-spot, thus resolving the contradiction between precision and computational time
Solution Approach 2:
The patent creates simplified copies or approximations of the complex LVM optimization model that can be executed quickly. These copies capture the essential relationships and patterns from the full model but use reduced complexity algorithms or pre-computed lookup tables, enabling fast individualized underwriting decisions without full computational overhead
2Productivity
If segment-level aggregation is used to simplify computations, then productivity is improved, but underwriting precision deteriorates as individual client variations are lost
Solution Approach 1:
The patent applies different levels of analysis to different aspects of underwriting: segment-level aggregation is used for identifying general patterns and trends, while individual client-level analysis is applied to specific decision-making. This hierarchical approach allows efficient use of segment data while preserving individual client nuances where they matter most for underwriting decisions
Solution Approach 2:
The patent divides the client base into segments for efficient group-level analysis, but then further segments or individualizes the approach within each segment for final decision-making. This multi-level segmentation strategy maintains productivity benefits of grouping while recovering individual client information needed for precise underwriting
Data Source
AI summary
Systems and methods are disclosed for determining one or more credit lines based on statistical approximations of credit line optimization models. By approximating the results of the optimization model, the disclosed embodiments may provide real-time account level credit line determinations based on fewer criteria than used in the optimization models. Other aspects of the disclosed embodiments are described herein.