Metadata-Driven Credit Risk Model Execution for Parallel Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing credit risk modeling and analysis systems face challenges in efficiently processing large credit data with multiple credit risk metrics, leading to prolonged processing times and the need for specialized components or cloud platforms to manage complex data scenarios.
Innovation Solution
A system utilizing a model wrapper that generates code based on relational database metadata to execute credit risk models and methodologies without manual coding, enabling parallel and distributed processing of credit data using column-wise calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional statistical analysis systems process large credit data with multiple credit risk metrics, then comprehensive analysis can be performed, but processing time becomes prolonged
Solution Approach 1:
The patent segments the credit risk analysis into multiple independent models (e.g., PD model, LGD model, EAD model) that can be processed separately and in parallel. Each model operates on specific data partitions, enabling concurrent execution that reduces overall processing time while maintaining comprehensive analysis coverage.
Solution Approach 2:
The patent introduces a distributed computing dimension by executing models across multiple processors or nodes. The system partitions data and distributes model execution across different computational units, transforming a sequential processing approach into a parallel dimensional approach that significantly improves processing speed.
2Productivity
If specialized components or cloud platforms are used to manage complex data scenarios, then processing efficiency improves, but system complexity increases
Solution Approach 1:
The patent creates a universal framework that can handle multiple credit risk metrics and data scenarios through a single integrated system. The framework uses standardized model wrappers and metadata-driven execution that can accommodate various models without requiring specialized components for each specific analysis type.
Solution Approach 2:
The system employs self-service mechanisms where the metadata automatically describes the data structures and model requirements, enabling the system to configure and execute models autonomously without manual intervention. This reduces the need for complex specialized components by allowing the system to manage its own configuration and execution.
3Adaptability or versatility
If manual coding is used to implement credit risk models, then model customization is possible, but development time increases
Solution Approach 1:
The patent uses model wrappers that generate standardized code templates representing credit risk models. Instead of manually coding each model from scratch, the system copies and instantiates pre-defined model structures through metadata configuration, significantly reducing development time while maintaining customization capability through parameter adjustment.
Solution Approach 2:
The system enables model customization by changing parameters and metadata rather than rewriting code. The model wrappers accept configuration through metadata that defines data types, relationships, and model parameters, allowing customization through parameter modification instead of manual coding, thus reducing development time.
Data Source
AI summary
A system may include one or more processors configured to receive a set of data containing relational database metadata, model data, and customer loan data, configure a first model and a second model based on the model data and the metadata, and partition, based on the relational database metadata, the customer loan data into first data and second data. The one or more processors may execute, by executing a first execution unit, execute the first model using the first data to output a first credit risk score, execute, by executing a second execution unit different from the first execution unit, the second model using the second data to output a second credit risk score, generate a third credit risk score based on the first credit risk score and the second credit risk score, and transmit a notification based on the third credit risk score.