Centralized Model Repository for Metadata and Output Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive and descriptive modeling systems lack a centralized, structured approach for storing and managing metadata and outputs, leading to inefficiencies in model deployment, retrieval, and integration, with ad-hoc database structures required for new models and tacit knowledge dispersed across networks.
Innovation Solution
A computer-implemented method and system for storing model metadata and outputs in a centralized repository, using a model instance identifier to track executions and store outputs, facilitating consistent retrieval and integration, and enabling real-time score management and analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If ad-hoc database structures are used for new models, then flexibility in storing different model types is improved, but system complexity and deployment time increase
Solution Approach 1:
The patent implements a universal repository structure that can store metadata and outputs for multiple types of models (predictive, descriptive, prescriptive) using a single standardized schema. The repository accepts various model types through a common interface, eliminating the need for separate ad-hoc database structures for each model type while maintaining flexibility.
Solution Approach 2:
The system segments model management into distinct modular components: a centralized repository for metadata storage, separate execution environments for different model types, and standardized interfaces for data input/output. This segmentation allows the repository to maintain a simple, consistent structure while the execution layer handles the complexity of different model implementations.
2Reliability
If centralized repository is implemented, then consistency in model output retrieval is improved, but initial implementation complexity increases
Solution Approach 1:
The patent introduces a standardized metadata schema as an intermediary layer between diverse model outputs and the centralized repository. This schema acts as a mediator that translates various model output formats into a unified structure, enabling consistent retrieval while simplifying the implementation by providing a clear translation framework.
Solution Approach 2:
The system creates standardized templates for model metadata and output structures that can be copied and reused across different model types. Instead of creating unique database structures for each model, the system copies and adapts proven templates, reducing implementation complexity while ensuring consistency.
3Loss of time
If standardized storage method is used, then deployment time is reduced, but flexibility for new model types decreases
Solution Approach 1:
The patent implements a dynamic metadata schema that can adapt to new model types while maintaining structural consistency. The standardized repository structure includes flexible fields and configurable parameters that can be dynamically adjusted to accommodate emerging model types without requiring changes to the core repository architecture, thus reducing deployment time while preserving flexibility.
Solution Approach 2:
The system uses parameterized metadata structures where key characteristics can be changed without altering the fundamental storage schema. New model types are accommodated by modifying parameters such as data types, field names, and validation rules within the standardized framework, enabling rapid deployment while maintaining structural flexibility.
4Productivity
If model metadata and outputs are stored separately, then retrieval efficiency is improved, but data integration complexity increases
Solution Approach 1:
The patent merges the storage of model metadata and outputs into a unified centralized repository while maintaining logical separation through standardized schemas. This integration allows efficient retrieval through single-query operations while the standardized structure simplifies data integration by providing consistent formats and relationships between metadata and outputs.
Solution Approach 2:
The repository structure is designed to universally handle both metadata and outputs using the same storage and retrieval mechanisms. This multi-functional approach eliminates the need for separate retrieval systems while maintaining efficiency, as the same standardized interface and query methods work for both metadata and output data.
Data Source
AI summary
Model metadata for each of a plurality of models is stored. The model metadata includes a statistical analysis technique identifier and one or more model input data identifiers. A request to execute a model is received. The request includes data identifying one of the plurality of models, and a model execution start date and end date. On the model execution start date, execution of the model associated with the model execution request is commenced. Outputs of the executed model are stored in a database. The outputs are associated with a model instance identifier, information describing a context for execution of the model, and model output type information. The outputs are retrieved, using the model instance identifier, for analysis.