Model Chaining System with Metadata-Driven Variable Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing modeling systems lack interoperability, leading to misalignment of inputs and outputs between distinct models, resulting in inefficiencies and a reliance on human expertise for bridging gaps, and suffer from lack of transparency, customizability, granularity, and scalability, especially when chaining multiple models together to predict complex scenarios like climate change impacts.
Innovation Solution
A computer-implemented method and system for automatically chaining models by identifying likely matches between output and input variables based on metadata, with data transformations and feedback loops to enhance accuracy and scalability, allowing for the execution of Integrated Assessment Modeling (IAM) to Computable General Equilibrium (CGE) models and beyond.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If models are chained together to predict complex scenarios, then the accuracy and comprehensiveness of scenario assessment is improved, but the complexity of integrating multiple models with misaligned inputs and outputs increases
Solution Approach 1:
The patent introduces an intermediary layer consisting of standardized data schemas, metadata annotations, and transformation routines that mediate between disparate models. This intermediary infrastructure enables models with different input/output structures to be chained together by automatically translating and adapting data formats, thereby maintaining assessment accuracy while managing integration complexity.
Solution Approach 2:
The patent creates universal model interfaces and standardized data structures that can accommodate multiple different models and scenarios. By designing flexible schemas that can represent various model outputs and inputs within a common framework, the system enables diverse models to be integrated without requiring custom integration logic for each model pair.
2Adaptability or versatility
If custom software functions are written to bridge model gaps, then the interoperability between models is improved, but the time and expertise required for development increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining standardized data schemas, metadata structures, and transformation templates before model integration is needed. These pre-established frameworks contain common data representations and conversion routines that can be directly applied to connect models, eliminating the need to write custom bridging software from scratch for each integration scenario.
Solution Approach 2:
The system enables models to be self-describing through metadata annotations that automatically provide information about data formats, units, and relationships. This self-service capability allows the integration framework to automatically generate appropriate transformation routines based on metadata comparisons, reducing or eliminating the need for manual software development to bridge model gaps.
3Manufacturing precision
If detailed transformations are performed to align model outputs, then the precision of data matching is improved, but the computing time and processing overhead increases
Solution Approach 1:
The patent applies partial transformation strategies by performing only the necessary data conversions needed to achieve sufficient matching precision for each specific integration scenario. Rather than always applying comprehensive transformation pipelines, the system selectively applies transformation routines based on the actual compatibility gaps identified between models, thereby maintaining precision while minimizing unnecessary processing overhead.
Data Source
AI summary
A computer-implemented method and computing system for automatically chaining together a series of models and facilitating automatic execution of the series of models is disclosed. For each sequential pair of models in the series, a set of output variables from a first model from the pair and a set of input variables to a second model from the pair are identified. The method automatically identifies a set of likely matches, each likely match pairing one output variable from the first model with one input variable to the second model, based on a title for each variable or other metadata associated with each variable, and receives confirmation that each likely match represents an accurate association between the output variable and input variable. After executing the first model, a software module imports output from the first model as input to the second model, and the second model is executed.