Automated Predictive Model Deployment Using Modular Code Wrappers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for deploying predictive models are inefficient and resource-intensive, requiring significant time and effort, especially during refitting and deployment phases.
Innovation Solution
A method and system for automating the deployment of predictive models by independently managing component inputs, utilizing feature generation, data grouping, and modeling codes, and applying a model wrapper to facilitate rapid deployment and refitting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional methods are used for deploying predictive models, then model deployment can be achieved, but the process is inefficient and resource-intensive requiring significant time and effort
Solution Approach 1:
The patent segments the model deployment process into independent components: feature generation code, data grouping code, modeling code, and model wrapper code. Each component can be developed, tested, and updated independently through separate code repositories, eliminating the need to redeploy entire models when making minor adjustments to individual components.
Solution Approach 2:
The patent implements preliminary action by pre-compiling and storing feature generation code, data grouping code, and modeling code in separate repositories before deployment. The model wrapper is pre-configured to orchestrate these components, allowing rapid deployment by simply assembling pre-prepared components rather than building everything from scratch during deployment.
2Productivity
If traditional deployment methods are used, then models can be deployed, but the process requires significant computational resources and is resource-intensive
Solution Approach 1:
The patent creates reusable copies of code components (feature generation code, data grouping code, modeling code) that are stored in repositories and can be copied and applied multiple times across different model deployments. This eliminates redundant computation by reusing proven code rather than redeveloping and retesting similar functionality for each deployment.
Solution Approach 2:
The model wrapper serves as a universal orchestration layer that can work with multiple different combinations of feature generation code, data grouping code, and modeling code. This multi-functional design allows the same wrapper infrastructure to support various model types and updates without requiring separate deployment mechanisms for each case.
3Adaptability or versatility
If models are refitted with updated code, then model improvements can be made, but the entire deployment process must be repeated consuming additional time and resources
Solution Approach 1:
The patent implements a dynamic deployment architecture where the model wrapper can selectively incorporate updated code components from repositories without requiring complete redeployment. When feature generation code, data grouping code, or modeling code is updated, the wrapper can dynamically assemble the new configuration by pulling updated components while maintaining the overall deployment structure.
Solution Approach 2:
The patent applies local quality by allowing specific components (feature generation, data grouping, or modeling code) to be updated independently while leaving other components unchanged. This enables targeted updates to only the portions of the model that need improvement, rather than forcing complete redeployment of the entire model system.
4Reliability
If comprehensive code updates are performed for model refitting, then model performance can be improved, but the complexity of managing multiple code versions and dependencies increases
Solution Approach 1:
The patent introduces the model wrapper as an intermediary layer between the various code components and the deployment target. This wrapper manages the complexity of coordinating feature generation code, data grouping code, and modeling code by providing a standardized interface and orchestration logic, simplifying the management of code versions and dependencies.
Solution Approach 2:
The patent implements feedback mechanisms through the code repository system that tracks versions and dependencies of feature generation code, data grouping code, and modeling code. This feedback loop allows the system to automatically manage code relationships, validate compatibility between components, and coordinate updates across multiple components without manual intervention.
Data Source
AI summary
A method includes obtaining feature generation code from, which is configured to determine features relating to input data. The method further includes obtaining data grouping code, which is configured to generate training data by determining a plurality of data groupings for the features relating to the input data. The method further includes obtaining modeling code, which is derived at least in part by applying one or more machine learning algorithms to the training data. The method further includes applying a model wrapper code to the feature generation code, the data grouping code, and the modeling code to generate a model wrapper and deploying the model wrapper such that the model wrapper may receive a first application programming interface (API) call including an input data value, determine a score relating to the input data value, and send a second API call including the score in response to the first API call.