Hierarchical ML Workflow Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current insurance claims processing systems rely heavily on manual human intervention for validating workflow completion, which is time-consuming and costly, and lacks efficiency in processing workflow data.
Innovation Solution
A machine learning system utilizing a hierarchical modeling process with a filter and expert model to automatically classify digital images of work actions, reducing the need for manual validation by processing images to determine if they depict before, during, or after the work action, and applying specific or parallel modeling processes based on workflow types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual human review is used to validate workflow completion, then validation accuracy can be maintained through careful inspection, but processing time increases and costs increase
Solution Approach 1:
The patent segments the validation process into two distinct machine learning models: a filter model that performs initial screening to identify potentially fraudulent workflows, and an expert model that provides detailed analysis and validation for flagged cases. This segmentation allows the system to maintain high validation accuracy through specialized analysis while reducing overall processing time by automatically filtering out legitimate cases without requiring full expert review.
Solution Approach 2:
The filter model serves as an intermediary between complete manual review and automated processing. It pre-screens workflows and identifies only those requiring expert model analysis, thereby reducing the volume of cases needing detailed review while maintaining accuracy through the expert model's focused analysis on suspicious cases.
2Measurement precision
If manual human review is used to validate workflow completion, then thorough inspection can be performed, but labor costs increase
Solution Approach 1:
The system enables self-service validation through automated machine learning models that independently analyze workflow data without requiring human intervention for routine cases. The filter and expert models work autonomously to validate workflows, reducing dependency on human labor while maintaining validation accuracy through sophisticated algorithmic analysis.
Solution Approach 2:
By segmenting validation into filter and expert models, the system allocates human resources efficiently—humans only need to review cases flagged by the filter model, significantly reducing total labor hours while maintaining thorough inspection through the expert model's detailed analysis of suspicious cases.
3Productivity
If automated machine learning models are used to validate workflows, then processing speed increases, but system complexity increases
Solution Approach 1:
The patent reduces system complexity by segmenting the machine learning architecture into two specialized models with distinct functions: the filter model handles high-volume initial screening, while the expert model provides detailed analysis only when needed. This segmentation allows each model to be optimized for its specific task, improving processing speed while keeping individual model complexity manageable.
Solution Approach 2:
The filter model acts as an intermediary layer that simplifies the overall system architecture by handling routine validation cases independently. This intermediary structure prevents the need for a single complex model to handle all cases, thereby reducing overall system complexity while maintaining high processing speed through efficient task distribution.
4Reliability
If comprehensive workflow validation is performed, then payment accuracy is ensured, but processing time increases
Solution Approach 1:
The patent segments payment validation into two stages: the filter model quickly screens workflows to identify potential issues, and the expert model performs comprehensive validation only on flagged cases. This ensures payment accuracy through thorough expert review of suspicious cases while reducing overall processing time by rapidly clearing legitimate cases through automated filtering.
Solution Approach 2:
The system implements feedback mechanisms where the filter model learns from expert model decisions and adjusts its screening criteria accordingly. This feedback loop improves payment accuracy over time by refining the filter's ability to identify truly suspicious cases, thereby reducing the burden on expert review while maintaining comprehensive validation where needed.
Data Source
AI summary
Machine learning systems and methods for validating workflows are provided. The system receives one or more digital images of a work action being performed, and processes the image using a hierarchical modeling process that includes a filter model and a cascaded expert model. The filter model processes the image to ascertain whether the image is suitable for use in validating completion of the work action. If the system determines that the image is suitable, the image is then processed by the expert model to classify whether the image depicts a scene occurring before, during, or after performance of the work action. The hierarchical modeling process can be utilized to validate different workflows based upon a specified workflow type. A plurality of hierarchical modeling processes can be applied in parallel to an input image in the event that a type of workflow to be validated is not specified in advance.