Predictive Model Pipeline Automation Using Record-Graph Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional machine learning models require significant manual customization and subjective judgment from highly trained professionals, limiting their scalability and accuracy due to compounding inaccuracies from non-optimal judgments.

Innovation Solution

An enhanced pipeline that automatically generates, validates, and deploys predictive models by analyzing record relationships, iteratively optimizing hyper-parameters, and employing user-guided interfaces to determine relevant data sets, reducing the need for manual customization and subjective judgment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional ML models are deployed with manual customization by highly trained professionals, then model accuracy can be maintained through expert judgment, but the deployment scope is limited due to the scarcity of such professionals and compounding inaccuracies from subjective judgments

Engineering Contradiction:

Improvedeployment scopeVSAvoidmanual customization complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system performs automatic model generation, validation, and deployment without requiring manual customization by experts. The pipeline autonomously executes all steps including data preprocessing, model training, hyperparameter optimization, and performance evaluation, eliminating dependency on highly trained professionals while maintaining deployment scope across diverse tasks

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system automatically optimizes hyperparameters through iterative validation processes, changing model parameters systematically rather than relying on expert judgment. This includes automatic selection of learning rates, batch sizes, network architectures, and other critical parameters that previously required manual tuning by data scientists

Inventive Principle:

Principle #35Parameter changes

2Productivity

If manual customization steps are required for each predictive task, then model accuracy can be adjusted through expert judgment, but the process becomes time-consuming and scales poorly due to the need for highly trained professionals at each step

Engineering Contradiction:

Improvemodel deployment speedVSAvoidcustomization time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary automatic validation and optimization of multiple model configurations before final deployment. By pre-executing validation steps and hyperparameter searches automatically, the system eliminates the need for time-consuming manual customization during deployment, significantly improving productivity while reducing the time loss associated with expert intervention

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The pipeline enables continuous automated model generation and validation without interruption by expert intervention. The system continuously processes predictive tasks through automated workflows, maintaining productive action across multiple tasks simultaneously without the downtime required for manual customization and expert review cycles

Inventive Principle:

Principle #20Continuity of useful action

3Measurement precision

If subjective judgments are introduced at each customization step, then model can be adapted to specific tasks, but predictive performance deteriorates due to compounding non-optimal judgments from multiple data scientists

Engineering Contradiction:

Improvepredictive accuracyVSAvoidmanual intervention level

Core Design Contradiction:

Measurement precisionVSExtent of automation

Solution Approach 1:

The system implements automated feedback loops where model performance is continuously evaluated against validation datasets, and hyperparameters are adjusted based on quantitative performance metrics rather than subjective expert judgment. This feedback mechanism systematically optimizes predictive accuracy by iterating through parameter configurations and selecting those that demonstrably improve performance measures

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system replaces the mechanical process of human expert judgment with automated computational algorithms for model customization and validation. Statistical and machine learning algorithms automatically determine optimal model configurations, substituting human subjective assessment with objective, repeatable computational processes that eliminate compounding inaccuracies from multiple experts

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12387131B2Enhanced pipeline for the generation, validation, and deployment of machine-based predictive models

Publication Date: 2025.08.12 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

An enhanced pipeline for the generation, validation, and deployment of machine-based predictive models (PMs) is provided. The pipeline analyzes records to generate a graph that indicates various relationships between the records. A user provides a selection of a data element of interest (DEOI). The generated PM predicts values for the DEOI based on input records that do not include values for the DEOI. The user provides selections for values of the DEOI that represent positive outcomes associated with the DEOI. The user provides selections for values of the DEOI that represent negative outcomes associated with the DEOI. A subgraph of the graph is determined based on the DEOI. A relevant set of records is determined based on the subgraph. The PM is automatically trained, validated, and deployed based on the relevant set of records, the DEOI, and the representative values for the DEOI.