Interactive Machine Learning Model Retraining via User Feedback

Solution Overview

Problem

Current machine learning applications in industries often lack direct user involvement, and users find it difficult to understand or trust predictions made by machine learning models, especially since they have no opportunity to influence the predicted outcomes.

Innovation Solution

The system allows users to interact with machine learning predictions by changing input data or retraining the model, enabling human feedback to improve the accuracy of predictions related to supply chain information and allowing users to influence the model based on their preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning models are used for predictions without user involvement, then automation and productivity are improved, but user trust and understanding of predictions deteriorate

Engineering Contradiction:

Improveprediction efficiencyVSAvoiduser trust

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system implements feedback loops where users can provide corrections to model predictions, and this feedback is used to retrain and improve the model. Users can amend predictions by moving items between clusters or modifying cluster assignments, and the system learns from these corrections to enhance future predictions, thereby building user trust while maintaining automation.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system introduces an interactive interface as an intermediary between the machine learning model and the user. This interface allows users to view predictions, make adjustments, and provide feedback without directly modifying the model code or data processing pipelines, thus maintaining automation while enabling user influence and understanding.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If machine learning models operate as black boxes without user interaction, then device complexity is reduced, but ease of operation and user control deteriorate

Engineering Contradiction:

Improvemodel structureVSAvoiduser control

Core Design Contradiction:

Device complexityVSEase of operation

Solution Approach 1:

The system allows users to directly interact with and modify prediction results through an intuitive interface. Users can move items between clusters, amend cluster assignments, and the system automatically retrains the model based on these user-driven modifications, enabling users to control and refine predictions without requiring technical expertise in machine learning.

Inventive Principle:

Principle #25Self-service

3Productivity

If clustering is performed automatically without user feedback, then productivity is improved, but measurement precision and user preference alignment deteriorate

Engineering Contradiction:

Improveclustering speedVSAvoidclustering accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system implements feedback loops where users can provide corrections to model predictions, and this feedback is used to retrain and improve the model. Users can amend predictions by moving items between clusters or modifying cluster assignments, and the system learns from these corrections to enhance future predictions, thereby building user trust while maintaining automation.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12242954B2Interactive machine learning

Publication Date: 2025.03.04 KINAXIS INC

AI summary

A computer-implemented method of interactive machine learning in which a user is provided with predicted results from a trained machine learning model. The user can take the predicted results and adjust the predicted data to retrain the model.