Machine Learning Model Retargeting via Extrapolated Feedback Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine-learning model training techniques for service provider systems are inefficient and resource-intensive, requiring significant computational resources and time, and necessitate retraining from scratch for even minor changes, such as switching between different event labels, which hinders the management of digital services.
Innovation Solution
The implementation of machine-learning model retargeting techniques that allow for training using extrapolated feedback data to identify a wider range of thresholds and entities, enabling the model to be retargeted for secondary labels without starting from scratch, thus improving efficiency and accuracy in predicting entity failures and managing service operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine-learning model training techniques are used to train a model for a primary label, then the model achieves accurate predictions for that label, but significant computational resources and time are consumed, and the model must be retrained from scratch for any changes
Solution Approach 1:
The patent applies preliminary action by collecting and storing feedback data during normal model operation before actual retraining is needed. This feedback data includes predictions, actual outcomes, and entity attributes. When model retargeting is required, this pre-collected data serves as a foundation, eliminating the need to start training from scratch and significantly reducing both time and computational resource requirements while maintaining prediction accuracy
Solution Approach 2:
The patent changes the parameter of training data scope by using extrapolated feedback data that covers a wider range of thresholds and entities than originally collected. This allows the model to be retargeted to secondary labels with different threshold ranges without requiring complete retraining, thus reducing training time while preserving accuracy through the use of expanded parameter ranges in the feedback data
2Reliability
If conventional machine-learning model training techniques are used with extensive training data, then the model achieves reliable predictions, but significant computational resources are consumed
Solution Approach 1:
The patent applies universality by creating feedback data that serves multiple functions: it validates model predictions, provides training data for retargeting to secondary labels, and enables exploration of additional thresholds. This multi-functional feedback data reduces the need for separate extensive training datasets, thereby maintaining model reliability while reducing computational resource consumption
Solution Approach 2:
The patent implements feedback by systematically collecting actual outcomes compared to model predictions and using this feedback data for model retargeting. This continuous feedback loop ensures model reliability is maintained while the feedback data itself serves as training material, reducing the need for additional computational resources for separate training processes
3Adaptability or versatility
If the machine-learning model is retrained from the beginning for different labels or thresholds, then the model adapts to new requirements, but the process is inefficient and resource intensive
Solution Approach 1:
The patent applies preliminary action by pre-collecting feedback data that includes entity attributes, predictions, and actual outcomes across various thresholds. This pre-collected data serves as a foundation for rapid retargeting to different labels, enabling model adaptability without requiring complete retraining and thus maintaining high training efficiency
Solution Approach 2:
The patent changes parameters by using feedback data that spans wider threshold ranges and includes diverse entity attributes. This allows the model to adapt to different labels and threshold requirements by leveraging existing feedback data with varied parameter ranges, rather than retraining from scratch, thereby maintaining both adaptability and productivity
4Quantity of substance
If feedback data is collected only within a specific threshold range, then the data is manageable, but the ability to identify wider ranges of thresholds and entities is limited
Solution Approach 1:
The patent applies dimensionality change by extrapolating feedback data beyond the original threshold range into new dimensional spaces. This allows the system to identify entities and thresholds outside the originally collected range without proportionally increasing data volume, thus expanding threshold range coverage while managing data quantity through intelligent extrapolation rather than brute-force collection
Data Source
AI summary
Machine-learning model retargeting techniques are described. In one example, training data is generated by extrapolating feedback data collected from entities. These techniques supports an ability to identify a wider range of thresholds and corresponding entities than those available in the feedback data. This also provides an opportunity to explore additional thresholds than those used in the past through extrapolating operations outside of a range used to define a segment, for which, the feedback data is captured. These techniques also support retargeting of a machine-learning model for a secondary label that is different than a primary label used to initially train the machine-learning model.