Machine Learning Model Segmentation via Confidence-Based Dataset Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting large-scale datasets for machine learning models are resource-intensive, require subject matter expertise, and can produce biased results, making them inflexible to dynamic dataset changes.
Innovation Solution
A system that trains a base machine learning model to generate a confidence score, evaluates it using an evaluation dataset, augments the dataset based on these scores, and re-trains the model to produce a final model, allowing for adaptive segmentation without external expertise and reducing bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If heuristics are used to divide the dataset into groups based on geographic region and specific characteristics, then the dataset can be segmented into sub-populations for model training, but the process requires significant resources including subject matter expertise and produces biased results
Solution Approach 1:
The system enables self-service segmentation by allowing the data itself to determine groupings through unsupervised learning algorithms. The machine learning model automatically identifies patterns and segments the dataset without requiring external subject matter expertise to define segmentation criteria, thereby reducing resource requirements while maintaining segmentation accuracy.
Solution Approach 2:
The patent replaces the manual, expert-driven mechanical process of heuristic segmentation with an automated machine learning system. Instead of experts manually analyzing and dividing data based on domain knowledge, the system uses computational algorithms to perform segmentation, substituting human expertise with automated intelligence that reduces bias and resource requirements.
2Adaptability or versatility
If heuristics are used for dataset segmentation, then initial model training can be performed, but the technique is static and hard to adapt to dynamics and evolving patterns of the dataset
Solution Approach 1:
The system implements dynamic segmentation by continuously monitoring data patterns and automatically adjusting segmentations as new patterns emerge. The machine learning model is designed to adapt to evolving data characteristics over time, allowing the segmentation structure to change dynamically rather than remaining fixed, thereby improving adaptability to evolving patterns.
Solution Approach 2:
The system incorporates feedback mechanisms where the performance and characteristics of segmented data are continuously evaluated, and this information feeds back into the segmentation process. This closed-loop approach allows the system to automatically adjust segmentation strategies based on observed data patterns, improving adaptability while maintaining implementation simplicity through automated feedback-driven adjustments.
Data Source
AI summary
Provided is a system for segmenting large scale datasets according to machine learning models based on transfer learning that includes at least one processor programmed or configured to train a base machine learning model using a training dataset to generate a trained machine learning model, evaluate the trained machine learning model using an evaluation dataset, wherein, when evaluating the trained machine learning model using the evaluation dataset, the at least one processor is programmed or configured to generate a confidence score for each data instance of the evaluation dataset with the trained machine learning model, augment the evaluation dataset based on the confidence score for each data instance of the evaluation dataset to generate an augmented evaluation dataset, and retrain the trained machine learning model using the augmented evaluation dataset to generate a final machine learning model. Methods and computer program products are also provided.