Synthetic Instance Generation for Machine Learning Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing technique for generating synthetic instances to improve inference accuracy in machine learning models may result in instances being generated farther from the decision boundary than intended, leading to suboptimal enhancement of estimation accuracy.
Innovation Solution
An information processing apparatus and method that acquires training instances, selects instances with uncertain prediction results from multiple machine learning models, and generates synthetic instances by combining these selected instances, allowing for more precise location of the synthetic instance in the feature space to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a virtual instance is generated by combining minority class instances using the technique in Non-patent Literature 1, then the number of training instances is increased, but the synthetic instance may be generated at a location farther from the decision boundary, reducing estimation accuracy improvement
Solution Approach 1:
The patent changes the selection parameters from simply choosing minority class instances to selecting instances that produce uncertain prediction results across multiple machine learning models. This parameter change ensures synthetic instances are generated at locations closer to decision boundaries, improving estimation accuracy while maintaining increased instance quantity
Solution Approach 2:
The patent introduces feedback by evaluating prediction results from multiple machine learning models to identify uncertain cases. This feedback mechanism guides the selection of parent instances for synthetic instance generation, ensuring they are located near decision boundaries where they will most improve model estimation accuracy
2Measurement precision
If training instances are selected based on uncertain prediction results from multiple machine learning models, then the location precision of synthetic instances near decision boundary is improved, but the selection process complexity increases
Solution Approach 1:
The patent segments the selection process into distinct steps: first identifying uncertain prediction results from multiple models, then selecting parent instances based on these uncertainties, and finally generating synthetic instances. This segmentation manages complexity by breaking down the complex selection process into manageable stages
Data Source
AI summary
In order to make it possible to generate a synthetic instance for more efficiently improve prediction accuracy of a machine learning model, an information processing apparatus (10) includes: an acquisition section (11) for acquiring a plurality of training instances; a selection section (12) for selecting, from the plurality of training instances, two or more training instances each of which derives one or more uncertain prediction results obtained using one or more machine learning models that output prediction results while using instances as input; and a generation section (13) for generating a synthetic instance by combining the two or more training instances which have been selected by the selection section (12).