AI Model Training via Majority Class Sub-group Segmentation
Solution Overview
Problem
Imbalanced training datasets in AI models for medical diagnostics often lead to inaccurate predictions, as they predominantly feature majority class samples over minority class samples, resulting in inefficient processor and memory usage and prolonged training times.
Innovation Solution
The method involves creating sub-group training datasets by dividing the majority class samples based on specific features, balancing the distribution of samples, and training AI models using these datasets to enhance prediction accuracy and reduce resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If imbalanced training datasets are used with vastly more majority class samples than minority class samples, then the training dataset comprehensively covers all patient data, but the AI models do not accurately predict medical conditions in the minority class
Solution Approach 1:
The patent segments the majority class samples into multiple sub-groups based on shared features (e.g., demographic characteristics, disease risk factors). Each sub-group is then paired with minority class samples to create separate balanced training datasets. This segmentation allows the model to learn distinct patterns within each subgroup while maintaining overall diversity, resolving the contradiction between comprehensive data coverage and accurate minority class prediction.
2Reliability
If techniques like under-bagging are used to address imbalance, then AI models can be generated, but a large number of AI models are created increasing memory consumption, processor consumption, and training time
Solution Approach 1:
Instead of creating numerous separate AI models through under-bagging, the patent segments the training data into sub-groups and trains a single AI model on each segmented dataset. This approach maintains prediction reliability by ensuring the model learns from balanced representations of all patient subgroups, while avoiding the computational overhead of maintaining multiple separate models in an ensemble.
3Adaptability or versatility
If a large number of AI models are created to address imbalance, then prediction coverage is improved, but processor consumption and training time increase
Solution Approach 1:
The patent segments the training data into multiple sub-groups based on feature similarities, then trains a single AI model on these segmented balanced datasets. This segmentation strategy improves prediction coverage by ensuring the model learns from diverse patient subgroups, while maintaining training efficiency by avoiding the need to train and maintain multiple separate models.
Solution Approach 2:
The patent creates a universal AI model that can handle all patient subgroups through single model training on segmented balanced data. This multi-functional approach allows the model to generalize across different patient populations without requiring separate specialized models, thereby improving prediction coverage while maintaining productivity.
Data Source
AI summary
Various systems and methods are provided for training and using a diagnostic model including artificial intelligence (AI) models. The diagnostic model including the AI models may be trained by receiving training data including a majority class of samples corresponding to medical data of patients that do not have the medical condition and a minority class of samples corresponding to medical data of patients that do have the medical condition, determining sub-groups of the majority class of samples based on features of the majority class of samples, generating sub-group training datasets that each include respective samples of the sub-groups of the majority class of samples and samples of the minority class of samples, and training the AI models of the diagnostic model using the sub-group training datasets. The diagnostic model including the AI models may be used to determine whether a patient has a medical condition.