Heuristic False Positive Reduction via Training Data Duplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive models, such as decision trees, face limitations in accuracy due to the size of the training corpus, leading to high false positives and negatives, necessitating re-training with misclassified samples, which can be costly and time-consuming.
Innovation Solution
The method involves training a heuristic, deploying it, identifying false positives in the field, duplicating them, modifying the training data to include these duplicates, and re-training the heuristic to amplify their weight, thereby reducing false positives without altering underlying algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the training corpus size is increased to improve heuristic accuracy, then false positives and negatives are reduced, but development time and computational cost increase significantly
Solution Approach 1:
The patent applies partial action by selectively duplicating only the misclassified samples (false positives and false negatives) rather than duplicating the entire training corpus. This targeted approach amplifies the weight of problematic cases without the computational burden of processing all training data multiple times, thereby improving accuracy while minimizing additional development time.
Solution Approach 2:
The patent uses copying by creating duplicate copies of misclassified samples and adding them back to the training corpus. These copied samples allow the heuristic to learn from its mistakes without requiring external data sources or significant corpus expansion, efficiently addressing accuracy issues through internal data reuse.
2Reliability
If the training corpus is re-trained with misclassified samples to reduce false positives, then accuracy improves, but the underlying algorithm complexity increases
Solution Approach 1:
The patent changes the parameter of sample weighting by duplicating misclassified samples within the existing algorithm framework. Instead of modifying the underlying algorithm's mathematical formulas or decision logic, it alters the distribution parameters of the training data, allowing the same algorithm to achieve better performance through adjusted input characteristics.
3Measurement precision
If misclassified samples are added to the training corpus and the heuristic is re-trained, then classification accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The patent applies partial action by focusing computational resources only on duplicating and re-processing misclassified samples rather than re-training on the entire corpus. This selective approach concentrates computational effort where it is most needed, improving classification accuracy while minimizing overall resource consumption.
Data Source
AI summary
An exemplary method for reducing false positives produced by heuristics may include: 1) training a heuristic using a set of training data, 2) deploying the heuristic, 3) identifying false positives produced by the heuristic during deployment, and then 4) tuning the heuristic by: a) duplicating at least a portion of the false positives, b) modifying the training data to include the duplicate false positives, and c) re-training the heuristic using the modified training data. Corresponding systems and computer-readable media are also disclosed.