Data Classification Method Adjusting Feature Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning-based data classification methods face a decrease in precision when the distribution of training data differs from the true distribution of the data being classified, leading to inaccurate classification results.
Innovation Solution
A data classification method that calculates appearance probabilities of feature quantities in the training data and adjusts the weighting of these features based on their likelihood in the true distribution, ensuring that the model generated reflects the true distribution of the data for more accurate classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is performed using training data with a distribution different from the true distribution, then the classification process can be completed, but the classification precision decreases
Solution Approach 1:
The patent changes the parameter of feature quantity weighting by calculating appearance probabilities that reflect the true distribution. Instead of using fixed weights from training data with mismatched distribution, the system dynamically adjusts weights based on calculated appearance probabilities, thereby adapting to the true distribution and improving classification precision.
Solution Approach 2:
The patent introduces a feedback mechanism where the system calculates appearance probabilities based on the true distribution and uses this information to adjust feature quantity weights. This closed-loop approach continuously refines the classification model by incorporating distributional information, ensuring the classifier adapts to the true data distribution.
2Measurement precision
If the weighting of feature quantities is adjusted based on appearance probabilities, then the classification precision is improved, but the computational complexity increases
Solution Approach 1:
The patent performs preliminary calculation of appearance probabilities before the actual classification process. By pre-computing the probabilities that reflect the true distribution, the system prepares the adjusted weights in advance, which then simplifies the subsequent classification operations rather than increasing overall complexity.
Data Source
AI summary
A data classification method which classifies a plurality of data into a plurality of classification items based on a feature quantity included in the data, the method includes calculating, by a processor, an appearance probabilities in which training data including the feature quantity appears in the classification items in a distribution of the data, generating, by the processor, a rule having the feature quantity and a weighting of the feature quantity based on a plurality of the training data having the feature quantity based on the appearance probabilities; and classifying, by the processor, the plurality of data according to the rule.