Data Analysis Apparatus for Feature Variable Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning methods struggle to accurately extract non-trivial feature variables contributing to predictions, often relying on trivial features and resulting in reduced prediction accuracy when these features are excluded.
Innovation Solution
A data analysis apparatus and method that selectively operates regularization coefficients for trivial and non-trivial feature variables within a prediction model, adjusting the loss function to prioritize non-trivial features and suppress trivial ones, thereby maintaining prediction accuracy while identifying important contributing features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing machine learning methods use trivial feature variables for prediction, then prediction accuracy is maintained, but the ability to extract non-trivial feature variables contributing to prediction is reduced
Solution Approach 1:
The patent segments feature variables into two distinct groups: trivial feature variables (first feature variable group) and non-trivial feature variables (second feature variable group). This segmentation allows the system to handle different types of features differently during the learning process, enabling accurate extraction of non-trivial features while maintaining prediction accuracy through trivial features.
Solution Approach 2:
The patent applies different regularization coefficients to different groups of feature variables. Specifically, a first regularization coefficient is applied to trivial feature variables and a second regularization coefficient is applied to non-trivial feature variables. This local differentiation in quality control allows the system to suppress trivial features while preserving and highlighting non-trivial features during the learning process.
2Device complexity
If regularization is applied uniformly to all feature variables, then the learning process is simplified, but the identification of important non-trivial feature variables is compromised
Solution Approach 1:
The patent divides the feature variable set into multiple groups (trivial and non-trivial) and applies different regularization coefficients to each group. This segmentation strategy balances the simplicity of uniform regularization with the precision needed for accurate feature importance identification, as it allows differentiated treatment without completely complicating the learning process.
Solution Approach 2:
The patent changes the regularization parameter (coefficient) based on the group to which each feature variable belongs. By setting different regularization coefficients for trivial and non-trivial feature variables, the system adapts the learning process to the specific characteristics of each feature group, improving feature importance identification while maintaining reasonable process complexity.
Data Source
AI summary
A data analysis apparatus executes: a selection process selecting a first feature variable group that is a trivial feature variable group contributing to prediction and a second feature data group other than the first feature variable group from a set of feature variables; an operation process operating a first regularization coefficient related to a first weight parameter group corresponding to the first feature variable group in a manner that the loss function is larger, and operating a second regularization coefficient related to a second weight parameter group corresponding to the second feature variable group in a manner that the loss function is smaller, among a set of weight parameters configuring a prediction model, in a loss function related to a difference between a prediction result output in a case of inputting the set of feature variables to the prediction model and ground truth data corresponding to the feature variables.