Synthetic Feature Screening Using a Pretrained Meta-Feature Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing feature engineering methods generate a large number of synthetic features, consuming significant time in evaluating their quality due to the need for extensive feature evaluation, particularly in machine learning processes.
Innovation Solution
A method and apparatus that involves screening synthetic features using a meta feature evaluation model to classify features into good and poor categories before evaluation, utilizing pre-trained samples and classifiers to improve efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a large number of synthetic features are generated through feature combination operations, then the quantity of features increases exponentially, but the time consumed for evaluating the quality of these synthetic features increases significantly
Solution Approach 1:
The patent divides the feature evaluation process into two segments: (1) a fast preliminary evaluation stage using a pre-trained meta feature evaluation model to quickly assess synthetic features, and (2) a detailed evaluation stage using a first classifier only for features that pass the preliminary screening. This segmentation allows the system to handle large quantities of synthetic features efficiently by applying different evaluation strategies to different subsets of features.
Solution Approach 2:
The patent implements preliminary action by pre-training a meta feature evaluation model before the actual feature evaluation process. This pre-trained model is used to perform initial quality assessment and screening of synthetic features, identifying high-quality candidates before they undergo more computationally intensive evaluation. This preliminary filtering action significantly reduces the time required for comprehensive feature evaluation.
2Measurement precision
If all synthetic features are evaluated using a first classifier, then the accuracy of feature selection is improved, but the calculation time and computational resources are excessively consumed
Solution Approach 1:
The patent applies local quality by using different evaluation methods with different levels of accuracy for different subsets of features. High-quality synthetic features identified by the meta feature evaluation model undergo detailed evaluation with the first classifier, while lower-quality features are either improved through feature engineering or discarded. This localized application of high-accuracy evaluation ensures that computational resources are focused on the most promising features.
Solution Approach 2:
The patent implements partial action by applying the computationally intensive first classifier only to a subset of synthetic features that are identified as high-quality candidates by the meta feature evaluation model, rather than evaluating all synthetic features with the same level of detail. This partial application of rigorous evaluation maintains measurement precision for the most important features while significantly improving overall productivity.
3Quantity of substance
If feature engineering performs extensive combination operations on original features, then more synthetic features are generated, but the main calculation time is consumed on evaluating the quality of new features rather than on feature generation
Solution Approach 1:
The patent introduces a meta feature evaluation model as an intermediary between feature generation and detailed feature evaluation. This intermediary model provides a fast preliminary assessment that guides the subsequent detailed evaluation process, allowing the system to maintain a large quantity of synthetic features while reducing the overall calculation time required for quality evaluation.
Solution Approach 2:
The patent changes the evaluation parameters by using a pre-trained meta feature evaluation model with optimized parameters for fast preliminary assessment, rather than using the same detailed evaluation parameters for all features. This parameter change enables efficient handling of large numbers of synthetic features while preserving the ability to perform accurate evaluation when needed.
Data Source
AI summary
A method for processing synthetic features is provided, and includes: the synthetic features to be evaluated and original features corresponding to the synthetic features are obtained. A feature extraction is performed on the synthetic features to be evaluated based on a number S of pre-trained samples, to obtain meta features with S samples. S is a positive integer. The meta features are input into the pre-trained meta feature evaluation model for a binary classification prediction, to obtain a probability of binary classification. Quality screening is performed on the synthetic features to be evaluated according to the probability of the binary classification, to obtain second synthetic features to be evaluated. The second synthetic features are classified in a good category. The second synthetic features and original features are input into a first classifier for evaluation. classified in a poor category.