Customizable Machine Learning Models via Dynamic Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine learning feature selection techniques fail to account for objective cost functions and customizable user performance criteria, leading to inefficient and inaccurate predictive models.
Innovation Solution
The method involves receiving data sets with information about individuals and specified criteria, training multiple models on different subsets of the data to classify n-grams, selecting combinations based on correlation measures, and using a cost function to evaluate performance, resulting in a customized predictive model that efficiently processes fewer input features while maintaining high predictive effectiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional feature selection techniques are used, then the model can be trained with all available features, but the model fails to account for objective cost functions and customizable user performance criteria, leading to inaccurate predictions
Solution Approach 1:
The system dynamically adapts the feature selection process to different user-defined cost functions and performance criteria. Instead of using a static feature selection method, the system re-trains models with different feature subsets based on specific user requirements, allowing the model to be customized for different prediction tasks while maintaining high accuracy.
Solution Approach 2:
The system changes the parameters of the machine learning model by selectively including or excluding features based on user-defined cost functions. By modifying which features are included in the training data (e.g., using only certain combinations of applicant attributes), the system optimizes the model to meet specific performance criteria while improving prediction accuracy for those criteria.
2Reliability
If all features are included in the training data, then the model has more information to work with, but the computing resources and data usage increase significantly
Solution Approach 1:
The system extracts only the most relevant features needed for a specific prediction task based on user-defined cost functions. Instead of using all available features, the system identifies and selects a subset of features that are most predictive of the outcome, thereby reducing computing resource requirements while maintaining or improving predictive effectiveness.
Solution Approach 2:
The system uses a partial set of features rather than all available features. By selectively including only the necessary features for each specific prediction task, the system achieves sufficient predictive power without the excessive computational cost of processing all possible features, thus improving resource efficiency.
3Measurement precision
If more features are selected for the model, then the model may capture more patterns, but the model complexity and training time increase
Solution Approach 1:
The system segments the feature selection process into multiple iterations, where different feature subsets are evaluated and selected based on user-defined cost functions. Instead of creating one complex model with all features, the system divides the problem into multiple simpler models, each trained on a specific feature subset optimized for particular performance criteria, thereby reducing overall model complexity while maintaining accuracy.
4Productivity
If traditional feature selection is used without cross-validation, then the process is simpler and faster, but the model performance cannot be reliably evaluated against specific cost functions
Solution Approach 1:
The system performs preliminary cross-validation testing with different feature subsets before final model selection. By pre-evaluating multiple feature combinations using k-fold cross-validation against user-defined cost functions, the system identifies the most effective feature set in advance, ensuring reliable performance evaluation while streamlining the final model generation process.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for customizable machine learning models. In some implementations, data is received, including (i) example data sets and (ii) data specifying one or more criteria to be assessed. A set of multiple models is trained, where each model in the set of models is trained using a training data set comprising a different subset of the example data sets. Output of the models is obtained for various example data sets, and a combination of n-grams is selected based on the outputs. The example data sets are used to train a classifier to evaluate input data with respect to the specified one or more criteria based on whether the input data includes the n-grams in the selected combination of n-grams.