Expand description
Automated Feature Engineering
This module provides automated feature engineering capabilities that can automatically generate new features from existing ones to improve model performance.
§Features
- Feature Generation: Automatically create polynomial, interaction, and transformation features
- Feature Selection: Select the most relevant features using various scoring methods
- Feature Transformation: Apply mathematical transformations to discover hidden patterns
- Feature Importance: Rank features by their predictive power
- Domain-Specific Engineering: Apply domain knowledge for specific feature types
§Examples
ⓘ
use sklears_preprocessing::automated_feature_engineering::{
AutoFeatureEngineer, AutoFeatureConfig, GenerationStrategy
};
use scirs2_core::ndarray::Array2;
fn example() -> Result<(), Box<dyn std::error::Error>> {
let config = AutoFeatureConfig::new()
.with_strategy(GenerationStrategy::Polynomial { degree: 2 })
.with_max_features(100)
.with_selection_threshold(0.01);
let mut engineer = AutoFeatureEngineer::new(config);
let data = Array2::from_shape_vec((100, 5), (0..500).map(|x| x as f64).collect())?;
let target = Array1::from_vec((0..100).map(|x| (x % 2) as f64).collect());
let engineer_fitted = engineer.fit(&data, &target)?;
let transformed = engineer_fitted.transform(&data)?;
println!("Original features: {}", data.ncols());
println!("Generated features: {}", transformed.ncols());
Ok(())
}Structs§
- Auto
Feature Config - Configuration for automated feature engineering
- Auto
Feature Engineer - Automated feature engineering transformer
- Auto
Feature Engineer Fitted - Fitted automated feature engineer
- Transformation
Function - Represents a transformation function for feature generation
Enums§
- Domain
- Domain-specific feature engineering
- Generation
Strategy - Feature generation strategies
- Math
Function - Mathematical functions for feature transformation
- Selection
Method - Feature selection methods
- Transformation
Type - Types of transformations