Struct FilterMethods 

pub struct FilterMethods;

Filter-based feature selection methods

Implementations

impl FilterMethods

pub fn variance_threshold( features: &Array2<f64>, threshold: f64, ) -> Result<FeatureSelectionResult>

Variance-based feature selection

Removes features with low variance (quasi-constant features).

Arguments

• features - Feature matrix (samples x features)
• threshold - Minimum variance threshold

Returns

• Feature selection result

Example

use scirs2_core::ndarray::Array2;
use scirs2_series::feature_selection::FilterMethods;

let features = Array2::from_shape_vec((100, 5), (0..500).map(|x| x as f64).collect()).expect("Operation failed");
let result = FilterMethods::variance_threshold(&features, 0.1).expect("Operation failed");
println!("Selected {} features", result.selected_features.len());

pub fn correlation_selection( features: &Array2<f64>, target: &Array1<f64>, threshold: f64, ) -> Result<FeatureSelectionResult>

Correlation-based feature selection

Selects features based on their correlation with the target variable.

Arguments

• features - Feature matrix (samples x features)
• target - Target variable
• threshold - Minimum absolute correlation threshold

Returns

• Feature selection result

pub fn mutual_information_selection( features: &Array2<f64>, target: &Array1<f64>, n_bins: usize, n_features: Option<usize>, ) -> Result<FeatureSelectionResult>

Mutual information-based feature selection for time series

Estimates mutual information between features and target using histogram-based approach.

Arguments

• features - Feature matrix (samples x features)
• target - Target variable
• n_bins - Number of bins for histogram estimation
• n_features - Number of top features to select

Returns

• Feature selection result

pub fn f_test_selection( features: &Array2<f64>, target: &Array1<f64>, alpha: f64, ) -> Result<FeatureSelectionResult>

F-test based feature selection

Performs F-test for feature relevance.

Arguments

• features - Feature matrix (samples x features)
• target - Target variable
• alpha - Significance level

Returns

• Feature selection result

pub fn autocorrelation_filter( features: &Array2<f64>, max_lag: usize, threshold: f64, ) -> Result<FeatureSelectionResult>

Autocorrelation-based feature filtering for time series

Selects features based on their autocorrelation structure.

Arguments

• features - Feature matrix (samples x features)
• max_lag - Maximum lag to consider
• threshold - Minimum autocorrelation threshold

Returns

• Feature selection result