deepbox/preprocess

Imputation & Feature Engineering

Imputers, discretizers, polynomial and spline transforms, and feature-selection utilities.

Feature engineering

type fClassif

export declare const fClassif: typeof f_classif;

fClassif is a public const in deepbox/preprocess.

type fRegression

export declare const fRegression: typeof f_regression;

fRegression is a public const in deepbox/preprocess.

type ScoreFunc

export type ScoreFunc = (X: Tensor, y: Tensor) => number[];

Scoring function type for SelectKBest.

type mutualInfoClassif

export declare const mutualInfoClassif: typeof mutual_info_classif;

mutualInfoClassif is a public const in deepbox/preprocess.

type mutualInfoRegression

export declare const mutualInfoRegression: typeof mutual_info_regression;

mutualInfoRegression is a public const in deepbox/preprocess.

KBinsDiscretizer

Bin continuous data into intervals.

RFE

Recursive Feature Elimination (RFE).

RFECV

Recursive Feature Elimination with Cross-Validation (RFECV).

SelectFromModel

Meta-transformer for selecting features based on importance weights from a fitted estimator.

SelectKBest

Select features according to the K highest scores.

VarianceThreshold

Feature selector that removes all low-variance features.

KNNImputer

KNN-based imputation for completing missing values.

MissingIndicator

Binary indicator for missing values.

SimpleImputer

Simple imputation transformer for completing missing values.

Binarizer

Binarize data (set feature values to 0 or 1) according to a threshold.

FunctionTransformer

Constructs a transformer from an arbitrary callable.

PolynomialFeatures

Generate polynomial and interaction features.

SplineTransformer

Generate B-spline basis features for each input feature.

f_classif

export declare function f_classif(X: Tensor, y: Tensor): number[];

ANOVA F-value between each feature and the target classes.

f_regression

export declare function f_regression(X: Tensor, y: Tensor): number[];

Univariate F-statistic from correlation between each feature and target.

mutual_info_classif

export declare function mutual_info_classif(X: Tensor, y: Tensor, options?: { nNeighbors?: number; randomState?: number; }): number[];

Estimate mutual information between each feature and a discrete target variable.

mutual_info_regression

export declare function mutual_info_regression(X: Tensor, y: Tensor, options?: { nNeighbors?: number; randomState?: number; }): number[];

Estimate mutual information between each feature and a continuous target variable.

preprocess-features.ts

import {  KBinsDiscretizer,  KNNImputer,  PolynomialFeatures,  SelectKBest,  VarianceThreshold,} from "deepbox/preprocess";import { tensor } from "deepbox/ndarray";const X = tensor([[1, 2], [1, 3], [2, 5], [4, 8]]);console.log(new KNNImputer().fitTransform(tensor([[1, NaN], [2, 3]])).toString());console.log(new PolynomialFeatures({ degree: 2 }).fitTransform(X).toString());console.log(new KBinsDiscretizer({ nBins: 3 }).fitTransform(X).toString());console.log(new VarianceThreshold().fitTransform(X).toString());console.log(new SelectKBest({ k: 1 }).fit(X, tensor([0, 0, 1, 1])).transform(X).toString());

Encoders & Vectorizers

Splitting & Cross-Validation