crossValidate

crossValidate(fitFn, scoreFn, X, y?, folds?): object

Defined in: ds/src/ml/validation.js:573

Execute cross-validation with a model.

Array API: crossValidate(fitFn, scoreFn, X, y, folds?)

Declarative table API: crossValidate(fitFn, scoreFn, { data, X, y, encoders? }, options?) Options can include { k, shuffle, folds } when using descriptors.

Returns: { scores, meanScore, stdScore, nFolds, metadata?, tableFolds? } When invoked with a descriptor, metadata/tableFolds include the training encoders and per-fold table views for further inspection.

Parameters

fitFn

Function

Fits a model given (XTrain, yTrain) and returns the model

scoreFn

Function

Scores a model given (model, XTest, yTest) and returns a number

X

any

Design matrix (n × p) or a table descriptor ({ data, X, y, encoders? })

y?

any = null

Response vector, or options object when X is a table descriptor

folds?

any[] = null

Optional fold definitions (each with trainIndices/testIndices); defaults to k-fold

Returns

object

Cross-validation results

meanScore

meanScore: number

metadata?

optional metadata?: any

nFolds

nFolds: number

scores

scores: number[]

stdScore

stdScore: number

tableFolds?

optional tableFolds?: any[]