Function Signature function crossValScore ( createEstimator : () => CrossValEstimator , X : Matrix , y : Vector , options ?: CrossValScoreOptions ) : number []
Parameters createEstimator
() => CrossValEstimator
required
Factory function that returns a new estimator instance for each fold. Must return an object with fit() and predict() methods.
Training data features (2D array)
Training data target values (1D array)
Configuration options for cross-validation cv
number | CrossValSplitter
default: "5"
Cross-validation splitting strategy. Can be:
Integer: number of folds (uses StratifiedKFold for binary classification, KFold otherwise)
CrossValSplitter object: custom splitter (e.g., KFold, StratifiedKFold)
scoring
BuiltInScoring | ScoringFn
Scoring metric to use. Built-in options:
"accuracy" - Classification accuracy"f1" - F1 score"precision" - Precision score"recall" - Recall score"r2" - R² coefficient of determination"mean_squared_error" - Mean squared error"neg_mean_squared_error" - Negative MSE (higher is better) Can also be a custom function: (yTrue: Vector, yPred: Vector) => number Group labels for samples (used by group-aware splitters)
Sample weights to use during model fitting
Returns Array of scores, one for each cross-validation fold
Description Evaluate a model’s performance using cross-validation. The function:
Splits the data into k folds
For each fold:
Creates a fresh estimator instance
Trains on k-1 folds
Evaluates on the remaining fold
Returns an array of scores from each fold
This provides a robust estimate of model performance and helps detect overfitting.
Example import { crossValScore } from 'bun-scikit' ; import { LinearRegression , LogisticRegression } from 'bun-scikit' ; // Regression example const X = [ [ 1 ], [ 2 ], [ 3 ], [ 4 ], [ 5 ], [ 6 ], [ 7 ], [ 8 ], [ 9 ], [ 10 ] ]; const y = [ 2 , 4 , 6 , 8 , 10 , 12 , 14 , 16 , 18 , 20 ]; const scores = crossValScore ( () => new LinearRegression (), X , y , { cv: 5 , scoring: 'r2' } ); console . log ( 'R² scores:' , scores ); console . log ( 'Mean R²:' , scores . reduce (( a , b ) => a + b ) / scores . length ); console . log ( 'Std R²:' , Math . sqrt ( scores . reduce (( sum , score ) => { const mean = scores . reduce (( a , b ) => a + b ) / scores . length ; return sum + ( score - mean ) ** 2 ; }, 0 ) / scores . length ));
Classification Example import { crossValScore , StratifiedKFold } from 'bun-scikit' ; import { LogisticRegression } from 'bun-scikit' ; const X = [ [ 2.5 , 3.0 ], [ 3.5 , 4.0 ], [ 1.5 , 2.0 ], [ 3.0 , 3.5 ], [ 5.0 , 6.0 ], [ 6.0 , 7.0 ], [ 4.5 , 5.5 ], [ 5.5 , 6.5 ] ]; const y = [ 0 , 0 , 0 , 0 , 1 , 1 , 1 , 1 ]; // Use stratified k-fold for classification const stratifiedScores = crossValScore ( () => new LogisticRegression (), X , y , { cv: new StratifiedKFold ({ nSplits: 4 , shuffle: true }), scoring: 'accuracy' } ); console . log ( 'Accuracy scores:' , stratifiedScores );
Custom Scoring Function import { crossValScore } from 'bun-scikit' ; import { LinearRegression } from 'bun-scikit' ; const X = [[ 1 ], [ 2 ], [ 3 ], [ 4 ], [ 5 ]]; const y = [ 1.1 , 2.0 , 2.9 , 4.2 , 5.1 ]; // Custom scoring: Mean Absolute Error (negated for consistency) const customScoring = ( yTrue : number [], yPred : number []) : number => { let sum = 0 ; for ( let i = 0 ; i < yTrue . length ; i ++ ) { sum += Math . abs ( yTrue [ i ] - yPred [ i ]); } return - ( sum / yTrue . length ); // Negate so higher is better }; const scores = crossValScore ( () => new LinearRegression (), X , y , { cv: 3 , scoring: customScoring } ); console . log ( 'Negative MAE scores:' , scores );
With Sample Weights import { crossValScore } from 'bun-scikit' ; import { LinearRegression } from 'bun-scikit' ; const X = [[ 1 ], [ 2 ], [ 3 ], [ 4 ], [ 5 ]]; const y = [ 1 , 2 , 3 , 4 , 5 ]; const sampleWeight = [ 1 , 1 , 2 , 2 , 1 ]; // Give more weight to middle samples const scores = crossValScore ( () => new LinearRegression (), X , y , { cv: 3 , sampleWeight } ); console . log ( 'Weighted scores:' , scores );
Notes
The createEstimator parameter must be a factory function that returns a new instance each time it’s called
Do not pass a single estimator instance - each fold needs its own independent model
If scoring is not provided, the estimator’s score() method is used
For classification with binary targets, StratifiedKFold is used by default
Scores are returned in the order of folds (not sorted)
Use crossValPredict if you need predictions rather than scores