VotingClassifier combines multiple classification algorithms using majority voting.
Constructor
import { VotingClassifier } from 'scikitjs';
const voting = new VotingClassifier(estimators, options);
estimators
[string, estimator | () => estimator][]
required
List of (name, estimator) tuples. Estimators can be instances or factory functions.
options.voting
'hard' | 'soft'
default:"'hard'"
Voting strategy:
'hard': uses predicted class labels for majority rule voting'soft': predicts class label based on sum of predicted probabilities (requires estimators with predictProba)
Sequence of weights for each estimator. If None, uses uniform weights.
Methods
fit
Fit all estimators.
fit(X: number[][], y: number[], sampleWeight?: number[]): VotingClassifier
predict
Predict class labels using majority voting.
predict(X: number[][]): number[]
predictProba
Predict class probabilities (soft voting only).
predictProba(X: number[][]): number[][]
score
Return accuracy score.
score(X: number[][], y: number[]): number
Properties
The fitted sub-estimators
namedEstimators_
Record<string, estimator>
Access estimators by name
Examples
Hard Voting
import {
VotingClassifier,
LogisticRegression,
DecisionTreeClassifier,
KNeighborsClassifier
} from 'scikitjs';
const X = [
[1, 2], [2, 3], [3, 4], [4, 5],
[5, 6], [6, 7], [7, 8], [8, 9]
];
const y = [0, 0, 0, 0, 1, 1, 1, 1];
// Create voting classifier with hard voting
const voting = new VotingClassifier(
[
['lr', new LogisticRegression()],
['dt', new DecisionTreeClassifier()],
['knn', new KNeighborsClassifier({ nNeighbors: 3 })]
],
{ voting: 'hard' }
);
voting.fit(X, y);
const predictions = voting.predict([[3.5, 4.5], [6.5, 7.5]]);
console.log(predictions); // [0, 1]
const accuracy = voting.score(X, y);
console.log(accuracy);
Soft Voting
import {
VotingClassifier,
LogisticRegression,
RandomForestClassifier,
GaussianNB
} from 'scikitjs';
const X = [
[1, 2], [2, 3], [3, 4], [4, 5],
[5, 6], [6, 7], [7, 8], [8, 9]
];
const y = [0, 0, 0, 0, 1, 1, 1, 1];
// Soft voting averages predicted probabilities
const voting = new VotingClassifier(
[
['lr', new LogisticRegression()],
['rf', new RandomForestClassifier({ nEstimators: 10 })],
['nb', new GaussianNB()]
],
{ voting: 'soft' }
);
voting.fit(X, y);
const probabilities = voting.predictProba([[3.5, 4.5], [6.5, 7.5]]);
console.log(probabilities);
// [[0.85, 0.15],
// [0.12, 0.88]]
Weighted Voting
import {
VotingClassifier,
LogisticRegression,
SVC,
DecisionTreeClassifier
} from 'scikitjs';
const X = [[1, 2], [2, 3], [3, 4], [4, 5]];
const y = [0, 0, 1, 1];
// Give more weight to certain estimators
const voting = new VotingClassifier(
[
['lr', new LogisticRegression()],
['svc', new SVC({ probability: true })],
['dt', new DecisionTreeClassifier()]
],
{
voting: 'soft',
weights: [2, 2, 1] // LR and SVC have double weight
}
);
voting.fit(X, y);
const predictions = voting.predict(X);
Using Factory Functions
import { VotingClassifier, RandomForestClassifier } from 'scikitjs';
// Factory functions create fresh instances for each fit
const voting = new VotingClassifier([
['rf1', () => new RandomForestClassifier({ nEstimators: 10, randomState: 42 })],
['rf2', () => new RandomForestClassifier({ nEstimators: 20, randomState: 43 })],
['rf3', () => new RandomForestClassifier({ nEstimators: 15, randomState: 44 })]
]);
voting.fit(X, y);
Cross-Validation
import {
VotingClassifier,
LogisticRegression,
SVC,
crossValScore
} from 'scikitjs';
const X = [
[1, 2], [2, 3], [3, 4], [4, 5],
[5, 6], [6, 7], [7, 8], [8, 9]
];
const y = [0, 0, 0, 0, 1, 1, 1, 1];
const voting = new VotingClassifier([
['lr', () => new LogisticRegression()],
['svc', () => new SVC()]
]);
const scores = crossValScore(voting, X, y, { cv: 3 });
console.log('CV Scores:', scores);
console.log('Mean:', scores.reduce((a, b) => a + b) / scores.length);
Compare with Individual Models
import {
VotingClassifier,
LogisticRegression,
DecisionTreeClassifier,
SVC
} from 'scikitjs';
const XTrain = [[1, 2], [2, 3], [3, 4], [4, 5]];
const yTrain = [0, 0, 1, 1];
const XTest = [[1.5, 2.5], [3.5, 4.5]];
const yTest = [0, 1];
// Individual models
const lr = new LogisticRegression();
lr.fit(XTrain, yTrain);
console.log('LR accuracy:', lr.score(XTest, yTest));
const dt = new DecisionTreeClassifier();
dt.fit(XTrain, yTrain);
console.log('DT accuracy:', dt.score(XTest, yTest));
const svc = new SVC();
svc.fit(XTrain, yTrain);
console.log('SVC accuracy:', svc.score(XTest, yTest));
// Voting ensemble
const voting = new VotingClassifier([
['lr', new LogisticRegression()],
['dt', new DecisionTreeClassifier()],
['svc', new SVC()]
]);
voting.fit(XTrain, yTrain);
console.log('Voting accuracy:', voting.score(XTest, yTest));
Access Individual Estimators
const voting = new VotingClassifier([
['lr', new LogisticRegression()],
['svc', new SVC()]
]);
voting.fit(X, y);
// Access by name
const lr = voting.namedEstimators_['lr'];
console.log('LR coefficients:', lr.coef_);
// Iterate over all estimators
for (const [name, estimator] of voting.estimators_) {
const score = estimator.score(XTest, yTest);
console.log(`${name}: ${score}`);
}