Logistic Regression

import { loadIris } from "deepbox/datasets";
import { tensor } from "deepbox/ndarray";
import { trainTestSplit } from "deepbox/preprocess";

// Load the Iris dataset
const iris = loadIris();
console.log(`Dataset: ${iris.data.shape[0]} samples, ${iris.data.shape[1]} features`);

// Convert to binary classification: setosa (0) vs non-setosa (1)
const y_binary: number[] = [];
for (let i = 0; i < iris.target.size; i++) {
  const label = Number(iris.target.data[iris.target.offset + i]);
  y_binary.push(label === 0 ? 0 : 1);
}
const y = tensor(y_binary);

Dataset: 150 samples, 4 features

import { StandardScaler } from "deepbox/preprocess";

const [X_train, X_test, y_train, y_test] = trainTestSplit(iris.data, y, {
  testSize: 0.3,
  randomState: 42,
});

console.log(`Training set: ${X_train.shape[0]} samples`);
console.log(`Test set: ${X_test.shape[0]} samples`);

// Standardize features (mean=0, std=1)
const scaler = new StandardScaler();
scaler.fit(X_train);
const X_train_scaled = scaler.transform(X_train);
const X_test_scaled = scaler.transform(X_test);

console.log("Features scaled");

Training set: 105 samples
Test set: 45 samples
Features scaled

import { LogisticRegression } from "deepbox/ml";

const model = new LogisticRegression({ 
  maxIter: 1000, 
  learningRate: 0.1 
});
model.fit(X_train_scaled, y_train);

console.log("Model trained!");

Model trained!

import { accuracy, precision, recall, f1Score, confusionMatrix } from "deepbox/metrics";

const y_pred = model.predict(X_test_scaled);

const acc = accuracy(y_test, y_pred);
const prec = precision(y_test, y_pred);
const rec = recall(y_test, y_pred);
const f1 = f1Score(y_test, y_pred);

console.log("Model Performance:");
console.log(`Accuracy:  ${(Number(acc) * 100).toFixed(2)}%`);
console.log(`Precision: ${(Number(prec) * 100).toFixed(2)}%`);
console.log(`Recall:    ${(Number(rec) * 100).toFixed(2)}%`);
console.log(`F1-Score:  ${(Number(f1) * 100).toFixed(2)}%`);

const cm = confusionMatrix(y_test, y_pred);
console.log("\nConfusion Matrix:");
console.log(cm.toString());

Model Performance:
Accuracy:  100.00%
Precision: 100.00%
Recall:    100.00%
F1-Score:  100.00%

Confusion Matrix:
Tensor([[15,  0],
        [ 0, 30]])