All Examples
Example 24
intermediate
24
Metrics
Evaluation
Classification
Regression

Model Evaluation Metrics

Choosing the right evaluation metric is as important as choosing the right model. This example covers three categories: Classification metrics (accuracy, precision, recall, F1 score, confusion matrix, ROC AUC, classification report, log loss, Matthews correlation, Cohen's kappa, balanced accuracy, Jaccard score), Regression metrics (MSE, RMSE, MAE, MAPE, R², adjusted R², explained variance, max error, median absolute error), and Clustering metrics (silhouette score, Calinski-Harabasz index, Davies-Bouldin index, adjusted Rand index, normalized mutual info, completeness, homogeneity, V-measure, Fowlkes-Mallows). Each metric is computed on sample data with an explanation of what it measures and when to use it.

Deepbox Modules Used

deepbox/ndarraydeepbox/metrics

What You Will Learn

  • Accuracy is misleading for imbalanced data — use F1 or balanced accuracy
  • Precision measures false positive rate; Recall measures false negative rate
  • MSE penalizes large errors quadratically; MAE treats all errors equally
  • R² close to 1.0 means the model explains most variance in the data
  • Silhouette score evaluates clustering quality without ground truth labels

Source Code

24-metrics/index.ts
1import { tensor } from "deepbox/ndarray";2import {3  accuracy, precision, recall, f1Score, confusionMatrix,4  mse, rmse, mae, r2Score,5  silhouetteScore6} from "deepbox/metrics";78// Classification metrics9const yTrue = tensor([0, 1, 1, 0, 1, 0, 1, 1]);10const yPred = tensor([0, 1, 0, 0, 1, 1, 1, 1]);1112console.log("=== Classification ===");13console.log("Accuracy: ", accuracy(yTrue, yPred).toFixed(3));14console.log("Precision:", precision(yTrue, yPred).toFixed(3));15console.log("Recall:   ", recall(yTrue, yPred).toFixed(3));16console.log("F1 Score: ", f1Score(yTrue, yPred).toFixed(3));17console.log("Confusion Matrix:");18console.log(confusionMatrix(yTrue, yPred).toString());1920// Regression metrics21const yTrueReg = tensor([3.0, -0.5, 2.0, 7.0]);22const yPredReg = tensor([2.5, 0.0, 2.1, 7.8]);2324console.log("\n=== Regression ===");25console.log("MSE: ", mse(yTrueReg, yPredReg).toFixed(4));26console.log("RMSE:", rmse(yTrueReg, yPredReg).toFixed(4));27console.log("MAE: ", mae(yTrueReg, yPredReg).toFixed(4));28console.log("R²:  ", r2Score(yTrueReg, yPredReg).toFixed(4));

Console Output

$ npx tsx 24-metrics/index.ts
=== Classification ===
Accuracy:  0.750
Precision: 0.800
Recall:    0.800
F1 Score:  0.800
Confusion Matrix:
[[2, 1],
 [1, 4]]

=== Regression ===
MSE:  0.2150
RMSE: 0.4637
MAE:  0.3500
R²:   0.9717
PreviousNext