Model Evaluation Metrics

1/**2 * Example 24: Model Evaluation Metrics3 *4 * Learn to evaluate models using various performance metrics.5 * Different metrics for classification, regression, and clustering.6 */78import {9  // Classification metrics10  accuracy,11  confusionMatrix,12  f1Score,13  mae,14  mape,15  mse,16  precision,17  // Regression metrics18  r2Score,19  recall,20  rmse,21  // Clustering metrics22  silhouetteScore,23} from "deepbox/metrics";24import { tensor } from "deepbox/ndarray";2526console.log("=== Model Evaluation Metrics ===\n");2728// Classification Metrics29console.log("1. Classification Metrics:");30console.log("-".repeat(50));3132const y_true_class = tensor([1, 0, 1, 1, 0, 1, 0, 0, 1, 1]);33const y_pred_class = tensor([1, 0, 1, 0, 0, 1, 0, 1, 1, 1]);3435console.log("True labels:", y_true_class.toString());36console.log("Predictions:", `${y_pred_class.toString()}\n`);3738const acc = accuracy(y_true_class, y_pred_class);39const prec = precision(y_true_class, y_pred_class);40const rec = recall(y_true_class, y_pred_class);41const f1 = f1Score(y_true_class, y_pred_class);4243console.log(`Accuracy:  ${(Number(acc) * 100).toFixed(2)}%`);44console.log(`Precision: ${(Number(prec) * 100).toFixed(2)}%`);45console.log(`Recall:    ${(Number(rec) * 100).toFixed(2)}%`);46console.log(`F1-Score:  ${(Number(f1) * 100).toFixed(2)}%\n`);4748const cm = confusionMatrix(y_true_class, y_pred_class);49console.log("Confusion Matrix:");50console.log(cm.toString());51console.log("Format: [[TN, FP], [FN, TP]]\n");5253// Regression Metrics54console.log("2. Regression Metrics:");55console.log("-".repeat(50));5657const y_true_reg = tensor([3.0, -0.5, 2.0, 7.0, 4.2]);58const y_pred_reg = tensor([2.5, 0.0, 2.1, 7.8, 4.0]);5960console.log("True values:", y_true_reg.toString());61console.log("Predictions:", `${y_pred_reg.toString()}\n`);6263const r2 = r2Score(y_true_reg, y_pred_reg);64const mseVal = mse(y_true_reg, y_pred_reg);65const rmseVal = rmse(y_true_reg, y_pred_reg);66const maeVal = mae(y_true_reg, y_pred_reg);67const mapeVal = mape(y_true_reg, y_pred_reg);6869console.log(`R² Score: ${r2.toFixed(4)}`);70console.log(`MSE:      ${mseVal.toFixed(4)}`);71console.log(`RMSE:     ${rmseVal.toFixed(4)}`);72console.log(`MAE:      ${maeVal.toFixed(4)}`);73console.log(`MAPE:     ${(mapeVal * 100).toFixed(2)}%\n`);7475// Clustering Metrics76console.log("3. Clustering Metrics:");77console.log("-".repeat(50));7879const X_cluster = tensor([80  [1, 2],81  [1.5, 1.8],82  [5, 8],83  [8, 8],84  [1, 0.6],85  [9, 11],86]);87const labels = tensor([0, 0, 1, 1, 0, 1]);8889const silhouette = silhouetteScore(X_cluster, labels);90console.log(`Silhouette Score: ${silhouette.toFixed(4)}`);91console.log("Range: [-1, 1], higher is better");92console.log("Measures how similar points are to their own cluster\n");9394console.log("Metric Selection Guide:");95console.log("• Classification: Use F1-score for imbalanced data");96console.log("• Regression: R² for variance explained, MAE for interpretability");97console.log("• Clustering: Silhouette for cluster quality");9899console.log("\n✓ Metrics evaluation complete!");100