Classification Report

EBM — Adult Income — 9769 samples · 2 classes

0.930
0.9300
AUC
Area Under the ROC Curve
87.72%
Accuracy
Fraction of all predictions that are correct: (TP+TN) / Total
80.13%
Balanced Acc
Average per-class recall, correcting for class imbalance
0.7951
Precision
Of predicted positives, how many are correct: TP / (TP+FP)
0.6557
Recall
Of actual positives, how many are found: TP / (TP+FN). Also called sensitivity
0.7187
F1 Score
Harmonic mean of precision and recall: 2·P·R / (P+R)
0.6459
MCC
Matthews Correlation Coefficient: balanced measure even with imbalanced classes. Range [-1, 1]
0.6411
Cohen κ
Agreement beyond chance between predictions and true labels. Range [-1, 1]
0.9300
AUC
Area Under the ROC Curve: probability that a random positive ranks above a random negative
0.2735
Log Loss
Negative log-likelihood of predicted probabilities. Lower is better
0.0868
Brier Score
Mean squared error of predicted probabilities. Lower is better. Range [0, 1]
0.9468
Specificity
Of actual negatives, how many are correctly identified: TN / (TN+FP)
0.8973
NPV
Negative Predictive Value: of predicted negatives, how many are correct: TN / (TN+FN)
0.6025
Informedness
Recall + Specificity − 1. How much the model informs beyond chance. Also called Youden's J
0.6924
Markedness
Precision + NPV − 1. How marked the predictions are beyond chance

Confusion Matrix

ROC Curve

Precision-Recall Curve

Calibration Plot

Threshold Analysis

Per-Class Metrics

Cumulative Gains

Feature Importances (Top 14)

Class Distribution (Test Set)

Prediction Score Distribution

Per-Class Metrics

Class                 Precision     Recall         F1    Support
----------------------------------------------------------------
<=50K                    0.8973     0.9468     0.9214       7431
>50K                     0.7951     0.6557     0.7187       2338