Confusion matrix

NxN matrix, where N is the number of classes for classification.

• Accuracy: ratio of correct prediction vs total predictions
• Positive Predictive Value or Precision: ratio of correctly predicted positive case to total positive predictions
• Negative Predictive Value: ratio of correctly predicted negative case to total negative predictions
• Sensitivity or Recall (True Positive rate): ratio of correctly positive prediction to actual number of positives
• Specificity (false positive rate): ratio of correctly negative prediction to actual number of negatives

ROC AUC

The ROC curve is the plot between sensitivity and (1- specificity) Here different lines are for different model. Model W is better than X, model X is better than Y and so on.