The Receiver operating charachteristics is a plot displaying the efficiency of a classification model as the threshold value is varied.

The ROC curve is the True positive rate(TPR) plotted against the False positive rate(FPR) at a specified threshold.

ROC Curve

Confusion Matrix

Error matrices

Score

• It is the estimated probability of true in a logistic regression

Baseline Score

• This is the score when a random guess predicts the classification. The value is mostly 0.5. This value is considered as the threshold and the model is expected to outperform the baseline score.

How to improve the ROC curve?

• The threshold can be varied based on the business requirement to improve the error matrices like benefits/cost.

Credits: Extracts from Wikipedia