Definition: ROC curve which stands for Receiver Operating Characteristic is a useful visual tool for comparing two classification models.
Rules for ROC curve:
1. In order to plot a Roc curve for a given classification model (M), the model most be able to return a probability or ranking for the predicted class of each test tuple.
2. The vertical axis of an Roc curve represents the true positive rate 'yes', where the horizontal axis represent the false positive rate 'no'.
3. The closer the ROC curves to the diagonal line, the less accurate the model and vice versa.
4. To asses the accuracy of the model, we can measure the area under the curve, the closer the area is to 0.5 the less accurate corresponding model is, a model with perfect accuracy will have an area of 1.0.
Example:
Assume we have two new medicines that were tested on different patients: A, B and C. given in the table the percentage of True positive and false positive.
Medicine 1
Test Value True positive(yes) False positive(no)
A 30% 2%
B 60% 20%
C 90% 40%
Medicine 2
Test Value True positive(yes) False positive(no)
A 40% 2%
B 68% 6%
C 88% 16%
Q: Which Medicine is better? Why?
