[Machine Learning]---ROC(Receiver Operating Characteristic) Curve

1. Some Basic Concepts

• True Positive Rate:
  $$TPR=\frac{TP}{TP+FN}$$
• False Positive Rate
  $$FPR=\frac{FP}{FP+TN}$$

2. ROC Curve

ROC use the TPR as the y axis and FPR as the x axis, so we hope that y is as large as possible and x is as small as possible.
In a picture we hope that the ROC cure in the high and left.
So, how to choose the threshold $\theta$.
1. if $h(x) > \theta$, we classify the $x$ is Positive as 1.
2. If $h(x) < \theat$, we classify the x as $Negative$ as 0.
Let’s plot the picture.

3. Reference

http://blog.csdn.net/abcjennifer/article/details/7359370