From the test results of the target detection, always with false positve, precision rate, the recall rate of the several nouns. Always in the use of the time to clear the meaning of these words, but a long time, and to forget. A few days ago to discuss this issue with my brother, still vaguely, ambiguous. Decide to summarize it and write it down for a rainy future.
The Chinese meanings of these three words are: false detection rate (also known as false alarm rate), precision ratio and recall. How they are calculated can be seen in the following diagram:
Figuring out the meaning of TP, FP, FN, and TN is basically even figuring it out. False positve is also widely used in medical testing. We'll use the tumor to do the example:
A patient, had a tumor, went to the hospital to check, respectively, the following conditions:
1, the real situation: malignant, examination results: Malignant, this situation is called: true positve, let's call true positive. (Unlucky enough)
2, the real situation: benign, examination results: Malignant, this situation is called: false positve, false positive. (No disease, no illness, no more reasoning)
3, the real situation: malignant, examination results: Benign, this situation is called: false negative, false negative. (Quack delay illness AH)
4, the real situation: benign, examination results: Benign, this situation is called: true negtive, true negative. (This guy has a good personality.)
After these nouns are understood, it is possible to know how the three "rate" is calculated:
False detection rate: FP = SUM (FP)/(SUM (FP) + sum (TN))
Precision ratio: Precision rate = SUM (TP)/(SUM (TP) + SUM (FP))
Recall: Recall rate = SUM (TP)/(SUM (TP) + SUM (FN))
There is also a common missing rate: Miss = SUM (FN)/(SUM (TP) + SUM (FN))
Recall rate + Miss Rate = 1
The rate of false detection is relative to the total number of false targets is mistaken for the real target, the precision is the detected target, the proportion of the real target, recall is the real target detected, in all the real target proportion.
Oh, I do not know you understand that no, anyway, I always look more dizzy. Good luck.