1
$\begingroup$

I have a data mining assignment where I make a content-based image retrieval system. I have 20 images of 5 animals. So in total 100 images.

My code returns the 10 most relevant images to an input image.

Now I need to evaluate the performance of my system with a Precision-Recall curve. However, I do not understand the concept of a Precision-Recall curve.

Let's say my system returns 10 most relevant images of a gorilla, but only 4 of them are gorillas. The other 6 images returned are other animals'.

Now my precision is 4/10 = 0.4 and my recall is 4/20 = 0.2

So I only have a point <0.2,0.4> not a curve.

How do I have a curve?

By changing the number of images returned (which is fixed = 10 in my case) ?

Any help would be appreciated,

Thanks !

$\endgroup$

closed as off-topic by Geoff Oxberry Apr 18 '14 at 21:32

  • This question does not appear to be about computational science within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ Please don't post identical questions on multiple StackExchange sites. This is a dupe of stats.stackexchange.com/questions/94189/… $\endgroup$ – Marc Claesen Apr 18 '14 at 21:06
  • $\begingroup$ This question is not really off-topic. It is a cross-post, and cross-posting is discouraged. $\endgroup$ – Geoff Oxberry Apr 18 '14 at 21:32
  • $\begingroup$ @GeoffOxberry if the question is not really off-topic, why did you close it with reason "off-topic"? $\endgroup$ – Federico Poloni Apr 20 '14 at 17:06
  • $\begingroup$ @FedericoPoloni: We close cross-posted questions as a matter of policy. "Off-topic" is one of the more flexible reasons; it allows a person to enter a custom reason for closure, which is converted to a comment. The "duplicate" reason for closing will not work because it requires that the duplicate question also be on SciComp; for cross-posts, that condition is never satisfied by definition. $\endgroup$ – Geoff Oxberry Apr 20 '14 at 20:30
  • $\begingroup$ @GeoffOxberry I see, thanks for the explanation. So it seems to be an interface issue in Stack Exchange. $\endgroup$ – Federico Poloni Apr 20 '14 at 21:28
1
$\begingroup$

Please check this link: http://en.wikipedia.org/wiki/Precision_and_recall

Check the ROC section as well. This is a very clear description I think. For a more thorough understanding, I would recommend:

http://machinelearning.wustl.edu/mlpapers/paper_files/icml2006_DavisG06.pdf

$\endgroup$
  • $\begingroup$ Thanks! The wikipedia article didn't help. I will try to read the paper. $\endgroup$ – jeff Apr 18 '14 at 18:15

Not the answer you're looking for? Browse other questions tagged or ask your own question.