I am studying antiviruses how antiviruses work at my university. This is just the very basics.

He asked in our opinion which we think is better: recall or precision.

Why does my professor say he cares about recall and precision, but does not care about accuracy?

We were talking about false positives and false negatives so it's somehow related to this.

Thank you.

EDIT: To clarify, the professor said he does not care about accuracy, and then asked us why.

closed as off-topic by Steffen Ullrich, Xander, NULLZ, Overmind, kalina Apr 28 at 4:10

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "This question does not appear to be about Information security within the scope defined in the help center." – Steffen Ullrich, Xander, NULLZ, kalina
If this question can be reworded to fit the rules in the help center, please edit the question.

  • and which you think is better ? – Soufiane Tahiri Apr 9 at 12:36
  • @SoufianeTahiri I prefer precision for less false positives. But why do we not care about accuracy? – John Smith Apr 9 at 12:44
  • 1
    This seems like an excellent question to ask to your professor, along with your own thoughts. Recalling my time from university, I'm sure my professors would have enjoyed a student actually caring enough about the subject at hand to ask for reasoning. – MechMK1 Apr 9 at 12:48
  • @MechMK1 We cannot get in contact with him and this is a group study session. I think there might be a maths reason for this, but none of us studied A-Level maths... – John Smith Apr 9 at 12:49
  • 1
    "He asked in our opinion which we think is better: recall or precision." - just because he lets you compare the explanatory power of recall vs. precision does not say at all that he does not care about accuracy. That's only what you read into this for some unknown reason. He does not explicitly say that he cares about money either but you would likely not assume that he does not. So "does not care about accuracy" is only in your mind, not necessarily in reality - at least from what you describe in your question. – Steffen Ullrich Apr 9 at 13:35

This is because of the severe class imbalance problem in malware detection. Instances of the class of interest, programs humans refer to as “malware”, are far outnumbered by instances of “not malware”. Let’s say that out of the set of all programs 1% are malware and 99% are not. A trivial classifier can achieve 99% accuracy by simply assigning all programs it analyzes to the majority class (the “not malware” class). Clearly this approach is not well suited to the purpose of distinguishing between malware and not-malware. Therefore, in classification problems with severe class imbalance there are alternative metrics that can be used to asses model performance besides simple accuracy, particularly in cases where the minority class is the class of interest, as it is with malware. Precision and recall are examples of such alternative metrics.

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