-2

I presented a testing (specifically, symbolic execution) approach for special language. I have a problem in evaluating my approach. I am going to evaluate correctness and usefulness of the approach. My own perception is computing precision for the correctness. I created 100 mutations of one case study by injecting one error in each of them and calculating the fraction of (number of real errors/number of detecting errors). For calculating usefulness of the approach, I studied in the papers that the mutation also used for this purpose. By the way, I have no idea about computing the usefulness of the approach and the differences between those. Could you please help me in this regard?

  • Uhhhhhhh, what? – Eppicurt Nov 24 '17 at 10:08
  • @Eppicurt how can I compute the usefulness of my tool? – user148494 Nov 24 '17 at 10:17
  • @Eppicurt: My feelings exactly. Are we talking about something in biology, linguistics, computer science, or what? Probably only 2 or 3 extra words somewhere near the beginning would be all that's necessary to provide the present missing context for those readers of this general academic site who are not steeped in the OP's specific technical area. (moments later) I noticed the tag "software" just now, which I guess somewhat blunts my criticism, but I still think 2 or 3 extra words of context would have been nice. – Dave L Renfro Nov 24 '17 at 11:57
  • This post is cross-posted on CS.SE – scaaahu Nov 24 '17 at 13:49
  • @DaveLRenfro All the information should be in the question itself -- people tend not to read the tags and really they're just there for indexing. This question looks like a "content of my research" question, to me, but I don't understand it, either. – David Richerby Nov 24 '17 at 14:43
1

"Usefulness" is a subjective term, not something that you compute. You typically present some concrete results, then when discussing these results, make a case of why in your opinion the results show that your method is useful. And ultimately it's up to reviewers and readers to decide whether they agree.

You should think about what makes something "useful" in your field, and what concrete analyses you can do that provide supporting evidence.

  • Typically that involves measuring desirable performance traits. "Correctness" would be one, but there may be others: speed, efficiency, rate of certain types of errors, performance on certain types of problems, etc.
  • But you can't just measure your own approach; you have to measure others, too, for comparison. If your method is 98% correct, but someone else is 99% correct, then arguably your method isn't "useful" (unless it has some other advantages).

However, these analyses wouldn't be described as "usefulness" directly.

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