I want to determine, given a project, "How long will it take for this project to be successful ?"

Therefore, survival analysis seems like a perfect fit in this case (as I do have some projects that ended up as censored data as they took too long to be successful).

However, I also want to build my model so that, if a project seems like it will take too long (for instance it will be more than 2 months - the duration of the study) it will tell me not to accept it.

To sum it up ideal model for me would be :

  • Input : a project (with features like Founding, Past success, Location, etc.)
  • The models tells me if I should accept the project or not
  • If I should accept it, how long will it take before being successful ?

Is it doable with only survival analysis or should I consider adding a classification algorithm before that ?

  • 1
    $\begingroup$ The answer may depend on what decisions you are making with the model. If it's i) always a simple yes/no action, and ii) you apriori know the threshold (you allude to 2 months), then classification makes sense. If either of these isn't true, then survival analysis would make more sense. $\endgroup$ Jul 31 '19 at 19:00
  • 1
    $\begingroup$ Also, if you choose to go classification, and your threshold is N months, then I think using any "starts" from the past N months would cause bias (as you'll only see successes in the past N months - know what I mean). This will limit the amount / recency of your dataset $\endgroup$ Jul 31 '19 at 19:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.