1
$\begingroup$

I have 5 years of historical data. I need to build an model forecasting attrition for the next 3-5 years. I want to build a survival analysis model, but need to know if I can estimate the attrition (if they left and time) for each observation in my data. Will the output of the Cox Survival Model allow me to do that?

Ideally, I would like to append two columns to the data set for each unique id. 1. Did they attrit? 2. When?

$\endgroup$
4
  • $\begingroup$ Have you taken a look at lifelines here? This should have some good information for you. $\endgroup$ Jan 15 at 15:02
  • $\begingroup$ Thank Demetri, I can't find an answer to my specific question. $\endgroup$
    – Andy
    Jan 15 at 16:38
  • $\begingroup$ Do you think we need an exponential model because it provides a y-intercept? That way we can manually calculate the likelihood of attrition for each observation. I think that would mean the Cox model wouldn't work, but an exponential regression would work. $\endgroup$
    – Andy
    Jan 15 at 16:38
  • 1
    $\begingroup$ no a cox regression should be fine if you want to model attrition. I suggest you familiarize yourself a little more with survival analysis. Its the best tool for the job $\endgroup$ Jan 15 at 17:47
0
$\begingroup$

I think this chapter could be quite helpful to you: http://peopleanalytics-regression-book.org/survival.html. It focuses specifically on HR-related analyses, which should translate to what you are trying to do here.

One clarification that may be helpful: your input data should already tell you which employees left, and when. The goal of survival analysis in this context is to take that data and create a probability model that can be used to answer questions such as: "given the pattern of past employee behaviour, what is the probability that a new employee will still be with us in 1 year? Or 2 years, or 3 years, etc.

Another thing I would note: if you have no predictor variables, you could just use a Kaplan-Meier curve, instead of the Cox model. There's slightly less overhead in learning how that works.

$\endgroup$

Your Answer

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

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