3
$\begingroup$

In a employee attrition analysis with a table having rowwise data for a (employee like Id, name, Date_Join Date_Relieving Dept Role etc)

eID eName    Joining Releiving Dept Married Experience  
123 John Doe 10Oct15 12Oct16   HR   No      12  
234 Jen Doee 01jan16 -NA-      HR   No      11         (ie she is available)  

I can run regression on this data to find the beta coefficients

eID eName    Joining Releiving Dept Married Experience  
123 John Doe 10Oct15 12Oct16   HR   No      12  
234 Jen Doee 01jan16 -NA-      HR   No      11  

But I've seen other approach too.. where employee have multiple entries depending on their difference between joining date and current month or relieving month(say Employee A joined in Jan and Left in Dec so he'll have 12 entries updating corresponding columns like experience and marriage etc)

eID eName    Dept Married Experience  
123 John Doe HR   No      0  
123 John Doe HR   No      1  
123 John Doe HR   Yes     2  
123 John Doe HR   Yes     3  

can someone tell what differentiate two approaches.. and what are the benefits of this second approach.

$\endgroup$
2
  • $\begingroup$ IMO, First approach seems like used for predicting a new employee's experience or releiving date, given other info. Or even classification whether the new employee is married or not, given other data. But the second approach seems like predicting the current employee's future data whether he will remain married or will his experience improve. $\endgroup$
    – chmodsss
    Nov 9 '16 at 9:04
  • $\begingroup$ Thanks, @null_9 I'm trying to understand for same output, to check whether employee will stay or leave the company, $\endgroup$
    – Makarand
    Nov 9 '16 at 18:35
1
$\begingroup$

Creating model on second scenario is difficult but more information driven and give you more prediction capibility.

What you can do is use second scenario and create some featured variable. These variable will give information like after marrying in how many months attrition happen and so on.

So think what other info you can take out of this trend.

First scenario is not of much use, because it's just giving you last information of employee which hides many important relevant information which are actually driving attrition.

$\endgroup$
0
$\begingroup$

Both the data could be used for a better model. As your aim is to classify whether the employee will quit or continue, use each person's monthly entries and construct them(marital status, experience, daily working hours, age, salary) as features with the output quit/continue. Train a classifier with the available data and use it to predict new data.

$\endgroup$

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.