Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Survival Analysis using EMiner: how to enter time based variables other than the target?

Accepted Solution Solved
Reply
New Contributor
Posts: 2
Accepted Solution

Survival Analysis using EMiner: how to enter time based variables other than the target?

In a telecommunication company, we have data on the customer lifecycle. The data has the time the customer joined and the time they left, and it has other events that happened to the customer during their lifecycle like daily data usage and support interactions (calling tech support). We would like to know the effect of each of these events on customer attrition.

I want to use EMiner survival analysis, it takes two time ids, one for activation and one for attrition, but how can I enter the other events and their times?

For example, the customer called support on three different dates, how can I enter these support interactions and their time. Another example is the customer daily data usage, how can I enter the data usage for each day. As I wish to know the impact of these events on attrition.

Appreciate any help.


Accepted Solutions
Solution
‎06-14-2017 12:18 AM
SAS Super FREQ
Posts: 269

Re: Survival Analysis using EMiner: how to enter time based variables other than the target?

In the Survival node in Enterprise Miner, you can use Change-Time or Fully Expanded for the Data Format property that I think will accommodate your data.  See this video for details: http://www.sas.com/apps/webnet/video-sharing.html?player=brightcove&width=640&height=360&autoStart=t...

 

And there are also details and examples in the EM Reference Help for the Survival node.

 

View solution in original post


All Replies
Super User
Posts: 17,749

Re: Survival Analysis using EMiner: how to enter time based variables other than the target?

I think accounting for the time in a standard survival model would be difficult. 

 

I would suggest creating new variables that indicate what you're after, ie # of service calls, daily usage total, Time between last service call and leaving. 

 

This sounds a bit too difficult for standard statistical models and I would consider simulation to look at the events and outcomes. 

Solution
‎06-14-2017 12:18 AM
SAS Super FREQ
Posts: 269

Re: Survival Analysis using EMiner: how to enter time based variables other than the target?

In the Survival node in Enterprise Miner, you can use Change-Time or Fully Expanded for the Data Format property that I think will accommodate your data.  See this video for details: http://www.sas.com/apps/webnet/video-sharing.html?player=brightcove&width=640&height=360&autoStart=t...

 

And there are also details and examples in the EM Reference Help for the Survival node.

 

New Contributor
Posts: 2

Re: Survival Analysis using EMiner: how to enter time based variables other than the target?

Thank you very much Wendy. This is what I was looking for. One thing though, I saw in your video how you are increasing number of complaints in fully expanded, but I have complaint type variable and complaint resolution quality as well for each complaint. How do I enter these two variables that are specific for each complaint? (as I would like to know which complaint type has more attrition probability, and if the complaint resolution quality affects attrition probability)
☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 3 replies
  • 225 views
  • 2 likes
  • 3 in conversation