BookmarkSubscribeRSS Feed
Wickywick
Calcite | Level 5

Hi All,

I am new using enterprise miner, currently working on getting a desicion tree based on a Good/Bad indicators, the posibility of an account going good/bad based on various characteristics.

I have my baseline dataset where the good/bad indicators have been defined, my next step is to get some historical data.

My question is, do I get the historical data for the same accounts in my base line dataset and obtain the same characteristics that I'm analysing on?

1 REPLY 1
DougWielenga
SAS Employee

When you are predicting future behavior, you typically obtain data from the more distant past (e.g. say January to May of a certain year) and then observe behavior during an intermediate time (e.g. say July/August of that year) so that the model you fit can predict future behavior for new observations available in September or later that year for which the outcome is not known yet.   Your training data must have the outcome variable you are trying to model so that requires the data to be old enough so that you already know the outcome.  You can then collect other information that might be helpful in predicting that outcome to build your training data.   Depending on the availability of data, the length of the window you are trying to predict into (e.g. one month ahead? two months ahead? etc...), and your business objectives, you might need to alter your time windows and which questions can be asked.   

 

I hope this helps!

Doug

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

How to choose a machine learning algorithm

Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 1 reply
  • 698 views
  • 0 likes
  • 2 in conversation