time window in classification model

bridge · Joined: 05 Apr 2006 Posts: 1

time window should be defined in a classification model such as churn model. could the data window overlap with the forecasting window? on the other words, could a classification problem be defined as:

to predict the probability about the total consumed amount of 3 month (such as May,June,July) will decrease to 200$ while the last 3 month's total amount (such as April,May,June) is more than 300$. on the assumption that it do have a good baseline rate in the training set,such as 3%.

thanks.

gabrielac · Contributor Joined: 16 Feb 2006 Posts: 8

It seems to me that if you already know the total amount for April and May, it comes down to what the amount for June will be. Maybe, you could define the problem with June as a target, based not only on the previous 3 month period, but also, with the 2 months already known of the present 3 month period.

Good luck.