DM: constructing new variables from raw data

We are starting a KDD project trying to predict fraud in telecommunications
area from raw call data. It seems that we ll have to construct new variables
from our data warehouse like:
-- Average duration of calls per day
-- Maximum number of international calls per month etc
Our approach of this subject is rather empirical for the moment. Could
anybody help us? Any suggestions, related work or readings are welcome!
Thanks a lot in advance
         Yannis

• Prev by Date: DM: Re: Datamining in Internet...
• Next by Date: DM: Re: predict fraud in telecommunications
• Prev by thread: DM: Re: predict fraud in telecommunications
• Next by thread: DM: Datamining in Internet...
• Index(es):
  • Date
  • Thread