DM: RE:

[Cunningham, Scott W]

Perhaps your neural network over-fit the data.

Many modeling techniques - if allowed - come up with excellent but 
highly
specialized solutions.  The models then do not generalize to new data 
sets. 

The best way I know to prevent this in neural networks is to 
prematurely
stop the learning.  (Stop at 2000 iterations, when you know the 
solution
converges to minimum error at 4000).  There are formal methods of 
setting up
training, testing, and validation data sets to determine when to stop
fitting.

There are also other techniques (such as Bayesian approaches) to 
prevent
neural networks from over-fitting data.

Best Regards

____________
Scott Cunningham, D.Phil.
Research Engineer
NCR Human Interface Technology Center







        -----Original Message-----
        From:   Hallberg Rassy [SMTP:hr38@hotmail.com]
        Sent:   Thursday, April 09, 1998 6:23 AM
        To:     datamine-l@nautilus-sys.com
        Subject:        

        Dear Friend

        I have now a problem and would like to share with you and to
receive, if 
        possible your opinion.
        Hereinafter the picture of situation:
        1.I'm presently involved in a customer profiling project for 
a large

        mobile operator.
        2.The goal is to set up a system able to anticipate the 
likelyhood
of 
        churn of  customers
        3.As a pilot step I extracted call records for 10000 active
customers 
        plus 4000  churned
        4.Using SPSS neural connection I made up a neural network 
based on a
set 
        of 4000  active+4000 churned
        5.The data was: calling patterns of july, agoust and 
september the 
        target was:  churn/no churn situation in december
        6.The results was promising: 90% of real churn anticipated, 
with a 
        cut-off  probability of 80%
        7.The same network was used on october, nov, dec. data to 
anticipate

        march churn  the results dropped to a terrific 11% with the 
same
cut-off 
        of 80%: totally  useless


        I have formulated some hypotheses
        A.The low time span (three month) is affected by seasonality
        B.The data used are not sufficient to build a reliable network
        C.The tool (SPSS Neural Connection) is not reliable

        Could you give your opinion?
        Many thanks in advance



        ______________________________________________________
        Get Your Private, Free Email at http://www.hotmail.com