Re: DM: Queries...

[Ken Collier]

Srikanth Jagannath T 211 R G-Floor WS blrsn47 writes:
> 
> Hello,
> 
>       I am really sorry for such a long mail, but I have some 
>queries.
> 
>       I have a few queries regarding the application of Data Mining 
>and
>       Data Warehousing. Yesterday I had posted to this group about 
>some
>       practical case studies, and Mr. Dorothy Firsching had 
>responded asking
>       me to have a look at http://www.nautilus-systems.com/. Yep,I 
>had a
>       look at this site, and got some information on applicaiton of 
>Data 
>       Mining to the marketing of medicines. Here again it is 
>mentioned that
>       the data was analysed, extracted and from this knowledge was 
>extracted.
> 
>       But my quesion is, what sort of data would be helpful for such
>       analysis. Usually, I guess, that such companies perform a 
>market
>       survey by going to the consumers with a questionaire, and they
>       perform analysis on this data obtained from the questionaire. 
>If again 
>       such data is gathered and analysis done using this, then what 
>is the need 
>       for Data Mining. In the case of Data Mining, I guess, that 
>this 
>       information is obtained from the data recorded about each 
>consumer 
>       transaction in the database. I have a typical marketing 
>scenario for 
>       which I am wondering as to what should be the data that has 
>       to be recorded. I would give you a brief of the scenario:
> 
>       I have a company providing some service to the customers. The 
>company
>       stores data regarding the customer, the service being 
>provided,
>       customer transactions, billing details, payment details, 
>etc.,. Now my
>       aim is to find the customer base which would respond if the 
>company
>       starts a new service, I would also like to know as to what 
>would be
>       the response for the new service, what new service to start 
>and things
>       like this. For this what special data must be stored in the 
>database.
> 
>       Moreover I have been having some conflicting thoughts about 
>Data
>       Warehousing. As far as I perceived Data Warehousing is a data 
>store     
>       that stores historic data, and data analysis can be done on 
>this
>       data store. But when I explained this to a lay man, he was 
>asking me
>       as to if this is just the use of Data Warehousing then why do 
>you need
>       one, or why is it being called Data Warehouse, and not just 
>any other
>       database, say Historic Database. I was just wondering on this 
>point.
> 
>       Actually all this is happening because I have not had any 
>formal 
>       discussion with any experts/researchers in this area. The 
>knowledge
>       that I have gained is by reading research papers. So it would 
>be
>       great if someone gives me an insight into this questions, and 
>also
>       some suggestions on some book/paper that is more explanatory 
>in these
>       aspects.
> 
>       Thanks in advance,
>       regards
>       -srikanth

There appear to be two questions here:

Q1: How can my company use data mining to target/predict existing
    customers who will buy a new service?
Q2: What is data warehousing  and  how  does  it  differ  from  a
    traditional database?

Allow me to pose an answer to Q2 first:  A  data  warehouse  (DW)
forms  the core of an effective decision support system (DSS). It
should be designed to contain all of the  salient  business  data
(from all of your business units) necessary to enable your people
to make informed decisions.  Suppose  your  marketing  department
collects   data  relevant  to  sales,  the  financial  department
collects  data  relevent  to  fiscal  issues,  and  your  quality
assurance  people  collect  process  control  data. A DW helps to
consolidate the data from each of those units into a cohesive and
usable  architecture  so that business decisions are based on the
"big picture" rather than only a slice of the data.  Moreover,  a
DW  provides  meta-data,  multidimensional views of the data, and
the  ability  to  drill  down  and  roll  up  to  gain  aggregate
perspectives  on  your  data.  There's  more  of course, but that
should give a thumbnail sketch of DW. Once you have a DW, you can
then  provide  information back to each business unit in the form
of data marts the make the data more manageable to each unit.

Now, Q1 is a little trickier. It makes perfect sense to use  data
mining  technologies  to  try and understand which customers will
buy a  new  service.  However,  to  form  this  predictive  model
requires   relevant   historical   data.  If  your  business  has
introduced a new service in the past and has  collected  data  on
the customers (both former and new customers) who bought this new
service, then you could use that data to get  a  profile  of  the
customer  who  buys a new service. If you don't have this sort of
historical data then you  might  be  able  to  buy  a  commercial
dataset  that  contains  the  kind  of  information  necessary to
generate a reasonably decent predictive model.  While  I  am  not
aware  of any such specific datasets, there are several companies
that exist to sell business data for this sort of purpose. In any
case  you  will  need  some  relevant historical data in order to
generate the predictive model you seek. This is true whether  you
do  the analysis by hand using statistical methods or by computer
using data mining technology.

I've tried to be relatively brief. Hope it has been helpful.

---
Ken Collier, Ph.D.
Project Leader, Center for Data Insight
http://www.cse.nau.edu/~insight

Box 15600                               e-mail: Ken.Collier@nau.edu
College of Engineering and Technology   Phone: 520-523-5412 
Northern Arizona University             Fax: 520-523-2300
Flagstaff, Arizona  86011-1560          WWW: 
http://www.cse.nau.edu/~kwc