The Data Mining Forum 
IMPORTANT: This is the old Data Mining forum.
I keep it online so that you can read the old messages.
Please post your new messages in the new forum: https://forum2.philippe-fournier-viger.com/index.php
Data mining problem
Posted by:
J.J.
Date: September 13, 2014 12:30PM
Hello. I recently entered the world of corporate data mining, and I'm a bit stumped with a request from my boss. I'm pretty well versed on a variety of statistical methods, but for the life of me, I can't think of a solution. Here's my problem:
We have sales data for various products our client sells. We have the date that each product was purchased (entered as a separate row). My boss is asking me to do something with this sales data.
First thing I need to do as flatten the dataset so that each row is a unique purchaser. Then, I'm thinking each product should become a column, and every person that purchases the product would have the date the purchase was made in that respective column. Obviously, if a person did not make that purchase, then that column would be blank (or null).
Anyone have any thoughts on a procedure I can run? I've run simple binary logistic regressions and decision trees based only on product ownership (either owned or un-owned) to see what products predict ownership of other products. But my boss wants the "time" a product was purchased element to see if product ownership at time A predicts product ownership at time B. I can also append the data with age, gender, and some other very basic demographics (by going through a data provider service).
Any help anyone can provide would be greatly appreciated – what procedure can I do with this data (I have SPSS with several module add-ons). And if something isn't clear, I'd be happy to further explain myself. Thanks!
J.J.
We have sales data for various products our client sells. We have the date that each product was purchased (entered as a separate row). My boss is asking me to do something with this sales data.
First thing I need to do as flatten the dataset so that each row is a unique purchaser. Then, I'm thinking each product should become a column, and every person that purchases the product would have the date the purchase was made in that respective column. Obviously, if a person did not make that purchase, then that column would be blank (or null).
Anyone have any thoughts on a procedure I can run? I've run simple binary logistic regressions and decision trees based only on product ownership (either owned or un-owned) to see what products predict ownership of other products. But my boss wants the "time" a product was purchased element to see if product ownership at time A predicts product ownership at time B. I can also append the data with age, gender, and some other very basic demographics (by going through a data provider service).
Any help anyone can provide would be greatly appreciated – what procedure can I do with this data (I have SPSS with several module add-ons). And if something isn't clear, I'd be happy to further explain myself. Thanks!
J.J.
Re: Data mining problem
Posted by:
vidya
Date: September 18, 2014 09:25PM
Data mining research topics
Re: Data mining problem
Posted by:
jason best
Date: October 05, 2014 01:32PM
Hi there,
Can you clarify what the purpose of the time element is? Is it that you want to know whether AFTER product A has been purchased then product B is purchased because that's a better question to ask then whether the two products are owned together like your original idea.
You can absolutely make that into a binary regression as well labeling the 1 as if B is purchased after A is purchased.
Overall I think the question can be better answered if you have a better outline of your goal and perhaps some mocked up data.
Can you clarify what the purpose of the time element is? Is it that you want to know whether AFTER product A has been purchased then product B is purchased because that's a better question to ask then whether the two products are owned together like your original idea.
You can absolutely make that into a binary regression as well labeling the 1 as if B is purchased after A is purchased.
Overall I think the question can be better answered if you have a better outline of your goal and perhaps some mocked up data.
Re: Data mining problem
Posted by:
priyanka
Date: December 11, 2014 06:09AM
I have to implement pincer algorithm..plz provide me source code in C/C++...
Re: Data mining problem
Posted by:
Pooja jardosh
Date: December 18, 2014 03:05AM
I want database(dataset)file for importing in DB2,purpose is to perform DataDirectQuery on DB2 and extracting Data from XMLdocuemnts stored in db.
And then performing DataMining algorithm.
Is there any1 who can suggest or provide??
Waiting for favorable reply
Thanking You,
Pooja
And then performing DataMining algorithm.
Is there any1 who can suggest or provide??
Waiting for favorable reply
Thanking You,
Pooja