Data Mining Cubes?

Post Reply
skan
Posts: 3
Joined: Mon Feb 23, 2015 9:40 pm
OLAP Product: SSAS
Version: 2012
Excel Version: 2013

Data Mining Cubes?

Post by skan » Mon Feb 23, 2015 10:51 pm

Hello.

I'm trying to study SQL Server Analysis Services and watching video tutorials on it.

Some videos explain how to create cubes and how useful they are.
Some other videos explain how to perform a data mining analysis.

But I don't understand why it seems that they don't create a cube as an earlier step for the data mining.
Why don't they use cubes as source for data mining? (and they just use views)

Regards

Admin note: Moved to correct forum.

Marcus Scherer
Community Contributor
Posts: 125
Joined: Sun Jun 29, 2008 9:33 am
OLAP Product: TM1
Version: 9.5 10.1 10.2
Excel Version: 2007
Location: Karlsruhe

Re: Data Mining Cubes?

Post by Marcus Scherer » Wed Feb 25, 2015 2:21 pm

Hello,
data mining is a wide field and not limited to OLAP cubes. They are probably using views on relational tables in a DWH or mining even directly in operational raw data. Difficult to say without knowing the tutorials. There are many approaches to data analysis depending on the requirement and available data structures. Using a view as a source might be a quick and flexible way compared to installing and eventually precalculating a cube.

regards,

skan
Posts: 3
Joined: Mon Feb 23, 2015 9:40 pm
OLAP Product: SSAS
Version: 2012
Excel Version: 2013

Re: Data Mining Cubes?

Post by skan » Wed Feb 25, 2015 2:29 pm

But, Can you use a cube as source for the data mining?
Or would it offer any advantage or disadvantage?


Any video tutorial on Youtube, they always do the same, for example
https://www.youtube.com/watch?v=EtFQv_B7YA8

Marcus Scherer
Community Contributor
Posts: 125
Joined: Sun Jun 29, 2008 9:33 am
OLAP Product: TM1
Version: 9.5 10.1 10.2
Excel Version: 2007
Location: Karlsruhe

Re: Data Mining Cubes?

Post by Marcus Scherer » Wed Feb 25, 2015 5:41 pm

In this example they are using the decision tree method which can be done with "flat" data records in tables or files. The classic market basket example uses flat tables as well. You don't need a cube here since you don't want to aggregate data or create multidimensional views. Since almost all the raw data of a company are captured in relational tables the preparation of the source for mining is easy then.
The advantage of cubes is to be able to "slice and dice" your cube views and drill down/up within dimensions. It is a different method of analysis but you may call it data mining as well. There is software on the market which can do more sophisticated data mining on existing cubes.

regards,

skan
Posts: 3
Joined: Mon Feb 23, 2015 9:40 pm
OLAP Product: SSAS
Version: 2012
Excel Version: 2013

Re: Data Mining Cubes?

Post by skan » Wed Feb 25, 2015 6:17 pm

OK, I don't need it,
but would it be possible to do data mining with cubes?

Post Reply