PDA

View Full Version : Need Advice



a.espil
10-12-2008, 08:12 AM
Hi All
I am working in developing BI system to ISP company.
I have just finished developing the first level of my BI system which include (reporting, analyzing and dashboards).. :)
now I am trying to add data mining and advanced analytics...
I know a lot about data mining and it's branches , BUT before I start..
I am wondering IF I have to extract more complex information about the objects I am working with, like build a profile for each sales point, area and region that allow me to perform a clustering and prediction algorithms on , OR I have to let the mining algorithems work directly on the core data I am using ....?

another question.. don't we have to use expert system in that field in order to perform the extracted rules ? :confused:

Mark
10-12-2008, 11:15 AM
I know a lot about data mining and it's branches , BUT before I start..
I am wondering IF I have to extract more complex information about the objects I am working with, like build a profile for each sales point, area and region that allow me to perform a clustering and prediction algorithms on , OR I have to let the mining algorithems work directly on the core data I am using ....?


I't is hard to say without knowing more about your data. In general, it is good practice to use your background knowledge of the domain and extract/derive as much information about each data point as you can. This will give the data mining algorithm more to work with.



another question.. don't we have to use expert system in that field in order to perform the extracted rules ? :confused:

It depends what the goal is. In an interactive diagnosis scenario, an expert system might be the way to go. In standard predictive analytics, you would feed a new data point into the rule set and obtain a prediction for the target field (whatever that might be for you problem).

Cheers,
Mark.