zondag 13 december 2009

Datamining with SQL Server

Today, i'm gonna tell you something about datamining with SQL Server. Datamining is part of Analysis Services. Data mining helps you make decisions about difficult business problems, It's possible to identify patterns in your data, so you can determine why things happen. It's even psosible to predict what will hapen in the future. You can create data mining models based on cubes, datawarehoues or even excel files.

I used a tutorial from MSDN (http://msdn.microsoft.com/en-us/library/bb510523.aspx) for some pica's. I created a datamining project and used the decision tree algoritm. I created the following decision tree:

The tutorial is about the adventureworks database and datamining is created to understand what kind of customers buys bicycles. The business value is to direct new market campaigns when you know more about your customers. You don't want to send offers to the wrong customer. Other examples are for example: what are common (root) causes of network problems, analyzing servicedesk incidents/changes , productionplanning problems. You can use it for every investigation in cause and effects between attributes.

In the above diagram you'll see that (potential) customers in the age between 36 - 44 are highly(?) predictable bike buyers.

The decisiontree algoritm has the following advantages:
  • Good visual presentation.
  • Rules are understandable.
  • predictability.
  • It shows what is important.
This tool is not end user tool but you can use the data mining toolbar in Office 2007 and 2010.

Geen opmerkingen:

Een reactie posten