Udostępnij za pośrednictwem


Creating a New Mining Structure

When you build data mining models in Microsoft SQL Server 2005 Analysis Services (SSAS), the first step is to create a mining structure, by using the Data Mining Wizard in Business Intelligence Development Studio. The mining structure defines the data domain from which mining models are built.

Mining structures can be based on either relational or online analytical processing (OLAP) data sources. Relational mining structures describe data that is stored in relational database systems. OLAP mining structures are built by using an OLAP cube that exists on the same database as the mining structure.

For More Information:Designing and Creating Databases, Working with Online Analytical Processing (OLAP)

The Data Mining Wizard automatically defines a mining structure and adds an initial mining model to the structure. Because a mining structure can contain multiple mining models, you can use Data Mining Designer to add more mining models to the structure.

The following sections provide more information about creating new mining structures with the Data Mining Wizard.

  • Creating a Relational Mining Structure
  • Creating an OLAP Mining Structure

Creating a Relational Mining Structure

Relational mining structures can be based on any data that is available through an OLE DB data source. If the source data is contained within multiple tables, you can feed it into the wizard as a single case table by using nested tables.

For More Information:Nested Tables

The Data Mining Wizard guides you through the following steps to create the structure for a new mining model:

  1. Selecting a data source type, in this case a relational database.
  2. Selecting an algorithm.
  3. Selecting a data source.
  4. Selecting a case table and, optionally, any nested tables.
  5. Selecting the type for each column, either predictable, input, or key.
  6. Specifying the column content types.
  7. Naming and saving the new mining structure and the associated mining model.

On the last page of the wizard, you have the option to enable drill through. If you enable this option, you can explore the data that the algorithm uses to create the mining model, after the model has been processed.

For More Information:Data Mining Algorithms, Designing and Creating Databases, Mining Model Columns, Mining Structure Columns, Data Types (Data Mining), Content Types (Data Mining)

Creating an OLAP Mining Structure

OLAP cubes frequently contain so many members and dimensions that it can be difficult to manually identify the patterns that the cubes contain. However, you can find many of these patterns by using data mining techniques, and you can then apply the knowledge that you derive from these patterns to important business decisions.

The following table lists several common OLAP data mining tasks, describes sample scenarios in which you might apply each task, and identifies the data mining algorithm to use for each task.

Task Sample scenario Algorithm

Group members into clusters

Segment a customer dimension based on customer member properties, the products that the customers buy, and the amount of money that the customers spend.

Microsoft Clustering Algorithm

Find interesting or abnormal members

Identify interesting or abnormal stores in a store dimension based on sales, profit, store location, and store size.

Microsoft Decision Trees Algorithm

Find interesting or abnormal cells

Identify store sales that go against typical trends over time.

Microsoft Time Series Algorithm

Use the Data Mining Wizard to define the initial structure of your OLAP mining model. The wizard guides you through the following process to create the structure for a new mining model:

  1. Selecting a data source type, in this case a cube.
  2. Selecting an algorithm.
  3. Selecting a source cube dimension.
  4. Selecting a case key.
  5. Selecting case columns.
  6. Selecting any nested tables.
  7. Selecting the usage for each column, either predictable, input, or key.
  8. Specifying the column content types.
  9. Slicing the source cube.
  10. Naming and saving the new mining structure and the associated mining model.

You can set the following options on the last page of the wizard:

  • Allow drill through
  • Create mining model dimension
  • Create a cube using mining model dimension

If you enable drill through, you can explore the data that the algorithm uses to create the mining model, after the model has been processed.

If you choose to create a new data mining dimension in the source cube, you can include information that the data mining algorithm finds in the OLAP data source. If you select the option to create a new cube, a new cube is defined on the database that includes the data mining dimension.

For More Information:Data Mining Algorithms, Mining Model Columns, Mining Structure Columns, Data Types (Data Mining), Content Types (Data Mining)

See Also

Concepts

Mining Structures (Analysis Services)
Data Mining Designer
Data Mining Wizard
Managing Mining Models in Data Mining Designer
Managing Mining Structures in Data Mining Designer
Setting Column Properties on a Mining Model
Setting Column Properties on a Mining Structure

Other Resources

Mining Structure Tab How-to Topics

Help and Information

Getting SQL Server 2005 Assistance