A data mining solution is an Analysis Services solution that contains one or more data mining projects.
The topics in this section provide information about how to design and implement an integrated data mining solution by using SQL Server Analysis Services. For an overview of the data mining design process and related tools, see Data Mining Concepts.
For more information about additional projects types that are useful for data mining, see Related Projects for Data Mining Solutions.
Relational vs. Multidimensional Solutions
Deploying Data Mining Solutions
A data mining solution can be based either on multidimensional data—that is, an existing cube—or on purely relational data, such as the tables and views in a data warehouse, or on text files, Excel workbooks, or other external data sources.
You can create data mining objects within an existing multidimensional database solution.
Typically you would create a solution like this if you have already created a cube and want to perform data mining by using the cube as a data source. When you move and backup models based on a cube, the cube must also be moved or copied.
You can create a data mining solution that contains only data mining objects, including the supporting data sources and data source views, and that uses relational data source only.
This is the preferred method for creating data mining models, as processing and querying is generally fastest against relational data sources. You can also easily move and backup models between servers by using the EXPORT and IMPORT commands.
The instance of Analysis Services to which you deploy the solution must be running in a mode that supports multidimensional objects and data mining objects; that is, you cannot deploy data mining objects to an instance that hosts tabular models or PowerPivot data.
Therefore, when you create a data mining solution in Visual Studio, be sure to use the template, Analysis Services Multidimensional and Data Mining Project.
When you deploy the solution, the objects used for data mining are created in the specified Analysis Services instance, in a database with the same name as the solution file.
For more information about how to deploy both relational and multidimensional solutions, see Deployment of Data Mining Solutions.
After you have created a basic data mining solution, including data sources and a mining structure, you can build on the solution by adding new models, testing and comparing models, creating predictions, and experimenting with subsets of data.
For more information, see the following links:
Test the models you create, validate the quality of your training data, and create charts that represent the accuracy of data mining models.
Train the model by populating the structure and related models with data. Update and extend models with new data.
Customize a mining model by applying filters to the training data, choosing a different algorithm, or setting advanced algorithm parameters.
Customize a mining model by applying filters to the data used in training the mode.
Update and manage data mining solutions.