Managing Mining Models in Data Mining Designer
On the Mining Models tab of Data Mining Designer, you can modify the mining models that exist in a mining structure and add new mining models to the structure. Mining models are based on the mining structures that you define with the Data Mining Wizard.
The Mining Models tab consists of a grid with one column that describes the mining structure, and additional columns that describe each associated mining model. Each cell in the structure column of the grid lists a column that is defined in the structure, whereas each cell in a mining model column of the grid lists how the model uses the columns from the mining structure.
Within the Mining Models tab, you can process mining models or create new mining models. You can adjust the properties for a mining structure and its associated mining models by using the Properties window. You can adjust the parameters of the algorithm that is used to define the mining model by using the Algorithm Parameters dialog box.
After you complete the Data Mining Wizard, the Mining Structures folder in Solution Explorer contains a single mining model that is based on the mining structure that you defined in the wizard. You can add other models to the mining structure by using Data Mining Designer. Although new models must all share the same mining structure, you can vary the algorithm type, column usage, and algorithm-specific parameters for each model.
There are several advantages to creating multiple models based on a single mining structure:
Each type of algorithm displays results in a different way. Defining more than one model from the same structure lets you obtain different information from the same data. For example, you might want to use a clustering model to explore the data, and a decision tree model to create predictions from the data.
The results of a mining model can be influenced by how the parameters are set. You can create several different models with the same algorithm and vary only the setting of a specific parameter. You can then compare the results so that you can select the best setting for the algorithm.
You can apply a filter to the model that controls the data that is used when training and testing the model.
The results of a mining model are affected by the input columns that you select. You can build several models that vary only in the input columns that are used, and then compare the results to determine which columns you should use as inputs.
For More Information: How to: Add a Mining Model to an Existing Mining Structure, Data Mining Algorithms (Analysis Services - Data Mining), Creating Filters for Mining Models (Analysis Services - Data Mining)
On the Mining Models tab, you can change the model after it has been created. The algorithm type, the model name, and the parameters that are specific to each algorithm can be modified. You can also change the usage of columns, add aliases to model columns, and create a filter that is applied to the mining model when training and testing.
For More Information: Mining Models Tab: How-to Topics
You can also modify a mining model by making changes to the underlying mining structure in the Mining Structure tab.
For More Information: Managing Mining Structures in Data Mining Designer
Changing Mining Column Usage
You can change which columns are included in a mining model and how each column is used, such as input, key, or predictable, by using the cells for that model column in the grid on the Mining Models tab. Each cell corresponds to a column in the mining structure. For key columns, you can set the cell to Key or Ignore. For input and output columns, you can set the cell to the following values:
If you set a cell to Ignore, the column is removed from the mining model, but that column can still be used by other mining models in the structure.
Aliasing Model Columns
When Analysis Services creates the mining model, it uses the same column names that are in the mining structure. You can add an alias to any column in the mining model. This might make it easier to understand the column contents or usage, or make the name shorter for convenience in creating queries.
You create an alias by editing the Name property of the mining model column. Analysis Services continues to use the original name as the ID of the column, and the new value that you type for Name becomes the column alias, and appears in the grid in parentheses next to the column usage.
This example shows related models that have multiple copies of a mining structure column related to income. Each copy of the structure column has been discretized in a different way. The models in the diagram each use a different column from the mining structure; however, for convenience in comparing the columns across the models, the column in each model has been renamed to [Income].
You can add a filter to a mining model. A filter is a set of WHERE conditions that restrict the data in the model cases to some subset. The filter is used when training the model, and can optionally be used when you test the model or create accuracy charts.
For more information, see Creating Filters for Mining Models (Analysis Services - Data Mining).