Documenting Mining Models (Data Mining Add-ins for Excel)

The Document Model wizard creates a report that provides useful information about the mining models that you have created. By documenting the models that you create, you can track the source of the data used to generate a model, get additional information about when the model was processed, and track parameter changes that affect the results of the model.

Using the Document Model Wizard

1. Click the Data Mining tab.

2. In the Model Usage group, click Document Model.

3. In the Select Model dialog box, select the model on which to report, and then click Next. You must run the Document Model wizard separately for each model that you want to document.

4. In the Select documentation details dialog box, choose one of two options: Complete information or Summary information.

5. Click Finish.

6. The wizard automatically creates a new worksheet that contains the specified report, titled Model Documentation,

Understanding the Report

When you create a report that documents a data mining model, you can create a summary,which contains basic information containing the name and description of the model, or a complete report, which contains details about the underlying structure and advanced information about the mining model.

Depending on the algorithm that was used to create the model, different types of information is provided. For example, in an association model, you are more interested in knowing the number of itemsets and rules that were generated. For a clustering model, the number of clusters is more interesting.

The following table lists the options and the information that is provided in the report for each option.

Note

The columns in the report are set to a particular size by default. Therefore, if any columns names or values are very long, they might not be visible, or might appear as ### in Excel. To make the values visible, you can resize the row. If the cell is selected, you can click and drag the double arrows at the right end of the formula bar to show the complete value or string.

Summary Report

Metadata	Model name Model description Algorithm name Date last processed
Model results	Association	Count of itemsets Count of rules
	Clustering	Count of clusters Support for each cluster
	Decision tree	Number of trees Number of nodes in each tree
	Linear regression	Number of trees (always 1) Number of nodes (always 1)
	Naïve Bayes	Important attributes
	Neural network	Number of input nodes Number of output nodes Number of hidden nodes
	Sequence clustering	Number of clusters

Complete Report

The complete report contains everything that is in the summary report, plus detailed information about the columns of data used in the model and the results of analysis:

Metadata	Model metadata	Algorithm parameters and values
	Column metadata	Column name Usage Data type Content type Values (list of discrete values, or range of values)
Model statistics	Continuous columns	Mean value Minimum value Maximum value Root mean square error Mean absolute error Log score Regression formula (for linear regression models only)
	Discrete columns	Count of passing Count of failing Log score Lift

Note

You can document any model type that is supported by SQL Server Analysis Services. Therefore, the table lists some model types that cannot be created by using the Table Analysis Tools or by using the wizards in the Data Mining Client. However, you can create all model types by using the Advanced Data Mining Query Editor. For more information, see Query (SQL Server Data Mining Add-ins).

See Also

Deploying and Scaling Mining Models (Data Mining Add-ins for Excel)