Megosztás a következőn keresztül:

  • Cikk

Viewing a Mining Model with the Microsoft Naive Bayes Viewer

The Microsoft Naive Bayes Viewer in Microsoft SQL Server Analysis Services displays mining models that are built with the Microsoft Naive Bayes algorithm. The Microsoft Naive Bayes algorithm is a classification algorithm for use in predictive modeling. For more information about this algorithm, see Microsoft Naive Bayes Algorithm.

Because one of the main purposes of a naive Bayes model is to provide a fast way to explore the data in a dataset, the Microsoft Naive Bayes Viewer provides several methods for displaying the interaction between predictable attributes and input attributes for a case table.

Note

To view detailed information about the equations used in the model and the patterns that were discovered, use the Microsoft Generic Content Tree viewer. For more information, see Viewing Model Details with the Microsoft Generic Content Tree Viewer or Microsoft Generic Content Tree Viewer (Data Mining Designer).

Viewer Tabs

When you browse a mining model in Analysis Services, the model is displayed on the Mining Model Viewer tab of Data Mining Designer in the appropriate viewer for the model. The Microsoft Naive Bayes Viewer provides the following tabs for exploring data:

  • Dependency Network

  • Attribute Profiles

  • Attribute Characteristics

  • Attribute Discrimination

Dependency Network

The Dependency Network tab displays the dependencies between the input attributes and the predictable attributes in a model. The slider at the left of the viewer acts as a filter that is tied to the strengths of the dependencies. Lowering the slider shows only the strongest links.

When you select a node, the viewer highlights the dependencies that are specific to the node. For example, if you choose a predictable node, the viewer also highlights each node that helps predict the predictable node.

The legend at the bottom of the viewer links color codes to the type of dependency in the graph. For example, when you select a predictable node, the predictable node is shaded turquoise, and the nodes that predict the selected node are shaded orange.

Back to Top

Attribute Profiles

The Attribute Profiles tab displays histograms in a grid. You can use this grid to compare the predictable attribute that you select in the Predictable box to all other attributes that are in the model. Each column in the tab represents a state of the predictable attribute. If the predictable attribute has many states, you can change the number of states that appear in the histogram by adjusting the Histogram bars. If the number you choose is less than the total number of states in the attribute, the states are listed in order of support, with the remaining states collected into a single gray bucket.

To display a Mining Legend that relates the colors of the histogram to the states of an attribute, click the Show Legend check box. The Mining Legend also displays the distribution of cases for each attribute-value pair that you select.

To copy the contents of the grid to the Clipboard as an HTML table, right-click the Attribute Profiles tab and select Copy.

Back to Top

Attribute Characteristics

To use the Attribute Characteristics tab, select a predictable attribute from the Attribute list and select a state of the selected attribute from the Value list. When you set these variables, the Attribute Characteristics tab displays the states of the attributes that are associated with the selected case of the selected attribute. The attributes are sorted by importance.

Back to Top

Attribute Discrimination

To use the Attribute Discrimination tab, select a predictable attribute and two of its states from the Attribute, Value 1, and Value 2 lists. The grid on the Attribute Discrimination tab then displays the following information in columns:

  • Attribute
    Lists other attributes in the dataset that contain a state that highly favors one state of the predictable attribute.

  • Values
    Shows the value of the attribute in the Attribute column.

  • Favors <value 1>
    Shows a colored bar that indicates how strongly the attribute value favors the predictable attribute value shown in Value 1.

  • Favors <value 2>
    Shows a colored bar that indicates how strongly the attribute value favors the predictable attribute value shown in Value 2.

Back to Top