Generating Nodes from Graphs

One of the most powerful features offered by IBM® SPSS® Modeler graphs is the ability to generate nodes from a graph or a selection within the graph. For example, from a time plot graph, you can generate Derive and Select nodes based on a selection or region of data, effectively "subsetting" the data. For example, you might use this powerful feature to identify and exclude outliers.

Whenever you can draw a band, you can also generate a Derive node. In graphs with two scale axes, you can generate Derive or Select nodes from the regions drawn in your graph. In graphs with marked elements, you can generate Derive nodes, Select nodes, and in some cases Filter nodes from these elements. Balance node generation is enabled for any graph showing a distribution of counts.

Figure 1. Graph with the generate menu showing

Whenever you generate a node, it is placed on the stream canvas directly so that you can connect it to an existing stream. The following nodes can be generated from graphs: Select, Derive, Balance, Filter, and Reclassify.

Select Nodes

Select nodes can be generated to test for inclusion of the records within a region and exclusion of all records falling outside the region or the reverse for downstream processing.

For bands. You can generate a Select node that includes or excludes the records within that band. Select node for Bands only is only available through contextual menus since you need to select which band to use in the Select node.
For regions. You can generate a Select node that includes or excludes the records within a region.
For marked elements. You can generate Select nodes to capture the records corresponding to the marked elements or web graph links.

Derive Nodes

Derive nodes can be generated from regions, bands, and marked elements. All graphs can produce Derive nodes. In the case of evaluation charts, a dialog box for selecting the model appears. In the case of web graphs, Derive Node (“And”) and Derive Node (“Or”) are possible.

For bands. You can generate a Derive node that produces a category for each interval marked on the axis, using the band names listed in the Edit Bands dialog box as category names.
For regions. You can generate a Derive node (Derive as flag) that creates a flag field called in_region with the flags set to T for records inside any region and F for records outside all regions. You can also generate a Derive node (Derive as set) that produces a set with a value for each region with a new field called region for each record, which takes as its value the name of the region into which the records fall. Records falling outside all regions receive the name of the default region. Value names become the region names listed in the Edit regions dialog box.
For marked elements. You can generate a Derive node that calculates a flag that is True for all marked elements and False for all other records.

Balance Nodes

Balance nodes can be generated to correct imbalances in the data, such as reducing the frequency of common values (use Balance Node (reduce) menu option) or boosting the occurrence of infrequent values (use Balance Node (boost) menu option). Balance node generation is enabled for any graph showing a distribution of counts, such as Histogram, Dot, Collection, Bar of Counts, Pie of Counts, and Multiplot.

Filter Nodes

Filter nodes can be generated to rename or filter fields based on the lines or nodes marked in the graph. In the case of evaluation charts, the best fit line does not generate a filter node.

Reclassify Nodes

Reclassify nodes can be generated to recode values. This option is used for distribution graphs. You can generate a Reclassify node for groups to recode specific values of a displayed field depending upon their inclusion in a group (select groups using Ctrl+click on the Tables tab). You can also generate a reclassify node for values to recode data into an existing set of numerous values, such as reclassifying data into a standard set of values in order to merge financial data from various companies for analysis.

Note: If the values are predefined, you can read them into IBM SPSS Modeler as a flat file and use a distribution to display all values. Then generate a Reclassify (values) node for this field directly from the chart. Doing so will put all the target values in the Reclassify node's New values column (drop-down list).

When setting options for the Reclassify node, the table enables a clear mapping from old set values to new values you specify:

Original value. This column lists existing values for the select field(s).
New value. Use this column to type new category values or select one from the drop-down list. When you automatically generate a Reclassify node using values from a Distribution chart, these values are included in the drop-down list. This allows you to quickly map existing values to a known set of values. For example, healthcare organizations sometimes group diagnoses differently based upon network or locale. After a merger or acquisition, all parties will be required to reclassify new or even existing data in a consistent fashion. Rather than manually typing each target value from a lengthy list, you can read the master list of values in to IBM SPSS Modeler, run a Distribution chart for the Diagnosis field, and generate a Reclassify (values) node for this field directly from the chart. This process will make all of the target Diagnosis values available from the New Values drop-down list.

For more information about the Reclassify Node, see Setting Options for the Reclassify Node.

Generating Nodes from Graphs

You can use the Generate menu in the graph output window to generate nodes. The generated node will be placed on the stream canvas. To use the node, connect it to an existing stream.

To generate a node from a graph:

Verify that you are in Explore mode. From the menus, choose View > Explore Mode.
In the Explore mode toolbar, click the Region button.
Define bands, regions, or any marked elements needed to generate your node.
From the Generate menu, choose the kind of node you want to produce. Only those which are possible are enabled.

Note: Alternatively, you can also generate nodes directly from the graph by right-clicking and choosing the generate option you want from the context menus.