Create new nodes for IBM SPSS Modeler 16 using R


Visual programming in IBM SPSS Modeler is based on icons called SPSS Modeler nodes. The user creates a process that runs data through a series of nodes called streams. The nodes represent operations to be performed on the data, and the links between the nodes indicate the direction of data flow. Typically, you use a data stream to read data from a collection of data sources, you manipulate the data, and you send the data to a destination, which can be a table or a viewer.

The nodes can be linked to form a stream. The stream represents a flow of data through a number of operations to a destination. This user-friendly interface makes it possible to analyze complex data sets and create powerful predictive models without programming.

Extend function for IBM SPSS Modeler with new nodes

To add functions not included in SPSS Modeler, you can create nodes. The more nodes you have in SPSS Modeler, the more operations you can perform. Use the Component-Level Extension Framework, a mechanism that enables you to add user-provided extensions to the standard functions of SPSS Modeler.

Because SPSS Modeler 16 is integrated with the R programming language, you can now run R scripts. This article describes how to develop new extensions using R. Figure 1 shows three new nodes: R Transform, R (for modeling), and R Output. You can insert your R script directly into these nodes in SPSS Modeler.

Figure 1. Nodes in SPSS Modeler where you can insert R scripts
Image shows nodes in SPSS Modeler where you can insert R                         scripts

What you'll need to get started

To run R code in SPSS Modeler, you need to download and install R 2.15.2 and the IBM SPSS Modeler Essentials for R Plugin.

SPSS Modeler is not intended to serve as a workbench to write the R code; rather, SPSS Modeler is only able to run R code. Use a workbench such as RStudio to write and test R code before porting it to IBM SPSS Modeler.

To learn how to configure IBM SPSS Modeler 16 and the R integration, see the developerWorks article "Calling R from SPSS."

R programming language and environment

R is a powerful open source statistical language and environment, which offers a rich analytic ecosystem for data exploration, visualization, statistical analysis, modeling, machine learning, simulations and more. It is an emerging competitor to proprietary platforms.

R is taught at universities; many statistics and math courses now use R. IT shops are adopting R, and many companies are integrating it into their products. A vibrant, active user community is growing around R and jobs, demand for R skills are on the rise, and the number of R packages are increasing. Also, emerging algorithms usually appear in the freely available R programming language, before they are available in commercial packages.

R does require the developer to overcome a learning curve. Before adopting R, organizations must consider the level of support offered because open source packages have varying levels of quality. R is based on an in-memory architecture.

The ability to create nodes for SPSS Modeler enables you to use the more than 5,500 R packages available to analysts, even if you don't have R programming skills, and even if you do not write any code.

Custom Dialog Builder

The Custom Dialog Builder (first available in IBM SPSS Statistics) enables you to create and manage nodes to be used in SPSS Modeler 16. To open the Custom Dialog Builder (shown in Figure 2), click Tools > Custom Dialog Builder for R in the main menu, which shows the following elements:

  • Dialog canvas— Area where you design the layout of the node dialog.
  • Properties pane— List of properties that make up the node dialog and the properties of the dialog, such as the node type.
  • Tools palette— Set of controls that can be included in a custom node dialog.
Figure 2. Custom Dialog Builder
Image shows Custom Dialog Builder
Image shows Custom Dialog Builder

Example 1. Create a geocoding node as your first node

As an example of how to create a new node, develop a node to perform geocoding of geographical data. Geocoding is the process to find geographic coordinates from other geographic data, such as street addresses, ZIP codes, and similar data. Many Internet companies — such as OpenStreetMap, Mapquest, Bing, and Google Maps — provide geocoding services. This article describes how to build a node based on the Google Maps API.

Step 1. Get and test the R code for geocoding

To create a new node, get the appropriate R code for the geocoding task. The R code can be written by any user with R skills or can be downloaded from the Internet. This article relies on the R code in Listing 1 from the article "Using Google maps API and R." Before you use this code in SPSS Modeler, test the code using RStudio or the R console.

  1. Open Rstudio and run the code from Listing 1.
    • Create a variable with an address in a string, such as address<-geoCode("The White House, Washington, DC").
    • Run the commands address[1] to get the latitude and address[2] to get the longitude. The results should be "38.8976831" and "-77.0364972."
      Listing 1. R code performing geocoding using Google Maps API
      #### This script uses RCurl and RJSONIO to download data from Google's API:
      #### Latitude, longitude, location type (see explanation at the end), formatted address
      #### Notice ther is a limit of 2,500 calls per day
      url <- function(address, = "json", sensor = "false") {
       root <- ""
       u <- paste(root,, "?address=", address, "&sensor=", sensor, sep = "")
      geoCode <- function(address,verbose=FALSE) {
       if(verbose) cat(address,"\n")
       u <- url(address)
       doc <- getURL(u)
       x <- fromJSON(doc,simplify = FALSE)
       if(x$status=="OK") {
       lat <- x$results[[1]]$geometry$location$lat
       lng <- x$results[[1]]$geometry$location$lng
       location_type <- x$results[[1]]$geometry$location_type
       formatted_address <- x$results[[1]]$formatted_address
       return(c(lat, lng, location_type, formatted_address))
       } else {
       return(c(NA,NA,NA, NA))
      ##Test with a single address
      #address <- geoCode("The White House, Washington, DC")
      #[1] "38.8976831"
      #[2] "-77.0364972"
      #[3] "APPROXIMATE"
      #[4] "The White House, 1600 Pennsylvania Avenue Northwest, Washington, D.C., DC 20500, USA"
  2. To verify that the coordinates are correct, point your browser to Google Maps and paste the coordinates in the search box, as shown Figure 3. In this case, the geocoding worked well, and the coordinates point to the White House in Washington.
    Figure 3. Plot of the coordinates [38.8976831, -77.0364972] in Google Maps
    Image shows coordinates [38.8976831, -77.0364972] in                         Google Maps
    Image shows coordinates [38.8976831, -77.0364972] in Google Maps

When you run the code in Listing 1, the following tasks are performed:

  1. The URL to the web service is created. Because this function uses a web service, the first action is to create the URL to call the service. The address passed as an argument is written with spaces, but the spaces are removed to create a URL in the specific format required by the Google Maps API. The result can be in XML or JSON. In this example, the JSON output is used.
  2. The connection is created and the JSON is received.
  3. The JSON is parsed to get the latitude and the longitude of the address given between others.

Third-party APIs, such as the Google Maps API, can have usage limits. Free web services do provide geocoding, and the code has only to be adapted to use them. Data provided by free services might be less accurate, however.

Step 2. Create the new node on IBM SPSS Modeler 16

Now that the code has been tested, the next step is to develop the new node.

  1. Open IBM SPSS Modeler 16 and open the Custom Dialog Builder.
  2. In the properties of the node, put the same parameters as in Figure 4. For Node Icon, download the image geocode_icon.gif from the sample code for this article and use it as an icon for the new node you are creating.
    Figure 4. Properties to set in the Custom Dialog Builder
    Image shows properties to set in the Custom Dialog                         Builder
    Image shows properties to set in the Custom Dialog Builder
  3. From the tools, drag and drop Field Chooser, double-click Field Chooser, change identifier to address, and change Title to Address field.
  4. Click Edit > Script Template and paste in the R code of Listing 2.

The node is almost created, but you must make some modifications in the R code to map the data coming from the SPSS Modeler stream. See the documentation on "Allowable Syntax" for SPSS Modeler. It describes the statements and functions recognized by R. For help, see the documentation for the script template, under Examples.

Specify control identifiers in the form %%Identifier%% at the appropriate location. Press Ctrl+Spacebar to show a list of available control identifiers. In this case, the value is %address%%.

location <- modelerData$%%address%%

The R object modelerData is a data frame that contains the original data. To add the new columns for latitude and longitude, use the cbind function to create a data frame with the original data plus the output generated, as shown.


The R object var1 sets up a new field in SPSS Modeler for the data model that describes the type and structure of the new data generated. The name of the new field and the type of storage are specified in this new field.

var1<-c(fieldName="Latitude",fieldLabel="",fieldStorage="real",fieldFormat="",fieldMeasure="",  fieldRole="")

The R object modelerDataModel contains the data model for the original data with the extra field generated. The extra field is called Latitude and it has characteristics specified.


When you are working on your R code in SPSS Modeler, you can perform limited debugging of the R script by using commands including print() and str().

Figure 5 shows the script template and the node at the end of development.

Figure 5. Properties to set in the Custom Dialog Builder
Image shows properties to set in the Custom Dialog                         Builder
Image shows properties to set in the Custom Dialog Builder

The following code listing shows the required modifications.

Listing 2. R code with modifications for geocoding
location <- modelerData$%%address%%
root <- ""
u <- paste(root,"json", "?address=", location, "&sensor=", "false", sep = "")
u <- gsub(' ','%20',u) #Encode URL Parameters
doc <- aaply(u,1,getURL)
 json <- alply(doc,1,fromJSON,simplify = FALSE)
coord = laply(json,function(x) {
    if(x$status=="OK") {
      lat <- x$results[[1]]$geometry$location$lat
      lng <- x$results[[1]]$geometry$location$lng
    } else {
var1<-c(fieldName="Latitude",fieldLabel="",fieldStorage="real",fieldFormat="",fieldMeasure="",  fieldRole="")
var2<-c(fieldName="Longitude",fieldLabel="",fieldStorage="real",fieldFormat="",fieldMeasure="",  fieldRole="")

Step 3. Save and install the new node

After the development is finished, save the node to distribute it to colleagues or to other SPSS Modeler users. Click File > Save in the Custom Dialog Builder. The file is saved with the extension .cfd. Install the new node. Click File > Install. Close the Custom Dialog Builder. The new node is in the Record Ops palette as specified in the properties.

Step 4. Test the geocoding node

Test the node by generating manual data using the User Input node:

  1. Click the Sources palette and drag and drop the User Input node into the canvas.
  2. Double-click the User Input node, create a new field called Location, and select String as storage. Specify addresses as values. In this example, the following addresses are used: "New York City""San Francisco,California""Paris, France".
  1. Click Preview to visualize the generated data in a table.
    Figure 6. Data manually generated with the User Input node
    Image shows data manually generated with the User Input                         node
  2. Select the new Geocoding node just created from the Field Ops palette. Connect the User Input node to the Geocoding node and a Table node as output.
    Figure 7. Stream to test new geocoding node
    Image shows stream to test  new geocoding node
  3. Double-click the Geocoding node. In the Address field, select Location.
  4. Run the stream. The expected output is a table with three columns: location, latitude, and longitude.
    Figure 8. Completed stream to test new geocoding node
    Image shows completed stream to test  new geocoding node
    Image shows completed stream to test new geocoding node

The first new node using R is now created and working.

Example 2. Create a visualization node for geospatial data

In Example 1, some addresses have been converted into coordinates. The output is a table that confirms that the code is running properly. However, there is no direct visualization to check that these coordinates are correct. The best way to visualize geospatial data is by using a map. This second example shows how to create a node that plot dots in a dynamic map that is run in a web browser. Because the map uses the Google Maps API, the name of the new node is the Google Maps node.

To create this node, use the R code described in "Mine spatial data with space-time-boxes in IBM SPSS Modeler and visualize the data with R." This R code uses the R package plotGoogleMaps, which provides an interactive plot for handling the geographic data within web browsers.

Step 1. Test and understand the code in RStudio

Test the code before you use SPSS Modeler. With Rstudio, you can easily modify and debug the code if it's not working.

  1. Open Rstudio and create the latitude and longitude variables as shown in Listing 3. For example, the following values are for the White House in Washington, D.C.
    Listing 3. Longitude and latitude variables
    latitude<- 38.8976831
    longitude<- -77.0364972
  2. Run the code in Listing 4. A browser opens to a page with the map and the dot is in the correct place.
    Listing 4. Example of space-time-box geohash
    coordinates(df)<-~longitude+latitude # convert to SPDF
    proj4string(df) <- CRS('+init=epsg:4326')
    # adding Coordinate Referent Sys.
    # Create web map of Point data

The code gets the coordinates and transforms them to the coordinate system that the Google Maps API understands by using the R function proj4string. The function plotGoogleMaps, which is included in the plotGoogleMaps package, is run. In this example, the output is stored on the C drive with the name of MyNewMap.htm.

Step 2. Create the Google Maps node in SPSS Modeler 16

Create the node by using the Custom Dialog Builder, then map the data:

  1. Open the Custom Dialog Builder and select the properties shown in Figure 9. Download the icon file GoogleMaps.gif from the sample code for this article. Note that now the Node Type is Output.
    Figure 9. Properties for Google Maps node
    Image shows properties for  Google Maps node
    Image shows properties for Google Maps node
  2. Select two Field Chooser elements and put them in the canvas; one is for the latitude and the other is for the longitude. The result looks similar to Figure 10.
    Figure 10. Properties for field choosers of the Google Maps node
    Image shows properties for field choosers of  Google Maps                         node
    Image shows properties for field choosers of Google Maps node
  3. Click Edit > Script and add the code from Listing 5. As explained, adapt the R code in Listing 4 to SPSS Modeler.
    Listing 5. R script to insert in the Custom Dialog Builder
    proj4string(modelerData) <- CRS('+init=epsg:4326')

Save and install the node.

Step 3. Test the node

To test the node, plot the data coming from the data previously created. Select the Google Maps from the Output palette and connect it to the Geocoding node.

Figure 11. Final stream with the new Google Maps node as an output
Image shows final stream with  new Google Maps node as                          output

Run the stream. The browser opens a page with a dynamic map, which plots the addresses generated.

Figure 12. Output with the three data points in the map
Image shows output with                     the three data points mapped
Image shows output with the three data points mapped

Dependencies to be aware of

The nodes created in this article use web services and maps that require an Internet connection. Before the nodes can be used, the libraries used by the nodes must be installed. In the two nodes created in this article, the R packages plotGoogleMaps, Rcurl, RJSONIO, and plyr must be installed. To install a package, build a stream as shown in Figure 13.

Figure 13. Stream used to install R packages in SPSS Modeler 16
Image shows stream used to install R packages in SPSS Modeler                         16

Double-click the R node, insert the code in Listing 6, and run the stream, which downloads and installs the library. This task must be performed only once. You can also run this code directly in the R console or RStudio.

Listing 6. Example of space-time-box geohash

You can create extensions using CLEF. Code development using CLEF is more complex than using R, but with CLEF, you can ensure that extension modules can look and act the same as native SPSS Modeler modules, and perform tasks at the same or similar speed and efficiency as native nodes.


This article describes how to use the integration of SPSS Modeler and R to create new nodes. Users who do not have R skills can use all of the powerful R packages, available for free. The R functions in SPSS Modeler are extended with technologies including SQL pushback to run the code faster. Features not originally included in SPSS Modeler can be added. An additional benefit is that nodes can be shared with the user community to make analytics technology available to non-experts.

To share the nodes in your organization, use the repository feature included in SPSS Collaboration and Deployment services. A public repository is available from the IBM SPSS DevCentral developerWorks community.

Downloadable resources

Related topics


Sign in or register to add and subscribe to comments.

Zone=Data and analytics
ArticleTitle=Create new nodes for IBM SPSS Modeler 16 using R