Getting Started with JDM Web Services - Java Data Mining: Strategy, Standard, and Practice - page 377

level, marital-status , and capital-gains. 1 The system executes the
singleRecordApply task using the selected models and collects the
probability that the caller will buy each of the respective products.
The scenario works as follows:

• The Web page provides fields to be filled by the call center
representative, who selects values from a drop-down list of
discrete attributes, and types in values for continuous
attributes. This portion of the HTML page is written in
javascript code for the field value selection. The list of
fields is defined at design time and will be the same for
all models called from this page.
• When requested by the user of the Web browser
(through the Score submit button shown in the screen
shot), the system connects to a JDM Web services pro-
vider chosen at design time, obtains the names of mod-
els available in the MOR whose name begins with Product ,
and, expecting that these models are classification models,
scores each input data record on each model and reports to
the representative the probabilities returned by the models.

Figure 14-2 is a screen shot of the Web page.

Figure 14-2 Product recommendation PHP page.

1 These three attribute names have been taken from a well-known dataset used in

data mining packages called “census” and do not claim to be truly useful
attributes for product recommendation.