Deploying a Model

This procedure describes how to configure the PMML Model Scoring stage to deploy an analytics model as part of a dataflow.

  1. Under Primary Stages > Deployed Stages > Analytics Scoring, click the PMML Model Scoring stage and drag it onto the canvas, placing it where you want on the dataflow and connecting it to input and output stages.
  2. Double click the PMML Model Scoring stage to show the PMML Model Scoring Options dialog box.
    By default the options dialog shows the details of the first model in the list of available models.
  3. Click the Type Filter drop-down and select the model type to filter by.
    Only model types which are associated with at least one model in the Analytics Scoring Repository are listed.
  4. Click the Model drop-down and select the model to deploy.
    The details of the chosen model are displayed in the dialog.
    Model type
    The type of the selected model as described in Supported Model Formats.
    Description
    A short text about the purpose of the model.
    Inputs
    A table containing information about all the required input fields for the model. Each row contains information about an input field. The model input field name will automatically be mapped to a valid Spectrum field name on publish (see Add a Model in a Web Browser Client or Add a Model via Spectrum Miner Spectrum Connector).
    Model Field Name
    The name of the field as specified in the model definition.
    Spectrum Field Name
    The name of the field as used in the Spectrum platform.
    Model Field Type
    the type of the field as specified in the model definition.
    Spectrum Field Type
    The Spectrum field type that is mapped to the model field type as described in QMML and PMML.
  5. Select the Configuration tab.
    Details about the model's Outputs are displayed in a table. It contains information about all the output fields for the model. Each row has information about an output field. The model output field name will automatically be mapped to a valid Spectrum field name on publish (see Add a Model in a Web Browser Client or Add a Model via Spectrum Miner Spectrum Connector).
    Model Field Name
    The name of the field as specified in the model definition.
    Spectrum Field Name
    The name of the field as used in the Spectrum platform.
    Model Field Type
    The type of the field as specified in the model definition.
    Spectrum Field Type
    The Spectrum field type that is mapped to the model field type as described in QMML and PMML.
    Include
    A check box to specify whether to use this output.
  6. Optional: Uncheck the Included column of any row in the Outputs table to exclude the output (stop it from being written to the output channel).
    At least one output must remain selected. If all outputs are excluded then a validation error symbol will appear beside the Outputs table. This means the current model configuration is invalid and the model cannot be deployed. The validation error symbol will remain visible until the error is corrected.
  7. Optional: Click the Spectrum Field Name column of any row in the Outputs table to rename the fields as required.
    No two outputs can share the same Spectrum Field Name and Spectrum Field Names must follow the standard Spectrum Technology Platform field naming conventions. If any validation errors are detected then a validation error symbol is displayed beside the Outputs table; hovering the mouse cursor over a validation error symbol shows the error details.
    Note: Renaming an output's Spectrum Field Name only affects the specific instance of the stage, it does not update the Analytics Scoring Repository .
    Note: Spectrum Technology Platform field names must:
    • Be unique.
    • Be non-empty.
    • Contain only alphanumeric, period or underscore characters.
    • Cannot start with a period.
  8. Finally click OK to save the chosen model and configuration.