Deploying SPSS Modeler flows and models
You can use SPSS Modeler flows and models in Watson Machine Learning if you have the Watson Machine Learning service.
In Watson Machine Learning, you add your flows and models to a deployment space, where you can test and manage them. In a deployment space, you can prepare your flows and models for use in pre-production or production environments to generate predictions and insights.
Models can be saved as either a scoring branch or as predictive model markup language (PMML). PMML is an XML format for describing data mining and statistical models. It includes inputs to the models, transformations that are used to prepare data for data mining, and the parameters that define the models themselves. If you save models as PMML, it is possible to share models with other applications that support this format. For more information about PMML, see the Data Mining Group website.
Models can be deployed in either online or batch deployments. Flows can be deployed only in batch deployments. For more information about deployments, see Creating online deployments in Watson Machine Learning and Creating batch deployments in Watson Machine Learning.
For flows in a deployment space, you can decide which terminal nodes to run in the flow each time that you create a batch job from the flow. You can use this flexibility to run the whole flow or only a few nodes from it.
Promoting SPSS Modeler models to Watson Machine Learning
For models, you can promote them to a deployment space after you save them to the project.
- In your SPSS Modeler flow, click the Save Model icon on the toolbar.
- In Branch terminal node, select the node that you want to make into a model.
- Enter a name and click Save to save the model to as a project asset.
- Click the Assets tab and find the model. For the model, click the
overflow menu
and select Promote to
space. - In Target space, select the deployment space.
- Click Promote.
In the deployment space, you can then create deployments and jobs to generate predictions for new data. For more information, see Deploying AI assets.
For a list of which data sources are supported for models in Watson Machine Learning, refer to the SPSS section under Batch deployment input details for SPSS models.
Importing SPSS Modeler flows into Watson Machine Learning
For flows, you can add them to a deployment space by exporting the project that they are in. You can then import the project in the deployment space. For more information about exporting and importing projects, see Exporting project assets and Importing space and project assets into deployment spaces.
You do not need to deploy the flow in the deployment space to create batch jobs. For more information about creating batch jobs for flows in deployment spaces, see Creating deployment jobs for SPSS Modeler flows.
When you create a batch job for a flow that contains multiple import nodes, Watson Machine Learning automatically maps the data sources and targets for nodes by name. Promote or import all your data sources and targets before you create any batch jobs for a flow. All the data assets and connections that are used in your flow need to be in your deployment space.