Deploying an FFNN-based alarm system
In this section, you will deploy the previously trained alarm system within a dedicated workflow. This deployment workflow will need the following:
- The ability to accept new data in the same shape and format as the original data
- The normalization model to apply to the new data
- The same preprocessing steps as for the testing part of the workflow
- The trained FFNN model
- The Keras Network Executor node
- The same postprocessing of the results to create the alarm system
We can do this manually by using the workflow shown in Figure 9.10. However, we can also create the deployment workflow automatically by replicating the testing part of that workflow. KNIME’s Integrated Deployment can do this and consists of three nodes: the Capture Workflow Start node, the Capture Workflow End node, and the Workflow Writer node.
The Capture Workflow Start and Capture Workflow End nodes are placed at the beginning and the end...
