0 ×

01_​Deploying_​and_​Validating_​models_​as_​WebServices

Workflow

Validating KNIME Workflows

This workflow demonstrates how to take a standard KNIME workflow and adapt it to be deployable as a web service using the KNIME Server and testable using the KNIME Testing Framework. This allows automatic testing to validate that the workflow and/or the web service are still producing the expected results.

reproducibility validation testing XGBoost RDKit
USER: provide data file here Adds reference dataif the input table isempty Tests the predictionsagainst a goldentable (if necessary) USER: here are theresults! Generate fingerprints and model predictions for the molecules in the table Deploying and validating models as web servicesThis workflow demonstrates a technique to deploy a model prediction workflow as a web service and to validate that it is doing what it'ssupposed to do. The key is the use of a reference set of input data and the Table Difference Checker in the "Run Tests if Necessary"wrapped metanode at the end of the workflow. The model can be validated in KNIME Analytics Platform using the KNIME TestingFramework and, once it has been deployed as a web service using KNIME Server, can be validated by calling the service without anyrows. The workflow itself generates bioactivity predictions for the input molecules using a set of pre-calculated and saved XGBoost models. Itis configured so that it can be easily deployed as a REST web service using KNIME server or used as a predictor directly in KNIMEAnalytics PlatformExtensions required;KNIME Testing FrameworkRDKit XGBoost Test configurationUser provided dataread modelsMFP2Append input columnsPrediction output Choose dataset Run Tests ifNecessary TestflowConfiguration Table Reader Table Reader RDKit Fingerprint ContainerInput (Table) Column Appender GeneratePredictions ContainerOutput (Table) USER: provide data file here Adds reference dataif the input table isempty Tests the predictionsagainst a goldentable (if necessary) USER: here are theresults! Generate fingerprints and model predictions for the molecules in the table Deploying and validating models as web servicesThis workflow demonstrates a technique to deploy a model prediction workflow as a web service and to validate that it is doing what it'ssupposed to do. The key is the use of a reference set of input data and the Table Difference Checker in the "Run Tests if Necessary"wrapped metanode at the end of the workflow. The model can be validated in KNIME Analytics Platform using the KNIME TestingFramework and, once it has been deployed as a web service using KNIME Server, can be validated by calling the service without anyrows. The workflow itself generates bioactivity predictions for the input molecules using a set of pre-calculated and saved XGBoost models. Itis configured so that it can be easily deployed as a REST web service using KNIME server or used as a predictor directly in KNIMEAnalytics PlatformExtensions required;KNIME Testing FrameworkRDKit XGBoost Test configurationUser provided dataread modelsMFP2Append input columnsPrediction output Choose dataset Run Tests ifNecessary TestflowConfiguration Table Reader Table Reader RDKit Fingerprint ContainerInput (Table) Column Appender GeneratePredictions ContainerOutput (Table)

Download

Get this workflow from the following link: Download

Nodes

01_​Deploying_​and_​Validating_​models_​as_​WebServices consists of the following 38 nodes(s):

Plugins

01_​Deploying_​and_​Validating_​models_​as_​WebServices contains nodes provided by the following 6 plugin(s):