Upload and Deploy a Forecasting Analytic

Steps to upload and deploy a sample Python forecasting analytic on Spark runtime.

Before You Begin

This procedure assumes that the following prerequisites tasks have been completed.
  • You have uploaded the required assets, tags and time series data required for this analytic.
  • Your tenant has been configured to stream to Predix timeseries.
  • You are signed into an OPM tenant with access privileges to manage analytics on Spark runtime.
  • You have downloaded the copy of sparkAnalytic_forecast_train_version_1_0_0.zip.
  • You have downloaded and extracted the contents of sparkAnalytic_forecast_train_io-definitions.zip locally.

Procedure

  1. Upload the analytics template to the catalog.
    Configure the following information for the analytic.
    OptionDescription
    RuntimeSpark
    NameSpark Forecast Train
    OwnerYour Name
    Analytic TypePython
    Type Version2.7.0
    Analytic FilesparkAnalytic_forecast_train_version_1_0_0.zip
    Analytic Version1.0.0
    Primary CategoryForecasting
  2. In the Analytic Template, configure the input definition, constant, and output definition through CSV upload
  3. Add and configure the deployment as follows:
    1. Enter deployment_forecast_train in the Deployment Name field and then select Submit.
    2. In the 1. Asset Selection step, select the asset defined in the analytic, and then select Save.
    3. Select Next to access the 2. I/O Mapping step.
    4. Select the Tag drop-down menu and then, select Add Tags....
    5. In the tag browser, search for the tag in the analytic. As represented in the sample example, search for OO_Tag_Temperature_ID20. After the search displays the tag, drag and drop it onto the input for mapping it.
    6. Select Save and Next to save the I/O Mapping configuration.
    7. In the 3. Schedule step, leave the selection at Streaming for Define how often a new run will be executed option.
    8. Select Time Span between May 23, 2017 to current date.
    9. Leave the Sample Interval at the default value of 1 Minute.
    10. Select Save and then select Deploy.
    Note: The input tag should have data evenly distributed for the Time Span selected during 3. Schedule step in deployment.
    The deployment is saved to the Spark runtime. After successful deployment the status updates to Run Once.

What To Do Next

Visualize the output in the Analysis Forecast App