We have the task to use PI system data as inputs for a machine learning studio.
One of the conditions of the project is not to use the intermediate database in any form. As a start point we consider to use PI Integrator for Business Analytics, Advanced edition.
The input module of machine learning studio supports this type of sources:
Get data that is hosted on a web URL that uses HTTP and that has been provided in the CSV, TSV, ARFF, or SvmLight formats
Get data from distributed storage in Hadoop. You specify the data you want by using the HiveQL language
Get data from Azure SQL Database or from Azure SQL Data Warehouse
Get data that is stored in the Azure table service
Get data that is stored in the Azure blob service
Get data exposed as a feed in OData format
Get data from an on-premises SQL Server database using the Microsoft Data Management Gateway
Get data stored in JSON format in Azure Cosmos DB.
On the other hand, PI Integrator for BA supports these types of target:
Microsoft SQL Server
Azure SQL Server
Azure SQL Data Warehouse
Azure Data Lake Store
Azure Event Hubs
Azure IoT Hub
As I see input source types of Azure Machine Learning Studio do not match with output target types of PI Integrator for BA.
Can you advise any approaches to achieve our goal with these constrain?