Alteryx ML model deployment and automation options

[c-lopez]

Alteryx ML model deployment and automation

It is not a controversial statement to say that data science is a highly iterative process, in fact there are several models of the data science process which try to break down problem solving into individual stages, for example the Cross Industry Standard for Data Mining (CRISP-DM)

In this post we would like to focus on the final stage of the process, once you have an ML model how can you put it to work? And what are your options?

Some models are only meant to be used a few times while others are meant to be integrated to Alteryx workflows or as part of a larger data engineering pipeline, in this post we will cover what your options are.

Ad-hoc data scoring through the Alteryx ML platform

This is the first option we will review since it is available to you as soon as you finish working on your ML use case.

In this scoring option you can chose two options:

• Browse
  • If you have incoming data from Trifacta Classic or Designer Cloud and stored it in the dedicated cloud storage for these technologies, then you will find all the CSV files you’ve exported are in this space
    

This means that if you’ve created a workflow or recipe to cleanse data in those products it can be picked up and scored directly and ad-hoc from Alteryx ML

• Import
  • If you have CSV extracts that you would like to quickly score, you can browse the files stored in your PC and import them directly
    
    
    

Ad-hoc data scoring using Alteryx Designer

Another option is to follow the Designer Integration guide to allow Alteryx Designer to interact with Alteryx ML programmatically via the integration tools.

This option is good if you do not run predictive analytics very often and only when needed, if your data is not constantly changing and you are dealing with small volumes this option might be sufficient.

Ad-hoc data scoring using Alteryx Server

This option is similar to the Designer approach but with added functionality. Imagine giving the ability to end users to enter their own data to score, or choose their own data sources to be scored instead of a rigid single source option

Automated data scoring using Alteryx Server

Arguably the most powerful combination of all is to allow Server to score our data in a specific cadence and writing the results to a central location to be used with other analytics across your organization or automating a report that needs to be sent in a regular schedule. The first step is to ensure that your Server administrator has set up the Alteryx ML integration tool, configured it to be used for all users, and shared a connection with you via DCM, administrators can refer to this guide to set up all integrations and Server system settings configu...