Free AutoML Tools For Every Need

Read time: 5 minutes

Hey there,

Last week I got a lot of feedback on my Auto ML Crash course. One of the most popular questions was:

Is there a free Auto ML tool to start with?

Well, I’m glad you asked!

In today's edition I'll introduce you to three different AutoML platforms that cost you $0 to use, and compare their pros and cons.

So let's go!

Want to improve your analytics skills? Join my upcoming Business Analytics Bootcamp - LIVE with O'Reilly! 

🎁 Use this promo code to get 30 days free access to the O'Reilly platform (incl. my bootcamp)! 🎁

Starting free isn't always cheap

As I mentioned last week, I like to use Azure ML Studio for several reasons:

1. You can build Auto ML workflows with or without code.

2. It's possible to build end-to-end workflows and easily scale things up.

3. If you're on a Microsoft stack, especially Power BI, Azure seems like a natural fit.

That said, Azure might be too much if:

• You're on your own and just want to try something quick.

• You have no previous experience with cloud platforms.

• You don't want to spend a dime until you can prove the value.

In these cases, starting with Azure (or any other cloud platform for this sake) might not be the best idea.

Here are three Auto ML alternatives you might want to consider instead:

3 Free Auto ML tools for different needs

TPOT: Open source Auto ML on top of sklearn

TPOT, or the Tree-based Pipeline Optimization Tool, is a Python-based Automated Machine Learning tool that streamlines the pipeline selection process using genetic programming.

TPOT explores thousands of options to find the optimal pipeline for your data. Once completed, it provides you with the Python code for the chosen pipeline, which can be further modified if desired. Built on scikit-learn, it generates familiar code for those familiar with the library.

TPOT requires previous machine learning expertise and coding skills to use it effectively. It does not cover model deployment.

Pros:

• Open source

• Built on top of sklearn

• Supports feature engineering

Cons:

• Previous ML expertise required

• Coding skills required

• No model deployment

H2O: The (kind of) open-source AutoML framework

H2O.ai isn't actually an AutoML framework per se but a company currently carrying a $1bn+ valuation. They offer an end-to-end machine learning platform at enterprise scale, but at the same time they are giving away some of their core frameworks as open source, which have grown quite a fanbase (10k+ github stars).

Having been an avid H2O.ai user myself, I can say that it's definitely a great tool with even greater capabilities, but it has a steep learning curve. Although it was initially designed for programmers, there is a graphical user interface called Flow that you can try out if you don't feel like hacking together scripts.

The H2O AutoML framework is available for both Python and R, making it a good (and maybe the only?) choice for R folks. Even though there are many great tutorials available, it can take a while to understand everything in the beginning.

Don't expect to get started in a few minutes. But once you do, it can be blazing fast:

Pros:

• Supports R and Python

• Offers a GUI

• End-to-end platform

• Enterprise-grade if needed

Cons:

• Steep learning curve

• Technical / coding skills recommended

• Feature engineering only part of paid product (Driverless AI)

Obviously.ai: NoCode AutoML platform with a free plan

Obviously.ai is the only Nocode AutoML platform on this list. Why? Because I wanted to give you free tools to get started, and the choice of easy-to-use, production-ready Nocode AutoML platforms with a "real" free tier is limited, to say the least.

Obviously.ai makes getting started with AutoML amazingly easy. You can upload or connect to a dataset, do some basic feature engineering, train a model, and deploy it as an API or Zapier integration with one click - all in probably less than 10 minutes.

A nice perk is that each model automatically comes with a hosted web app where you can enter prediction parameters and get the results through a pretty user front end.

Recently, they've also added a free academy where you can deepen your AI/ML knowledge. Give it a try!

Pros:

• Speed / easy to learn

• No coding skills required

• Free deployment (up to 1,200 predictions)

• Free university

Cons:

• Limited feature engineering

• Lack of customization

Conclusion

Don't be fooled - there are far more options than are mentioned here in this newsletter. But I don't want to overcomplicated things.

I think four options in total are enough for now:

• Azure ML Studio: Recommended if you're on the Microsoft stack and not afraid of a full-fledged cloud platform.

• TPOT: Recommended if you like coding and have some experience with Scikit-Learn.

• H2O.ai: Recommended if you're a programmer who might want to use enterprise features.

• Obviously.ai: Recommended if you want to get started fast without coding.

That's it for today!

As always, thanks for reading.

I'll be back next week with a new practical AI for BI use case.

Did you find today's issue helpful? Hit reply and let me know!

See you next Friday,

Tobias

If you liked this content then check out my book AI-Powered Business Intelligence (O’Reilly). You can read it in full detail here: https://www.aipoweredbi.com