Part 2:

Kaggle Competition

Automated Machine Learning

(Written January 2021)


I had come across this Intro to AutoML and thought it would be fun to try out Google's AutoML on the Kaggle dataset.

I was intrigued that it claimed to take care of most of the steps!



In short, my attempt did not yield a successful result.

Through a bit of trial and error I learned I did not have enough data to meet Google's minimum of 1000 rows for training of the model.

I knew I would try AutoML in the future with different dataset but I was hell bent on making this work now...so I fabricated additional data which would be considered heretic in this field. I do NOT recommend this and in my defence was purely for my own educational purpose.


I found it a valuable experience trying my first competition and experimenting with different Machine Learning models.

With Google AutoML (and knowing Amazon and Microsoft Azure also have similar offering), the trend is to make ML accessible for all.

Its amazing to see the technology evolving and sobering to know that the skills I learn today will be different tomorrow. All the more incentive to understand the fundamentals and science behind it.