I'm trying to build a model of Machine Learning for the first time, and I'm looking for some guidelines.
My final purpose is to identify customers who are planning to flight abroad.
From my existing data, I identified customers who flew in the past, and I've flagged some potential variables (features) that might indicate that one is about to flight abroad.

My questions are:
1. What techniques should I use, in order to choose the right features for my model?
2. What approaches (models) would you use in such a problem?

because this is the first time I'm experiencing ML, I'll be happy to hear any further suggestions you have to give me.

Your help would be appreciated.

  • $\begingroup$ There's not much to go on here, but you might want to try using a random forest. I think of them as a sort of Swiss Army knife of machine learning. They do an okay job at lots of different kinds of problems, and in the absence of much information they're a good place to start. Once you have a model, you can try omitting different features to see if the performance of your model improves, decreases, or remains the same. That might help you to identify extraneous features. $\endgroup$ Commented Jul 21, 2016 at 23:01

1 Answer 1


Which algo you should use is up to debate and sometimes comes down to the person

If you're going for accuracy then random forests are a great place to start (as mentioned in the comments) but personally I like logistic regressions as my starting place for this type of problem as they have decent accuracy and are comparatively explainable.

But if this is your first Machine Learning problem I'd recommend doing some wider reading first on how to formulate and approach machine learning problems.

It'll give you the perspective to decide which model to use on your own. Furthermore, you'll likely need to explain your model choices further down the line and using the "some dude from stack overflow told me to use it" excuse won't cut the mustard.


How to approach a ML problem

Intro to ML in Python


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.