How to choose the features from the given attached heat map & correlation factor for the classification algorithm?

I have 6 different features i.e., ac233fc01403, ac233fc02eaa, ac233fc015f6, ac233fc044aa, ac233fc0455d, ac233fc0451f. From the 6 different features, I should classify in which segment(i.e.,class) the test data is?

How to choose from the given plot, which are the feature has to be considered for the algorithm to achieve the better efficiency?

Heat Map

enter image description here

  • 1
    $\begingroup$ It's hard to say without knowing anything about what you're trying to classify, the size of your data, etc. Anyway it's probably a good idea to use all 6 features. $\endgroup$
    – Erwan
    Sep 26 '19 at 12:35

Why start with feature selection/engineering if you have so few features to begin with?

Best practice should be to include all features/variables in your first draft model and look at the accuracy, quality of resulting segments, etc.

If there is need to improve and to leave out a feature you could then use the metrics of variable importance of your chosen model (e.g. variable importance from random forest) to decide which feature was least relevant and could be discarded for a slimmer model.

In general model performance shouldn't suffer to much from including irrelevant features and therefore selection isn't the same issue as it would be for less robust models.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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