Classification of spread Data

Hi everyone.

I have data consisting of two classes with a total of eight features. So far I have tried solving classification problems using standard datasets like in example IRIS.

In this cases all the features were nicely distributed so that classification was not a very difficult task.

In my own dataset the distribution of the classes is a somewhat different:

As Iam new to machine Learning Iam interested to know what algorithm would be appropriate in that case and what the best way is to select my features. In my case it seems like the features are highly correlated and linearly dependent. Is that something preferable in machine Learning?

Any advice would be really great!

Cheers,

Mike

sad

[KSSV]

For ML correlation is important. You may try fitcsvm

[MikeSv]

Thanks for the comment! I tried to use a CART algorithm which seems to work quite nicely, though Iam not sure if its the correct one to choose for this kind of data.

Answer 1

[Image Analyst]

The Classification Learner app (on the Apps tab of the tool ribbon) will let you try out all the different methods and show you the error (misclassification rate) for each method.

You can attach a .mat file with your table variable in it if you want people to try things with it.