Options
Feature ranking
Contributor I
Hi all,
I apologise if this has come up before, I did a quick search but couldn't find anything specifically addressing my issue.
I have a dataset which consists of a number of variables: continuous, date, multinomial and binomial. The data label is binomial.
There are a number of examples and tutorials for running subset selection in order to find the most informative variables in the data. However, I would like to do something more simple to begin, merely rank the variables (i.e. rank the features by a given metric).
Is there an easy way to do this using an operator? I.e. to feed my dataset into a method, and get an ordered list of variables out? Of course, the added complication is that I have different types of variable (i.e. continuous vs. categorical), but I suppose ranking by p.value would allow me to fuse the outputs.
Thanks in advance for any help you can give
I apologise if this has come up before, I did a quick search but couldn't find anything specifically addressing my issue.
I have a dataset which consists of a number of variables: continuous, date, multinomial and binomial. The data label is binomial.
There are a number of examples and tutorials for running subset selection in order to find the most informative variables in the data. However, I would like to do something more simple to begin, merely rank the variables (i.e. rank the features by a given metric).
Is there an easy way to do this using an operator? I.e. to feed my dataset into a method, and get an ordered list of variables out? Of course, the added complication is that I have different types of variable (i.e. continuous vs. categorical), but I suppose ranking by p.value would allow me to fuse the outputs.
Thanks in advance for any help you can give
0
Answers
The approach you are looking for is called "Filter" in the area of feature subset selection. Rapidminer provides a good amount of operators for this. See lefthandside Modelling -> Attribute Weighting.
greetings,
steffen