Friday, September 14, 2012

R data mining IDE, rattle tutorial

Besides the Rcmdr IDE, Rattle is another one IDE for me to do interactive data minding or just data discovery.
installation is easy, just install rattle library, then load the library , and click yes to install the prerequisites components.

image

once done, run rattle() to load the Rattle IDE.

image

like the Rcmdr, you can load/import data. then you can explore the data. Always remember to click Execute to make it happen.
since it’s used for modeling also, you can specify which column is primary key(ident here,) which one is the response value here it is called Target.
image

then click explore tab, to do more exploring. like a summary.

image

click distribution, you can do some plotting.

image

in the correlation, you can tell which variable are correlated to others, blue means positive.

for the sample data, of course, revenue is correlated with request, impression and clicks.
has no relationship with fill_rate.
imagef

for the interactive , you can use plot builder to do a lot plotting. rember to install JDK and rJava library.
image

here is one cool bubble plot,
image

Also you can run transform, clustering.

No comments:

 
Locations of visitors to this page