[R] machine learning goal (new to R )

Sarah Goslee sarah.goslee at gmail.com
Tue Jan 3 20:11:20 CET 2017


There are a lot of machine learning options in R:
https://cran.r-project.org/web/views/MachineLearning.html

It sounds like you need to back up a step, and do some reading on the
statistical underpinnings of machine learning before you try to figure
out how to implement a particular method.

There are an enormous number of references online, from brief articles
to full courses. Here's one possible starting point:
https://statweb.stanford.edu/~tibs/ElemStatLearn/

The options are far too complex and numerous for anyone here to be
able to tell you the "right method" to use.

Sarah

On Tue, Jan 3, 2017 at 12:15 PM, Chuck Snell
<chuck.snell.email at gmail.com> wrote:
> Hello,
>
> I am new to R, a computer programmer friend of mine recommended R for a
> project I have on my plate.
>
> (He is not a R guy but knows I need to consider it for the problem I
> described to him)
>
> Frist, I have plenty of data
>
> I have been doing this task with regression models but was asked to try to
> improve my accuracy.
>
> I am forecasting an "output" which is numerical based upon forecasted
> weather.
>
> for extreme weather and stable weather my regression does decent. Meaning,
> really cold and hot weather that has been cold or hot for a while.
>
> What I miss is when things change, meaning if we have had mild weather then
> a sudden change, intuitively we know things won't behave as if it had been
> cold (or hot) for the last week or so but my regression obviously does not
> consider the "history" or patterns.
>
> What was suggested to me was consider some machine learning to identify the
> patterns and so forth.
>
> I have R installed and started searching around the libraries - seems
> overwhelming.
>
> I have found an example of machine learning for R that did "categories" -
> maybe of flowers not sure.
>
> What I need is not categories but a number for an estimate / forecast,
>
> Can you recommend some routines / libraries / techniques to consider?
>
> Thanks
>
-- 
Sarah Goslee
http://www.functionaldiversity.org



More information about the R-help mailing list