Hi sklearn developers,

I am a huge R fan and R user but I am trying to use sklearn for predictive
analytics purposes.
I found sklearn is very-well made package. But there are still a few
limitations.

1) there is a lack of statistical terminologies and correct equations in
Documentation.
this makes a hard to understand the meaning of each metrics and how the
method does.

2) sklearn is not really good enough to do descriptive analytics (
explanation purpose)
It would be great if we can see the actual equation (like R) when we do
linear regression etc.

Additionally, I just noticed there is no adjusted R square calculation
function in sklearn.
Thus I quickly made my own adjusted R square function. I am sharing my
function with you.
please add adjusted R square function when you update the version:

def adj_r2_score(model,y,yhat):

    """Adjusted R square — put fitted linear model, y value, estimated y
value in order



        Example:

        In [142]: metrics.r2_score(diabetes_y_train,yhat)

        Out[142]: 0.51222621477934993



        In [144]: adj_r2_score(lm,diabetes_y_train,yhat)

        Out[144]: 0.50035823946984515"""

    from sklearn import metrics

    adj = 1 - float(len(y)-1)/(len(y)-len(model.coef_)-1)*(1 -
metrics.r2_score(y,yhat))

    return adj


Thanks,

 Joon
-- 
 *Joon Lim*
 Master of Science in Analytics
 Department of Industrial Engineering and Management Science
 Northwestern University
------------------------------------------------------------------------------
Precog is a next-generation analytics platform capable of advanced
analytics on semi-structured data. The platform includes APIs for building
apps and a phenomenal toolset for data science. Developers can use
our toolset for easy data analysis & visualization. Get a free account!
http://www2.precog.com/precogplatform/slashdotnewsletter
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to