Clever Geek Handbook

Is LASSO regression implemented in Statsmodels?

I would love to use LASSO linear regression in statsmodels to be able to use a "formula" formula to write the model, which would save me quite a bit of coding time when working with many categorical variables and their interactions. However, it looks like it hasn't been implemented in statistics models yet?

+9


source to share


2 answers


Lasso is really implemented in statsmodels. The documentation is in the url below:

http://www.statsmodels.org/dev/generated/statsmodels.regression.linear_model.OLS.fit_regularized.html



To be precise, the implementation in statsmodel has both L1 and L2 regularizations, and their relative weight is specified by the L1_wt parameter. You should take a look at the formula below to make sure you are doing exactly what you want.

In addition to implementing the elastic network, statsmodels also implements the square root Lasso method.

+12


source


You can use Patsy with scikit-learn to get the same results that you can get with formula notation in statsmodels. See the following code:

from patsy import dmatrices

# create dummy variables, and their interactions
y, X = dmatrices('outcome ~ C(var1)*C(var2)', df, return_type="dataframe")
# flatten y into a 1-D array so scikit-learn can understand it
y = np.ravel(y)


and now I can use any model implemented in scikit-learn with the usual notation having X as independent variables and y as dependent variables.

0


source






More articles:


All Articles