sklearn estimator wrappers for lifeline survival analysis Cox proportional hazard and Aalen Additive models from https://github.com/CamDavidsonPilon/lifelines
pip install git+https://github.com/sashaostr/sklearn-lifelines.gitimport lifelines
from sklearn_lifelines.estimators_wrappers import CoxPHFitterModel
from sklearn.pipeline import make_pipeline
from sklearn.cross_validation import train_test_split
from patsylearn import PatsyTransformer
data = lifelines.datasets.load_dd()
# create sklearn pipeline
coxph_surv_ppl = make_pipeline(PatsyTransformer('un_continent_name + regime + start_year -1', \
return_type='dataframe'),
CoxPHFitterModel(duration_column='duration',event_col='observed'))
#split data to train and test
data_train, data_test = train_test_split(data)
#fit CoxPH model
coxph_surv_ppl.fit(data_train, y=data_train)
#use pipeline to predict expected lifetime
exp_lifetime = coxph_surv_ppl.predict(data_test[0:1])
print 'expected lifetime: ' + str(exp_lifetime)
#or you can extract the model from the pipeline to access more methods
coxmodel = coxph_surv_ppl.named_steps['coxphfittermodel'].estimator
coxmodel.print_summary()> expected lifetime: 4.44003472211
>
> n=1356, number of events=1094
>
> coef exp(coef) se(coef) z p lower 0.95 upper 0.95
>un_continent_name[Africa] 6.664e+00 7.834e+02 nan nan nan nan nan
>un_continent_name[Americas] 7.375e+00 1.596e+03 nan nan nan nan nan
>un_continent_name[Asia] 7.171e+00 1.301e+03 nan nan nan nan nan
>un_continent_name[Europe] 8.190e+00 3.605e+03 nan nan nan nan nan
>un_continent_name[Oceania] 4.183e+00 6.554e+01 nan nan nan nan nan
>regime[T.Military Dict] 8.758e-02 1.092e+00 3.972e-02 2.205e+00 2.747e-02 9.707e-03 1.655e-01 *
>regime[T.Mixed Dem] 4.137e-01 1.512e+00 4.480e-02 9.235e+00 2.588e-20 3.259e-01 5.015e-01 ***
>regime[T.Monarchy] -1.924e-01 8.250e-01 4.860e-02 -3.959e+00 7.539e-05 -2.877e-01 -9.712e-02 ***
>regime[T.Parliamentary Dem] 3.740e-01 1.453e+00 4.931e-02 7.584e+00 3.356e-14 2.773e-01 4.706e-01 ***
>regime[T.Presidential Dem] 3.440e-01 1.411e+00 4.660e-02 7.383e+00 1.546e-13 2.527e-01 4.354e-01 ***
>start_year -4.382e-02 9.571e-01 3.351e-02 -1.308e+00 1.910e-01 -1.095e-01 2.188e-02
>
>Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
>
>Concordance = 0.654