I wanted to use sklearn.pipeline instead of using imblearn.pipeline to incorporate `RandomUnderSampler()'. My original data requires missing value imputation and scaling. Here I have breast cancer data as a toy example. However, it gave me the following error message. I appreciate your suggestions. Thanks for your time!
from numpy.random import seed
seed(12)
from sklearn.datasets import load_breast_cancer
import time
from sklearn.metrics import make_scorer
from imblearn.metrics import geometric_mean_score
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import cross_validate
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import MaxAbsScaler
from imblearn.under_sampling import RandomUnderSampler
gmean = make_scorer(geometric_mean_score, greater_is_better=True)
X, y = load_breast_cancer(return_X_y=True)
start_time1 = time.time()
scoring = {'G-mean': gmean}
LR_pipe = Pipeline([("impute", SimpleImputer(strategy='constant',fill_value= 0)),("scale", MaxAbsScaler()),("rus", RandomUnderSampler()),("LR", LogisticRegression(solver='lbfgs', random_state=0, class_weight='balanced', max_iter=100000))])
LRscores = cross_validate(LR_pipe,X, y, cv=5,scoring=scoring)
end_time1 = time.time()
print ("Computational time in seconds = " +str(end_time1 - start_time1) )
sorted(LRscores.keys())
LR_Gmean = LRscores['test_G-mean'].mean()
print("G-mean: %f" % (LR_Gmean))
Error message:
TypeError: All intermediate steps should be transformers and implement fit and transform or be the string 'passthrough' 'RandomUnderSampler()' (type <class 'imblearn.under_sampling._prototype_selection._random_under_sampler.RandomUnderSampler'>) doesn't
We should import make_pipeline from imblearn.pipeline and not from sklearn.pipeline: make_pipeline from sklearn needs the transformers to implement fit and transform methods. sklearn.pipeline import Pipeline was conflicting with imblearn.pipeline import Pipeline!