1

I perform many logistic regression analyses with different parameters. From time to time I get an annoying message that the iteration limit is reached.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(
/home/arnold/bin/anaconda/envs/vehicles/lib/python3.10/site-packages/sklearn/linear_model/_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

I don't want a message, I have 1000's of them in my project during one run. Is there a way to suppress it?

What i'd like is some indication that something had gone wrong, e.g. raising an exception so that I can check afterwards which analyses were ok and which were wrong. Is there a way to do that?

Arnold
  • 4,578
  • 6
  • 52
  • 91

1 Answers1

1

The message is a custom warning defined sklearn.exceptions. You can suppress it (as noted in the comments), and you can catch it as if it was an error. The catch feature allows you to record the message. That might help you check which analyses were okay afterward.

The following code sample should help you get started. It is based on the python warnings documentation. The with block catches and records the warning produced by the logistic regression.

import warnings
from sklearn import datasets, linear_model,exceptions
import matplotlib.pyplot as plt 

#>>>Start: Create dummy data
blob = datasets.make_blobs(n_samples=100,centers=1)[0]
x = blob[:,0].reshape(-1,1)
# y needs to be integer for logistic regression
y = blob[:,1].astype(int)
plt.scatter(x,y)
#<<<End: Create dummy data

#<<Create logistic regression. set max_iteration to a low number
lr = linear_model.LogisticRegression(max_iter=2)

with warnings.catch_warnings(record=True) as w:
    # Cause all warnings to always be triggered.
    warnings.simplefilter("always")
    # Trigger a warning.
    lr.fit(x,y)

After running the code, you can check the contents of variable w.

print(type(w))
print(w[-1].category)
print(w[-1].message)

Output:

<class 'list'>
<class 'sklearn.exceptions.ConvergenceWarning'>
lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.
Juancheeto
  • 556
  • 4
  • 16