CODEWITHSUNDEEP
pythonpandasnumpyscikit-learnlogistic-regression
Ivan Zhovannik
Ivan Zhovannik

Reputation: 1103

ValueError: Unknown label type: 'unknown'

I try to run following code.

import pandas as pd
import numpy as np
from sklearn.linear_model import LogisticRegression


# data import and preparation
trainData = pd.read_csv('train.csv')
train = trainData.values
testData = pd.read_csv('test.csv')
test = testData.values
X = np.c_[train[:, 0], train[:, 2], train[:, 6:7],  train[:, 9]]
X = np.nan_to_num(X)
y = train[:, 1]
Xtest = np.c_[test[:, 0:1], test[:, 5:6],  test[:, 8]]
Xtest = np.nan_to_num(Xtest)


# model
lr = LogisticRegression()
lr.fit(X, y)

where y is a np.ndarrayof 0s and 1s.

However, I receive the following error:

File "C:\Anaconda3\lib\site-packages\sklearn\linear_model\logistic.py", line >1174, in fit
  check_classification_targets(y)
File "C:\Anaconda3\lib\site-packages\sklearn\utils\multiclass.py", line 172, >in check_classification_targets
  raise ValueError("Unknown label type: %r" % y_type)

ValueError: Unknown label type: 'unknown'

From sklearn documentation, I see that

y : array-like, shape (n_samples,)
Target values (class labels in classification, real numbers in regression)

What is my error?

FYI, y is np.array([0.0, 1.0, 1.0, ..., 0.0, 1.0, 0.0], dtype=object) whose size is (891,).

Upvotes: 93

Views: 215852

Answers (4)

cottontail
cottontail

Reputation: 23489

Target variable in a Logistic Regression can be of type binary (e.g. np.random.randint(2, size=100)) or multiclass (e.g. np.random.randint(3, size=100)). It may be verified using sklearn.utils.multiclass.type_of_target. For example:

from sklearn.utils.multiclass import type_of_target
y = np.random.randint(3, size=100)
type_of_target(y)    # multiclass

If we look at the source code, a target can be unknown type in 3 cases.

  1. An array with more than 2 dimensions, e.g. [[[1, 2]]].
    Solution: Flatten the target variable, e.g. y = np.ravel(y).
  2. A multi-dimensional empty array, e.g. [[]].
    Solution: There is no solution here. The input data is probably not correct.
  3. A non-sparse array of object dtype whose first element is not a string, e.g. np.array([1, 2], dtype=object).
    Solution: Convert to any type other than object before model instantiation; even converting to a list works. All of the following should work: 
    y = y.astype(str)
y = y.astype(float)
y = y.tolist()

Upvotes: 0

Arun Nandam
Arun Nandam

Reputation: 91

I also got a similar type of error. I found out that my target is a non-integer type. After casting my Target variable to the integer type. Boom, error is solved

y = train_data['Y'].astype('int')

Upvotes: 8

Aryan Agarwal
Aryan Agarwal

Reputation: 59

Adding to Miriam ,I also got the similar error but in my case individual elements of y_pred was of type 'np.int32' and individual elements of y was of type 'int'. I solved it by doing:

for i,x in enumerate(y_pred):
    y_pred[i]=x.astype('int')

Upvotes: 1

Miriam Farber
Miriam Farber

Reputation: 19664

Your y is of type object, so sklearn cannot recognize its type. Add the line y=y.astype('int') right after the line y = train[:, 1].

Upvotes: 197

Related Questions