如何对测试数据使用逻辑回归

Question

I am using Logistic Regression on my Titanic model and PyCharm is asking me to pass DataFrames with bool values only:

Traceback (most recent call last):
  File "C:/Users/security/Downloads/AP/Titanic-Kaggle/TItanic-Kaggle.py", line 29, in <module>
    predictions = logReg.predict(test[test_data])
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\pandas\core\frame.py", line 2914, in __getitem__
    return self._getitem_frame(key)
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\pandas\core\frame.py", line 3009, in _getitem_frame
    raise ValueError('Must pass DataFrame with boolean values only')
ValueError: Must pass DataFrame with boolean values only

I don't understand why because the exact same features were used on Logistic Regression while training the model and it was well received then. Here is my code (ignore the code repetition. That's a problem I'm going to tackle after):

import pandas as pd
import warnings
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split

warnings.filterwarnings("ignore", category=FutureWarning)

train = pd.read_csv("https://raw.githubusercontent.com/oo92/Titanic-Kaggle/master/train.csv")
test = pd.read_csv("https://raw.githubusercontent.com/oo92/Titanic-Kaggle/master/test.csv")

train['Sex'] = train['Sex'].replace(['female', 'male'], [0, 1])
train['Embarked'] = train['Embarked'].replace(['C', 'Q', 'S'], [1, 2, 3])
train['Age'].fillna(train.groupby('Sex')['Age'].transform("median"), inplace=True)
train['HasCabin'] = train['Cabin'].notnull().astype(int)
train['Relatives'] = train['SibSp'] + train['Parch']
train_data = train[['Pclass', 'Sex', 'Relatives', 'Fare', 'Age', 'Embarked', 'HasCabin']]
x_train, x_validate, y_train, y_validate = train_test_split(train_data, train['Survived'], test_size=0.22, random_state=0)

test['Sex'] = test['Sex'].replace(['female', 'male'], [0, 1])
test['Embarked'] = test['Embarked'].replace(['C', 'Q', 'S'], [1, 2, 3])
test['Age'].fillna(test.groupby('Sex')['Age'].transform("median"), inplace=True)
test['HasCabin'] = test['Cabin'].notnull().astype(int)
test['Relatives'] = test['SibSp'] + test['Parch']
test_data = test[['Pclass', 'Sex', 'Relatives', 'Fare', 'Age', 'Embarked', 'HasCabin']]

logReg = LogisticRegression()
logReg.fit(x_train, y_train)

predictions = logReg.predict(test[test_data])
submission = pd.DataFrame({'PassengerId': test['PassengerId'], 'Survived': predictions})

filename = 'Titanic-Submission.csv'
submission.to_csv(filename, index=False)

Specifically, Python takes issue with this snippet:

test_data = test[['Pclass', 'Sex', 'Relatives', 'Fare', 'Age', 'Embarked', 'HasCabin']]

...

predictions = logReg.predict(test[test_data])

UPDATE

I've changed my predictions variable to this:

predictions = logReg.predict(test_data)

And now this is my stacktrace:

Traceback (most recent call last):
  File "C:/Users/security/Downloads/AP/Titanic-Kaggle/TItanic-Kaggle.py", line 29, in <module>
    predictions = logReg.predict(test_data)
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\sklearn\linear_model\base.py", line 281, in predict
    scores = self.decision_function(X)
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\sklearn\linear_model\base.py", line 257, in decision_function
    X = check_array(X, accept_sparse='csr')
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\sklearn\utils\validation.py", line 573, in check_array
    allow_nan=force_all_finite == 'allow-nan')
  File "C:\Users\security\Anaconda3\envs\TItanic-Kaggle.py\lib\site-packages\sklearn\utils\validation.py", line 56, in _assert_all_finite
    raise ValueError(msg_err.format(type_err, X.dtype))
ValueError: Input contains NaN, infinity or a value too large for dtype('float64').

Which means that my feature selection/engineering for the test data does not go through

Answer 1

You have a NaN value in the Fare column which you don't take care of. Replacing it similar as you do to Age takes care of the problem. Is this the best solution for the model? That's a different argument but this gets rid of the problem.

train['Fare'].fillna(train.groupby('Sex')['Age'].transform("median"), inplace=True)
test['Fare'].fillna(train.groupby('Sex')['Age'].transform("median"), inplace=True)

Answer 2

Predictions with x_validate work no problem. Try:

>>> predictions = logReg.predict(x_validate)

So there must be something wrong with test_data . Get some information on the dataframes and compare:

>>> x_validate.info(verbose=True)                                                                                                                                                          
<class 'pandas.core.frame.DataFrame'>
Int64Index: 197 entries, 495 to 45
Data columns (total 7 columns):
Pclass       197 non-null int64
Sex          197 non-null int64
Relatives    197 non-null int64
Fare         197 non-null float64
Age          197 non-null float64
Embarked     197 non-null int64
HasCabin     197 non-null int64
dtypes: float64(2), int64(5)
memory usage: 12.3 KB

>>> test_data.info(verbose=True)                                                                                                                                                           
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 418 entries, 0 to 417
Data columns (total 7 columns):
Pclass       418 non-null int64
Sex          418 non-null int64
Relatives    418 non-null int64
Fare         417 non-null float64
Age          418 non-null float64
Embarked     418 non-null int64
HasCabin     418 non-null int64
dtypes: float64(2), int64(5)
memory usage: 22.9 KB

Looks like there's a NaN here:

Fare         417 non-null float64