[Python] Different answers with imputation of missing values

I have made a classifier using Logistic Regression and for testing it I used the breast cancer dataset available at:

https://archive.ics.uci.edu/ml/datasets/Breast Cancer Wisconsin (Original)

This dataset contains missing values so I have changed those values with three options:

1. Fill them with a value that is below any data from the dataset
2. Use a Imputer with the data frame
3. Use a Imputer, but instead of using the data frame I have used an array of numpy

The issue is that the results from option (1) and (3) are almost similar, but option (2) makes a huge Type II error. My code and results are:

import pandas as pd
import numpy as np
from sklearn import preprocessing, model_selection, linear_model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import Imputer
from sklearn.impute import SimpleImputer
from sklearn.metrics import accuracy_score,confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt

def readfile(name):
df=pd.read_csv(nombre,names=['id', 'clump_thickness','unif_cell_size',
'unif_cell_shape', 'marg_adhesion', 'single_epith_cell_size',
'bare_nuclei', 'bland_chromatin', 'normal_nucleoli','mitoses','class'])
return df

def outlier(df):
#OPTION 1
df.drop(['id'], 1, inplace=True)
df.replace('?', -99999, inplace=True)
return df

def mediaFill(df):
#OPTION 2
df.replace('?',np.NaN,inplace=True)
imp=SimpleImputer(missing_values=np.NaN)
idf=pd.DataFrame(imp.fit_transform(df))
idf.columns=df.columns
idf.index=df.index
return idf

def funcFill():
#OPTION 3
data = np.genfromtxt("breast-cancer-wisconsin.data",delimiter=",")
X = data[:,1:-1]
X[X == '?'] = 'NaN'
imputer = Imputer()
X = imputer.fit_transform(X)
y = data[:, -1].astype(int)
X_train, X_test, y_train, y_test = train_test_split(X, y,test_size=0.2)
lg=linear_model.LogisticRegression(solver="liblinear")
lg.fit(X_train,y_train)
predictions = lg.predict(X_test)
cm=confusion_matrix(y_test,predictions)
print(cm)
score = lg.score(X_test, y_test)
print(score)

def LogisticFunc(df):
X = np.array(df.drop(['class'],1))
y = np.array(df['class'])
labels=[2,4]
X_train, X_test, y_train, y_test = model_selection.train_test_split(X, y, test_size = 0.2)
clf = linear_model.LogisticRegression(solver="liblinear")
clf.fit(X_train,y_train)
y_pred = clf.predict(X_test)
conf = confusion_matrix(y_test, y_pred, labels)
print (conf)
print (accuracy_score(y_pred,y_test))

def main():
file="breast-cancer-wisconsin.data"
df=readfile(file)
df=outlier(df)
LogisticFunc(df)

df=readfile(file)
df=mediaFill(df)
LogisticFunc(df)

df=readfile(file)
funcFill()

if __name__=="__main__":
main()


My results are: Option 1:

[[97 1]
[ 2 40]]


Option 2:

[[89 0]
[51 0]]


Option 3:

[[92 2]
[ 2 44]]


Why does it differ too much Option 2?

Continue reading...