machine_learning_pca.py

from sklearn.datasets import fetch_mldata
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import confusion_matrix

# get training and test sets
x_train, x_test, y_train, y_test = train_test_split(mnist.data, mnist.target, test_size=0.2, random_state=0)
print(x_train.shape)
print(x_test.shape)

# standardization
scaler = StandardScaler()

# calculate scaling parameters from training data set and scale the training data set
scaler.fit(x_train)
x_train = scaler.transform(x_train)

# PCA for training data set, only the first 80%  principal components will be choosed
pca = PCA(0.80)
pca.fit(x_train)
print(pca.n_components_)
x_train = pca.transform(x_train)

# use logistic regresssion for machine learning
clf = LogisticRegression()
clf.fit(x_train, y_train)


# prediction
x_test = scaler.transform(x_test)
x_test = pca.transform(x_test)
y_pred = clf.predict(x_test)

confusion_matrix(y_test, y_pred)
## array([[1277,    0,    2,    2,    4,    7,   13,    0,    7,    0],
##        [   0, 1555,   12,    3,    1,    7,    1,    3,   19,    3],
##        [  10,   12, 1211,   21,   19,    5,   14,   18,   33,    5],
##        [   6,   15,   37, 1260,    0,   46,    4,   11,   32,   16],
##        [   5,    9,   10,    4, 1252,    2,   13,    4,    9,   54],
##        [  14,    7,   14,   50,   20, 1086,   25,    7,   41,   16],
##        [   6,    3,    9,    0,    4,   12, 1357,    1,    3,    2],
##        [   9,    5,   18,    7,   18,    3,    0, 1351,    4,   46],
##        [   9,   33,   15,   46,   10,   41,   14,    4, 1196,   22],
##        [  11,    7,   10,   29,   58,   10,    1,   50,    7, 1236]])