下載數據到本地，加載數據
import numpy as np
import csv
import pandas as pd
def load_data(csv):
lines = csv.reader(open(csv))
l = []
for line in lines:
l.append(line)
return l
l = load_data('train.csv')
l = np.array(l[1:], dtype=float)
train = l[1:,1:]
label = l[1:,0]
a = pd.DataFrame(train)
# 二值化，不影響數字顯示
a[a > 1] = 1
l = load_data('test.csv')
test = np.array(l[1:], dtype=float)
a = pd.DataFrame(test)
# 二值化，不影響數字顯示
a[a > 1] = 1
import seaborn as sns
%matplotlib inline
df = pd.DataFrame(np.hstack((train, label[:,None])),
columns = range(train.shape[1]) + ["class"])
plt.figure(figsize=(8, 6))
_ = sns.heatmap(df.corr(), annot=False)
使用LogisticRegression分類
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import accuracy_score
X_train = train
y_train = label
sc = StandardScaler()
sc.fit(X_train)
X_train_std = sc.transform(X_train)
lr = LogisticRegression(C=10000.0, random_state=0)
lr.fit(X_train_std, y_train)
看下訓練集誤差，誤差大約6.7954%，這個還是蠻大的
y_pred = lr.predict(X_train_std)
print('Misclassified samples: %.8f' % ((y_train != y_pred).sum()/float(len(y_train))))
OUT:Misclassified samples: 0.06795400
對測試集預測
X_test = test
X_test_std = sc.transform(X_test)
'''sc.scale_標準差, sc.mean_平均值, sc.var_方差'''
y_pred = lr.predict(X_test_std)
print y_pred
OUT: [ 2. 0. 9. ..., 3. 9. 2.]
提交kaggle,得分排名比較靠后
畫一個像素圖片數字，第二個圖片，上面預測是0
from PIL import Image
import numpy as np
import matplotlib.pyplot as plt
import csv
test1 = test[1]
test2 = []
for el in test1:
test2.append([0,0,el])
img = np.array(test2)
print img.shape
img1 = img.reshape((28,28,3))
plt.figure("dog")
plt.imshow(img1)
plt.axis('off')
plt.show()
