手写数字识别问题

数据概要

数据来源于kaggle
其中，训练集是 $785\times 42000$ ，预测集是 $784\times 28000$

实际每个数字是 $28\times 28$ ，由0和1组成的图形所表示

导入必要的包

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.decomposition import PCA
from sklearn.pipeline import Pipeline

读取数据

train=pd.read_csv(r'C:\Users\yep\data\digit-recognizer\train.csv')
test=pd.read_csv(r'C:\Users\yep\data\digit-recognizer\test.csv')

观察样本点

train.head()

	label	pixel0	pixel1	pixel2	pixel3	pixel4	pixel5	pixel6	pixel7	pixel8	...	pixel774	pixel775	pixel776	pixel777	pixel778	pixel779	pixel780	pixel781	pixel782	pixel783
0	1	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
1	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2	1	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
3	4	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
4	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0

5 rows × 785 columns

test.head()

	pixel0	pixel1	pixel2	pixel3	pixel4	pixel5	pixel6	pixel7	pixel8	pixel9	...	pixel774	pixel775	pixel776	pixel777	pixel778	pixel779	pixel780	pixel781	pixel782	pixel783
0	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
1	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
3	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
4	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0

5 rows × 784 columns

train.shape,test.shape

((42000, 785), (28000, 784))

发现数据中只包含（0，1）这组元素，且为稀疏矩阵

target=train['label']
train=train.drop('label',1)

train.head(3)

	pixel0	pixel1	pixel2	pixel3	pixel4	pixel5	pixel6	pixel7	pixel8	pixel9	...	pixel774	pixel775	pixel776	pixel777	pixel778	pixel779	pixel780	pixel781	pixel782	pixel783
0	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
1	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0

3 rows × 784 columns

t1_x,t2_x,t1_y,t2_y=train_test_split(train,target,test_size=0.2,random_state=1)

plt.figure(figsize=(8,8))
for digit_num in range(0,64):
    plt.subplot(8,8,digit_num+1)
    grid_data = train.iloc[digit_num].as_matrix().reshape(28,28)  # reshape from 1d to 2d pixel array
    plt.imshow(grid_data, interpolation = "none", cmap = "bone_r")
    plt.xticks([])
    plt.yticks([])
plt.show()

C:\Users\yep\Anaconda3\lib\site-packages\ipykernel_launcher.py:4: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
  after removing the cwd from sys.path.

建立模型

PCA+KNN

pcn=Pipeline([('pca',PCA(n_components=0.95)),
            ('knn',KNeighborsClassifier())])

pcn.fit(t1_x,t1_y)
pcn.score(t2_x,t2_y)

0.9701190476190477

随机森林

利用交叉验证寻找最优参数

from sklearn.model_selection import GridSearchCV
from sklearn.ensemble import RandomForestClassifier
parameters={'n_estimators':[10,50,100,200,300,400,500,600],'min_samples_split':[2,3,4,5,6,7,8]}
random=RandomForestClassifier(n_jobs=-1)
Gridcv=GridSearchCV(random,parameters,cv=5,n_jobs=-1)
Gridcv.fit(train,target)
Gridcv.best_estimator_

C:\Users\yep\Anaconda3\lib\site-packages\joblib\externals\loky\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.
  "timeout or by a memory leak.", UserWarning





RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
                       max_depth=None, max_features='auto', max_leaf_nodes=None,
                       min_impurity_decrease=0.0, min_impurity_split=None,
                       min_samples_leaf=1, min_samples_split=2,
                       min_weight_fraction_leaf=0.0, n_estimators=600,
                       n_jobs=-1, oob_score=False, random_state=None, verbose=0,
                       warm_start=False)

from sklearn.ensemble import RandomForestClassifier
random=RandomForestClassifier(n_estimators=600,n_jobs=-1,min_samples_split=2,oob_score=True)
random.fit(train,target)
random.oob_score_

0.9676428571428571

梯度提升树

from xgboost import XGBClassifier
xgb=XGBClassifier(n_estimators=500,n_jobs=-1)
xgb.fit(t1_x,t1_y)
xgb.score(t2_x,t2_y)

0.9669047619047619

from lightgbm import LGBMClassifier

lgb1=LGBMClassifier(n_estimators=300)

lgb1.fit(t1_x,t1_y)
lgb1.score(t2_x,t2_y)

0.9777380952380952

LR

from sklearn.linear_model import LogisticRegression
lr=LogisticRegression(C=0.4)
lr.fit(t1_x,t1_y)
lr.score(t2_x,t2_y)

c:\users\yep\miniconda3\lib\site-packages\sklearn\linear_model\logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.
  FutureWarning)
c:\users\yep\miniconda3\lib\site-packages\sklearn\linear_model\logistic.py:469: FutureWarning: Default multi_class will be changed to 'auto' in 0.22. Specify the multi_class option to silence this warning.
  "this warning.", FutureWarning)
c:\users\yep\miniconda3\lib\site-packages\sklearn\svm\base.py:929: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.
  "the number of iterations.", ConvergenceWarning)





0.9096428571428572

模型融合

from sklearn.ensemble import VotingClassifier
eclf=VotingClassifier([('RF',random),('lgb',lgb1),('knn',pcn),('xgb',xgb),('lr',lr)],n_jobs=-1)
eclf.fit(t1_x,t1_y)
eclf.score(t2_x,t2_y)

0.9736904761904762

result=eclf.predict(test)


np.savetxt('result1_1.csv', 
           np.c_[range(1,len(test)+1),result], 
           delimiter=',', 
           header = 'ImageId,Label', 
           comments = '', 
           fmt='%d')

stacking

from mlxtend.classifier import StackingClassifier
clf=StackingClassifier([random,pcn,xgb,lr],lgb1)
clf.fit(train,target)

c:\users\yep\miniconda3\lib\site-packages\sklearn\linear_model\logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.
  FutureWarning)
c:\users\yep\miniconda3\lib\site-packages\sklearn\linear_model\logistic.py:469: FutureWarning: Default multi_class will be changed to 'auto' in 0.22. Specify the multi_class option to silence this warning.
  "this warning.", FutureWarning)
c:\users\yep\miniconda3\lib\site-packages\sklearn\svm\base.py:929: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.
  "the number of iterations.", ConvergenceWarning)





StackingClassifier(average_probas=False,
                   classifiers=[RandomForestClassifier(bootstrap=True,
                                                       class_weight=None,
                                                       criterion='gini',
                                                       max_depth=None,
                                                       max_features='auto',
                                                       max_leaf_nodes=None,
                                                       min_impurity_decrease=0.0,
                                                       min_impurity_split=None,
                                                       min_samples_leaf=1,
                                                       min_samples_split=2,
                                                       min_weight_fraction_leaf=0.0,
                                                       n_estimators=400,
                                                       n_jobs=-1,
                                                       oob_score=True,
                                                       random_st...
                                                  max_depth=-1,
                                                  min_child_samples=20,
                                                  min_child_weight=0.001,
                                                  min_split_gain=0.0,
                                                  n_estimators=300, n_jobs=-1,
                                                  num_leaves=31, objective=None,
                                                  random_state=None,
                                                  reg_alpha=0.0, reg_lambda=0.0,
                                                  silent=True, subsample=1.0,
                                                  subsample_for_bin=200000,
                                                  subsample_freq=0),
                   store_train_meta_features=False, use_clones=True,
                   use_features_in_secondary=False, use_probas=False,
                   verbose=0)

result=eclf.predict(test)


np.savetxt('result1.csv', 
           np.c_[range(1,len(test)+1),result], 
           delimiter=',', 
           header = 'ImageId,Label', 
           comments = '', 
           fmt='%d')

stacking的最终结果为97.3%，voting的最终结果为97%