OCTA数据集的交叉验证划分为训练集和测试机

https://blog.csdn.net/weixin_39531594/article/details/110804636

1、从excel数据中读取

import pandas as pd

file_path = r'OCT500-Text labels.xlsx'   # r对路径进行转义，windows需要
raw_data = pd.read_excel(file_path, header=0)  # header=0表示第一行是表头，就自动去除了
print(raw_data.values.shape)
raw_data.head()
raw_data.number.value_counts()

2、取出输入和输出数据

y = raw_data.number.values
X = raw_data.ID.values #raw_data.drop(['number'], axis = 1)
print(X)
print(y)

3、采用k-fold划分训练集和测试机

from sklearn.model_selection import StratifiedKFold
from iterstrat.ml_stratifiers import MultilabelStratifiedKFold
from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit

#kfold = StratifiedKFold(n_splits=5, shuffle=True, random_state=1) #StratifiedKFold, MultilabelStratifiedKFold
kfold = StratifiedKFold(n_splits=5,random_state=0,shuffle=True)

for train_ix, test_ix in kfold.split(X, y):
    print('train_ix is ************',train_ix.shape)
    print('test_ix is ************',test_ix.shape)
    print('_____________________________________')
    # select rows
    train_X = X[train_ix]
    test_X  = X[test_ix]
    train_y = y[train_ix]
    test_y  = y[test_ix]
    # summarize train and test composition
    train_0, train_1,train_2, train_3, train_4, train_5, train_6 = len(train_y[train_y==0]), len(train_y[train_y==1]), len(train_y[train_y==2]), len(train_y[train_y==3])\
                        , len(train_y[train_y==4]), len(train_y[train_y==5]), len(train_y[train_y==6])
    test_0, test_1,test_2, test_3, test_4, test_5, test_6 = len(test_y[test_y==0]), len(test_y[test_y==1]), len(test_y[test_y==2])\
                        , len(test_y[test_y==3]), len(test_y[test_y==4]), len(test_y[test_y==5]), len(test_y[test_y==6])
    print('>Train: 0=%d, 1=%d,2=%d,3=%d,4=%d,5=%d,6=%d;\
        Test: 0=%d, 1=%d ,2=%d,3=%d,4=%d,5=%d,6=%d' % (train_0, train_1, train_2, train_3, train_4, train_5, train_6,\
                                                    test_0, test_1,test_2, test_3, test_4, test_5, test_6 ))

结果如下

train_ix is ************ (400,)
test_ix is ************ (100,)
_____________________________________
>Train: 0=200, 1=39,2=52,3=12,4=12,5=8,6=77;        Test: 0=51, 1=10 ,2=12,3=4,4=2,5=2,6=19
train_ix is ************ (400,)
test_ix is ************ (100,)
_____________________________________
>Train: 0=201, 1=40,2=51,3=13,4=11,5=8,6=76;        Test: 0=50, 1=9 ,2=13,3=3,4=3,5=2,6=20
train_ix is ************ (400,)
test_ix is ************ (100,)
_____________________________________
>Train: 0=201, 1=39,2=51,3=13,4=11,5=8,6=77;        Test: 0=50, 1=10 ,2=13,3=3,4=3,5=2,6=19
train_ix is ************ (400,)
test_ix is ************ (100,)
_____________________________________
>Train: 0=201, 1=39,2=51,3=13,4=11,5=8,6=77;        Test: 0=50, 1=10 ,2=13,3=3,4=3,5=2,6=19
train_ix is ************ (400,)
test_ix is ************ (100,)
_____________________________________
>Train: 0=201, 1=39,2=51,3=13,4=11,5=8,6=77;        Test: 0=50, 1=10 ,2=13,3=3,4=3,5=2,6=19

版本2

from sklearn.model_selection import StratifiedKFold,StratifiedShuffleSplit
from iterstrat.ml_stratifiers import MultilabelStratifiedKFold
from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit


from sklearn.model_selection import KFold

#kfold = StratifiedKFold(n_splits=5, shuffle=True, random_state=1) #StratifiedKFold, MultilabelStratifiedKFold
kfold = StratifiedKFold(n_splits=5,random_state=0,shuffle=True)

ss = StratifiedShuffleSplit(n_splits=2, test_size=0.5, random_state=0) 
ss = KFold(n_splits=2, shuffle=True, random_state=1)

for train_ix, test_ix in kfold.split(X, y):
    print('train_ix is ************',train_ix.shape)
    print('test_ix is ************',test_ix.shape)
    print('_____________________________________')
    # select rows
    train_X = X[train_ix]
    test_X  = X[test_ix]
    train_y = y[train_ix]
    test_y  = y[test_ix]
    
    # summarize train and test composition
    train_0, train_1,train_2, train_3, train_4, train_5, train_6 = len(train_y[train_y==0]), len(train_y[train_y==1]), len(train_y[train_y==2]), len(train_y[train_y==3])\
                        , len(train_y[train_y==4]), len(train_y[train_y==5]), len(train_y[train_y==6])
    test_0, test_1,test_2, test_3, test_4, test_5, test_6 = len(test_y[test_y==0]), len(test_y[test_y==1]), len(test_y[test_y==2])\
                        , len(test_y[test_y==3]), len(test_y[test_y==4]), len(test_y[test_y==5]), len(test_y[test_y==6])
    
    print('>Train: 0=%d, 1=%d,2=%d,3=%d,4=%d,5=%d,6=%d;\
        Test: 0=%d, 1=%d ,2=%d,3=%d,4=%d,5=%d,6=%d' % (train_0, train_1, train_2, train_3, train_4, train_5, train_6,\
                                                    test_0, test_1,test_2, test_3, test_4, test_5, test_6 ))
    
    for val_index, test_index in ss.split(test_X, test_y):
        print(val_index.shape, test_index.shape)
    test_X = X[test_index]
    test_y = y[test_index]
    
    val_X = X[val_index]
    val_y = y[val_index]
    
    test_0, test_1,test_2, test_3, test_4, test_5, test_6 = len(test_y[test_y==0]), len(test_y[test_y==1]), len(test_y[test_y==2])\
                        , len(test_y[test_y==3]), len(test_y[test_y==4]), len(test_y[test_y==5]), len(test_y[test_y==6])
    
    val_0, val_1,val_2, val_3, val_4, val_5, val_6 = len(val_y[val_y==0]), len(val_y[val_y==1]), len(val_y[val_y==2])\
                        , len(val_y[val_y==3]), len(val_y[val_y==4]), len(val_y[val_y==5]), len(val_y[val_y==6])
    print('>Train: 0=%d, 1=%d,2=%d,3=%d,4=%d,5=%d,6=%d;\
        Test: 0=%d, 1=%d ,2=%d,3=%d,4=%d,5=%d,6=%d:\
        Val: 0=%d, 1=%d ,2=%d,3=%d,4=%d,5=%d,6=%d' % (train_0, train_1, train_2, train_3, train_4, train_5, train_6,\
                                                      test_0, test_1,test_2, test_3, test_4, test_5, test_6,\
                                                    val_0, val_1,val_2, val_3, val_4, val_5, val_6 ))

---

OCTA数据集两分类的结果

import pandas as pd

file_path = r'./OCT300.xlsx'   # r对路径进行转义，windows需要
raw_data = pd.read_excel(file_path, header=0)  # header=0表示第一行是表头，就自动去除了
print(raw_data.values.shape)
raw_data.head()

y = raw_data.label.values
X = raw_data.ID.values #raw_data.drop(['number'], axis = 1)
#print(X)
#print(y)

from sklearn.model_selection import StratifiedKFold
from iterstrat.ml_stratifiers import MultilabelStratifiedKFold
from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit
from sklearn.model_selection import train_test_split
import numpy as np

#kfold = StratifiedKFold(n_splits=5, shuffle=True, random_state=1) #StratifiedKFold, MultilabelStratifiedKFold
kfold = StratifiedKFold(n_splits=5,random_state=0,shuffle=True)

i = 0
for train_val_ix, test_ix in kfold.split(X, y):
    i +=1
    print('_____________________________________')
    # select rows
    train_val_X = X[train_val_ix]
    train_val_y = y[train_val_ix]
    test_X  = X[test_ix]
    test_y  = y[test_ix]
    
    train_X,val_X,train_y,val_y= train_test_split(train_val_X,train_val_y,test_size=0.125,random_state = 20,shuffle = True)
    print('train_val_X is ************',train_val_X)
    print('train_val_ix ix is ************',train_val_ix)
    print('test is ************',test_X)
    print('test_ix is ************',test_ix)
    
    train_X = train_X-10000-1
    val_X = val_X-10000-1
    test_X = test_X -10000-1
    
    
    if i==1:
        print('train_val is ************',)
        print('val_X is ************',val_X-10000)
        print('test_X is ************',test_X-10000)
        
        np.savetxt('./OCT300/OCT300_train_Kfold1.txt',train_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_train_y_Kfold1.txt',train_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_val_Kfold1.txt',val_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_val_y_Kfold1.txt',val_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_test_Kfold1.txt',test_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_test_y_Kfold1.txt',test_y,fmt='%d')
    
    elif i==2:
        
        np.savetxt('./OCT300/OCT300_train_Kfold2.txt',train_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_train_y_Kfold2.txt',train_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_val_Kfold2.txt',val_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_val_y_Kfold2.txt',val_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_test_Kfold2.txt',test_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_test_y_Kfold2.txt',test_y,fmt='%d')
        
    elif i==3:
        
        np.savetxt('./OCT300/OCT300_train_Kfold3.txt',train_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_train_y_Kfold3.txt',train_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_val_Kfold3.txt',val_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_val_y_Kfold3.txt',val_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_test_Kfold3.txt',test_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_test_y_Kfold3.txt',test_y,fmt='%d')
    
    elif i==4:
        
        np.savetxt('./OCT300/OCT300_train_Kfold4.txt',train_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_train_y_Kfold4.txt',train_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_val_Kfold4.txt',val_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_val_y_Kfold4.txt',val_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_test_Kfold4.txt',test_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_test_y_Kfold4.txt',test_y,fmt='%d')
    
    elif i==5:
        
        np.savetxt('./OCT300/OCT300_train_Kfold5.txt',train_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_train_y_Kfold5.txt',train_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_val_Kfold5.txt',val_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_val_y_Kfold5.txt',val_y,fmt='%d')
        
        np.savetxt('./OCT300/OCT300_test_Kfold5.txt',test_X,fmt='%d')
        #np.savetxt('./OCT300/OCT300_test_y_Kfold5.txt',test_y,fmt='%d')
        
    # summarize train and test composition
    train_0, train_1 = len(train_y[train_y==0]), len(train_y[train_y==1])
    val_0, val_1 = len(val_y[val_y==0]), len(val_y[val_y==1])
    test_0, test_1 = len(test_y[test_y==0]), len(test_y[test_y==1])
    print('>Train: 0=%d, 1=%d; Val: 0=%d, 1=%d; Test: 0=%d, 1=%d; ' % (train_0, train_1, val_0, val_1, test_0, test_1 ))