python - 从 Numpy 数组中删除单引号

Question

我有一个 Numpy 数组，其中元素用单引号引起来，我想将数据类型转换为 float 。

array = 
 [['20150101' '0.12']
 ['20150102' '0.42']
 ['20150103' '0.12']
 ['20150104' '0.46']
 ['20150105' '0.14']
 ['20150106' '0.1']
 ['20150107' '0.27']
 ['20150108' '0.03']
 ['20150109' '0.16']
 ['20150110' '0.29']
 ['20150111' '0.32']
 ['20150112' '0.16']]

我尝试过:

values = array.item().split(' ')
new_array = np.asarray(values, dtype='float')

但我收到 ValueError: can only conversion an array of size 1 to a Python scalar。我希望输出如下所示:(没有单引号)

new_array = 
     [[20150101 0.12]
     [20150102 0.42]
     [20150103 0.12]
     [20150104 0.46]
     [20150105 0.14]
     [20150106 0.10]
     [20150107 0.27]
     [20150108 0.03]
     [20150109 0.16]
     [20150110 0.29]
     [20150111 0.32]
     [20150112 0.16]]

是否有一个 numpy 函数可以让我删除单引号？

Answer 1

您显示的是一个带有字符串dtype的二维数组，我可以使用它重新创建:

In [420]: arr = np.array([['20150101', '0.12'], 
     ...:  ['20150102', '0.42'], 
     ...:  ['20150103', '0.12'], 
     ...:  ['20150104', '0.46']])                                               
In [421]:                                                                       
In [421]: arr                                                                   
Out[421]: 
array([['20150101', '0.12'],     # the repr display
       ['20150102', '0.42'],
       ['20150103', '0.12'],
       ['20150104', '0.46']], dtype='<U8')
In [422]: print(arr)                        # the str display                                                
[['20150101' '0.12']
 ['20150102' '0.42']
 ['20150103' '0.12']
 ['20150104' '0.46']]

引号反射(reflect)了数组的基本性质；它们不仅仅是数组的偶然部分。

转换为 float 据类型数组:

In [423]: arr.astype(float)                                                     
Out[423]: 
array([[2.0150101e+07, 1.2000000e-01],
       [2.0150102e+07, 4.2000000e-01],
       [2.0150103e+07, 1.2000000e-01],
       [2.0150104e+07, 4.6000000e-01]])

广泛的值需要科学记数法。第一列本身显示为:

In [424]: _[:,0]                                                                
Out[424]: array([20150101., 20150102., 20150103., 20150104.])

我可以混合使用整数和 float :

In [426]: arr1 = np.zeros((4,), dtype='i,f')                                    
In [427]: arr1                                                                  
Out[427]: 
array([(0, 0.), (0, 0.), (0, 0.), (0, 0.)],
      dtype=[('f0', '<i4'), ('f1', '<f4')])
In [428]: arr1['f0'] = arr[:,0]                                                 
In [429]: arr1['f1'] = arr[:,1]                                                 
In [430]: arr1                                                                  
Out[430]: 
array([(20150101, 0.12), (20150102, 0.42), (20150103, 0.12),
       (20150104, 0.46)], dtype=[('f0', '<i4'), ('f1', '<f4')])

这是一个一维结构化数组。请注意表示法的差异，包括 () 的使用。