首页 > > 网络编程 > 其它 >

利用Python进行数据分析_Numpy_基础_2

2018-06-18 01:12:53来源：未知阅读 ()

Numpy数据类型包括：

int8、uint8、int16、uint16、int32、uint32、int64、uint64、float16、float32、float64、float128、complex64、complex128、complex256、bool、object、string_、unicode_

astype

显示转换数组类型的方法

例如：

NumPy数组的索引和切片

索引

和python列表差不多，基本上没啥区别

切片

NumPy数组的切片出来的数值改变，就会改变NumPy数组的源数组的值。NumPy数组的切片是源数组的视图，而不是新复制出来的一个数组。从下面的例子,我们可以看到arr[1,1]=0 ,arr的数组变化了，data数组对应位置的数值也变化了。

In [101]: data = np.random.randn(4,4)

In [102]: data
Out[102]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  1.42196278,  0.84758994]])

In [103]: arr = data[2:,1:]

In [104]: arr
Out[104]:
array([[-0.91963212,  3.09214837, -0.6070068 ],
       [-1.46576298,  1.42196278,  0.84758994]])

In [105]: arr = 0

In [106]: data
Out[106]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  1.42196278,  0.84758994]])

In [107]: arr
Out[107]: 0

In [108]: arr = data[2:,1:]

In [109]: arr
Out[109]:
array([[-0.91963212,  3.09214837, -0.6070068 ],
       [-1.46576298,  1.42196278,  0.84758994]])

In [110]: arr == 0
Out[110]:
array([[False, False, False],
       [False, False, False]], dtype=bool)

In [111]: arr
Out[111]:
array([[-0.91963212,  3.09214837, -0.6070068 ],
       [-1.46576298,  1.42196278,  0.84758994]])

In [112]: arr[1,1]=0

In [113]: arr
Out[113]:
array([[-0.91963212,  3.09214837, -0.6070068 ],
       [-1.46576298,  0.        ,  0.84758994]])

In [114]: data
Out[114]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  0.        ,  0.84758994]])

In [115]:

如果要复制NumPy数组的切片，则可以使用显示复制方法copy()

In [116]: data
Out[116]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  0.        ,  0.84758994]])

In [117]: arr = data

In [118]: arr
Out[118]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  0.        ,  0.84758994]])

In [119]: arr = np.copy(data)

In [120]: arr
Out[120]:
array([[-1.68867271, -0.89369286, -0.0288363 ,  0.73855122],
       [-0.13084603,  0.43972144,  0.73542583,  1.99925332],
       [ 0.04291022, -0.91963212,  3.09214837, -0.6070068 ],
       [-0.01416294, -1.46576298,  0.        ,  0.84758994]])

布尔类型索引

假设每个字符串对应data数组一行数据。需要注意布尔型数组的长度必须与被索引的轴长度一致。

通过布尔型索引查找数组数值的方式如下：

In [140]: names = np.array(['aaa','bbb','ccc','ddd','eee','fff'])

In [141]: data = np.random.randn(6,4)

In [142]: names
Out[142]:
array(['aaa', 'bbb', 'ccc', 'ddd', 'eee', 'fff'],
dtype='<U3')

In [143]: data
Out[143]:
array([[ 0.49394026, -0.65887621, -0.26946242, 0.22042355],
        [-1.11606179, -1.94945158, -0.4866134 , 0.67712409],
      [-2.33792045, 0.01639887, -0.46020647, 0.84180777],
        [-1.99622938, 1.937877 , -0.17134376, 0.56915872],
        [ 1.50980905, 0.07244016, -0.95650922, 1.23508517],
        [ 0.74706519, -0.03149619, -0.38235363, 0.69786257]])

In [144]: names == 'aaa'
Out[144]: array([ True, False, False, False, False, False], dtype=bool)

In [145]: data[names=='aaa']
Out[145]: array([[ 0.49394026, -0.65887621, -0.26946242, 0.22042355]])

In [146]: names =='ccc'
Out[146]: array([False, False, True, False, False, False], dtype=bool)

In [147]: data[names=='ccc']
Out[147]: array([[-2.33792045, 0.01639887, -0.46020647, 0.84180777]])

布尔数组索引结合切片进行查找数组的数值：

In [148]: data[names=='aaa',2]
Out[148]: array([-0.26946242])

In [149]: data[names=='aaa',2:]
Out[149]: array([[-0.26946242,  0.22042355]])

In [150]: data[names=='aaa',1:]
Out[150]: array([[-0.65887621, -0.26946242,  0.22042355]])

反向查找

In [155]: names !='aaa'
Out[155]: array([False,  True,  True,  True,  True,  True], dtype=bool)

In [156]: data[names!='aaa']
Out[156]:
array([[-1.11606179, -1.94945158, -0.4866134 ,  0.67712409],
       [-2.33792045,  0.01639887, -0.46020647,  0.84180777],
       [-1.99622938,  1.937877  , -0.17134376,  0.56915872],
       [ 1.50980905,  0.07244016, -0.95650922,  1.23508517],
       [ 0.74706519, -0.03149619, -0.38235363,  0.69786257]])

组合查找

In [171]: mask = (names == 'aaa')|(names == 'ccc')

In [172]: mask
Out[172]: array([ True, False,  True, False, False, False], dtype=bool)

In [173]: data[mask]
Out[173]:
array([[ 0.49394026, -0.65887621, -0.26946242,  0.22042355],
       [-2.33792045,  0.01639887, -0.46020647,  0.84180777]])

花式索引

其实就是利用整数列表或数组进行索引查找。花式索引与数组切片不同，花式索引会将数据复制到新的数组。

整数列表

创建一个二维数组arr，然后传入[3,1]，意思就是按 arr [3,:]、arr[1,:]的顺序显示出来。

In [203]: arr = np.array(([1,2,3,4],[2,3,4,5],[3,4,5,6],[7,8,9,10]))

In [204]: arr
Out[204]:
array([[ 1,  2,  3,  4],
       [ 2,  3,  4,  5],
       [ 3,  4,  5,  6],
       [ 7,  8,  9, 10]])

In [205]: arr[[3,1]]
Out[205]:
array([[ 7,  8,  9, 10],
       [ 2,  3,  4,  5]])

传入多个整数数组

一次传入多个整数数组，返回的是一个一维数组。

数组转置对轴对换

数组转置，是指将原数组A的行与列交换得到的一个新数组。

比如：

的转置是，的转置是

方法1：T

In [227]: arr = np.random.randn(10)

In [228]: arr
Out[228]:
array([-1.42853867,  1.54300781, -0.74079757, -1.20272388, -1.00416459,
       -0.59571731,  1.16744662,  0.05739806,  1.01660691, -0.84625494])

In [229]: arr.T
Out[229]:
array([-1.42853867,  1.54300781, -0.74079757, -1.20272388, -1.00416459,
       -0.59571731,  1.16744662,  0.05739806,  1.01660691, -0.84625494])

In [230]: arr = np.random.randn(3,5)

In [231]: arr
Out[231]:
array([[ 1.36114118,  0.48455027,  0.64847485,  0.01691785, -0.03622465],
       [-2.31302164,  1.14992892, -1.47836923,  1.08003907, -1.33663009],
       [-0.38005499,  1.3517217 ,  2.52024026, -0.3576492 ,  0.46016645]])

In [232]: arr.T
Out[232]:
array([[ 1.36114118, -2.31302164, -0.38005499],
       [ 0.48455027,  1.14992892,  1.3517217 ],
       [ 0.64847485, -1.47836923,  2.52024026],
       [ 0.01691785,  1.08003907, -0.3576492 ],
       [-0.03622465, -1.33663009,  0.46016645]])

方法2：transpose

三维数组 arr：4个3*4的数组

In [275]: arr = np.arange(48).reshape(4,3,4)

In [276]: arr
Out[276]:
array([[[ 0, 1, 2, 3],
[ 4, 5, 6, 7],
[ 8, 9, 10, 11]],

       [[12, 13, 14, 15],
         [16, 17, 18, 19],
         [20, 21, 22, 23]],

       [[24, 25, 26, 27],
         [28, 29, 30, 31],
         [32, 33, 34, 35]],

       [[36, 37, 38, 39],
         [40, 41, 42, 43],
         [44, 45, 46, 47]]])

transpose参数的真正意义在于这个shape元组的索引（轴编号）。

In [278]: arr.shape
Out[278]: (4, 3, 4)

arr数组的索引（轴编号）：0、1、2

下面是按索引 2、0、1进行对换

In [277]: arr.transpose(2,0,1)
 Out[277]:
 array([[[ 0,  4,  8],
         [12, 16, 20],
         [24, 28, 32],
         [36, 40, 44]],

       [[ 1,  5,  9],
         [13, 17, 21],
         [25, 29, 33],
         [37, 41, 45]],

       [[ 2,  6, 10],
         [14, 18, 22],
         [26, 30, 34],
         [38, 42, 46]],

       [[ 3,  7, 11],
         [15, 19, 23],
         [27, 31, 35],
         [39, 43, 47]]])

然后，我们再按（轴编号）0、1、2 对换回到原来的样子

In [279]: arr.transpose(0,1,2)
Out[279]:
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],

       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]],

       [[24, 25, 26, 27],
        [28, 29, 30, 31],
        [32, 33, 34, 35]],

       [[36, 37, 38, 39],
        [40, 41, 42, 43],
        [44, 45, 46, 47]]])

方法3：`swapaxes`

`swapaxes返回的是源数组的视图。`

`相比于transpose是需要传入一个索引元组（轴编号），swapaxes只需要一对索引元组（轴编号）。`

In [283]: arr.swapaxes(2,1)
Out[283]:
array([[[ 0,  4,  8],
        [ 1,  5,  9],
        [ 2,  6, 10],
        [ 3,  7, 11]],

       [[12, 16, 20],
        [13, 17, 21],
        [14, 18, 22],
        [15, 19, 23]],

       [[24, 28, 32],
        [25, 29, 33],
        [26, 30, 34],
        [27, 31, 35]],

       [[36, 40, 44],
        [37, 41, 45],
        [38, 42, 46],
        [39, 43, 47]]])