《利用python进行数据分析》（四）—NumPy_利用python进行数据分析百度云-CSDN博客

本文链接：https://blog.csdn.net/wuxioafan/article/details/103646002

ndarray：多位数组对象

创建ndarry：array函数 zeros empty range
ndarray数据类型：
1. dtype定义查看类型
2. astype：转换数据类型
数组运算：不用for循环就能实现数组之间的运算
基本索引和切片：
1. 类似列表
2. 改变切片中的值时，变动会体现在原始数据arr中
3. [:]给数组所有值赋值
4. 多维数组索引，访问
5. 多维切片索引号

In [60]: arr = np.arange(10)

In [61]: arr
Out[61]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

In [62]: arr[5]
Out[62]: 5

In [63]: arr[5:8]
Out[63]: array([5, 6, 7])

In [64]: arr[5:8] = 12

In [65]: arr
Out[65]: array([ 0,  1,  2,  3,  4, 12, 12, 12,  8,  9])
In [66]: arr_slice = arr[5:8]

In [67]: arr_slice
Out[67]: array([12, 12, 12])
In [68]: arr_slice[1] = 12345

In [69]: arr
Out[69]: array([    0,     1,     2,     3,     4,    12, 12345,    12,     8,   
  9])
In [72]: arr2d = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])

In [73]: arr2d[2]
Out[73]: array([7, 8, 9])
In [74]: arr2d[0][2]
Out[74]: 3

In [75]: arr2d[0, 2]
Out[75]: 3
In [76]: arr3d = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]])

In [77]: arr3d
Out[77]: 
array([[[ 1,  2,  3],
        [ 4,  5,  6]],
       [[ 7,  8,  9],
        [10, 11, 12]]])
In [78]: arr3d[0]
Out[78]: 
array([[1, 2, 3],
       [4, 5, 6]])
In [79]: old_values = arr3d[0].copy()

In [80]: arr3d[0] = 42

In [81]: arr3d
Out[81]: 
array([[[42, 42, 42],
        [42, 42, 42]],
       [[ 7,  8,  9],
        [10, 11, 12]]])

In [82]: arr3d[0] = old_values

In [83]: arr3d
Out[83]: 
array([[[ 1,  2,  3],
        [ 4,  5,  6]],
       [[ 7,  8,  9],
        [10, 11, 12]]])
In [85]: x = arr3d[1]

In [86]: x
Out[86]: 
array([[ 7,  8,  9],
       [10, 11, 12]])

In [87]: x[0]
Out[87]: array([7, 8, 9])
In [90]: arr2d
Out[90]: 
array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

In [91]: arr2d[:2]
Out[91]: 
array([[1, 2, 3],
       [4, 5, 6]])
In [92]: arr2d[:2, 1:]
Out[92]: 
array([[2, 3],
       [5, 6]])
In [95]: arr2d[:, :1]
Out[95]: 
array([[1],
       [4],
       [7]])

布尔型索引

In [98]: names = np.array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'])

In [99]: data = np.random.randn(7, 4)

In [100]: names
Out[100]: 
array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'],
      dtype='<U4')

In [101]: data
Out[101]: 
array([[ 0.0929,  0.2817,  0.769 ,  1.2464],
       [ 1.0072, -1.2962,  0.275 ,  0.2289],
       [ 1.3529,  0.8864, -2.0016, -0.3718],
       [ 1.669 , -0.4386, -0.5397,  0.477 ],
       [ 3.2489, -1.0212, -0.5771,  0.1241],
       [ 0.3026,  0.5238,  0.0009,  1.3438],
       [-0.7135, -0.8312, -2.3702, -1.8608]])
In [102]: names == 'Bob'
Out[102]: array([ True, False, False,  True, False, False, False], dtype=bool)
In [103]: data[names == 'Bob']
Out[103]: 
array([[ 0.0929,  0.2817,  0.769 ,  1.2464],
       [ 1.669 , -0.4386, -0.5397,  0.477 ]])

花式索引
1. 特定顺序索引
2. 负数索引
3. 与切片不一样，将数据复制到新数组

In [117]: arr = np.empty((8, 4))

In [118]: for i in range(8):
   .....:     arr[i] = i

In [119]: arr
Out[119]: 
array([[ 0.,  0.,  0.,  0.],
       [ 1.,  1.,  1.,  1.],
       [ 2.,  2.,  2.,  2.],
       [ 3.,  3.,  3.,  3.],
       [ 4.,  4.,  4.,  4.],
       [ 5.,  5.,  5.,  5.],
       [ 6.,  6.,  6.,  6.],
       [ 7.,  7.,  7.,  7.]])
In [120]: arr[[4, 3, 0, 6]]
Out[120]: 
array([[ 4.,  4.,  4.,  4.],
       [ 3.,  3.,  3.,  3.],
       [ 0.,  0.,  0.,  0.],
       [ 6.,  6.,  6.,  6.]])
In [121]: arr[[-3, -5, -7]]
Out[121]: 
array([[ 5.,  5.,  5.,  5.],
       [ 3.,  3.,  3.,  3.],
       [ 1.,  1.,  1.,  1.]])

In [122]: arr = np.arange(32).reshape((8, 4))

In [123]: arr
Out[123]: 
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]])

In [124]: arr[[1, 5, 7, 2], [0, 3, 1, 2]]
Out[124]: array([ 4, 23, 29, 10])
In [125]: arr[[1, 5, 7, 2]][:, [0, 3, 1, 2]]
Out[125]: 
array([[ 4,  7,  5,  6],
       [20, 23, 21, 22],
       [28, 31, 29, 30],
       [ 8, 11,  9, 10]])

数组转置和轴对换
1. T转置
2. transpose方法
3. swapaxes方法

In [126]: arr = np.arange(15).reshape((3, 5))

In [127]: arr
Out[127]: 
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14]])

In [128]: arr.T
Out[128]: 
array([[ 0,  5, 10],
       [ 1,  6, 11],
       [ 2,  7, 12],
       [ 3,  8, 13],
       [ 4,  9, 14]])
In [132]: arr = np.arange(16).reshape((2, 2, 4))

In [133]: arr
Out[133]: 
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7]],
       [[ 8,  9, 10, 11],
        [12, 13, 14, 15]]])

In [134]: arr.transpose((1, 0, 2))
Out[134]: 
array([[[ 0,  1,  2,  3],
        [ 8,  9, 10, 11]],
       [[ 4,  5,  6,  7],
        [12, 13, 14, 15]]])
In [135]: arr
Out[135]: 
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7]],
       [[ 8,  9, 10, 11],
        [12, 13, 14, 15]]])

In [136]: arr.swapaxes(1, 2)
Out[136]: 
array([[[ 0,  4],
        [ 1,  5],
        [ 2,  6],
        [ 3,  7]],
       [[ 8, 12],
        [ 9, 13],
        [10, 14],
        [11, 15]]])

通用函数

利用数组进行数据处理

矢量化数组运算
将条件逻辑表述为数组运算
1. np.where
数学和统计方法
1. mean
2. sum
3. cumsum
用于布尔型数组的方法
排序
1. sort
唯一化以及其它的集合逻辑
1. np.unique:找到数组中的唯一值，并返回排序结果
2. np.in1d
集合函数

In [177]: arr = np.random.randn(5, 4)

In [178]: arr
Out[178]: 
array([[ 2.1695, -0.1149,  2.0037,  0.0296],
       [ 0.7953,  0.1181, -0.7485,  0.585 ],
       [ 0.1527, -1.5657, -0.5625, -0.0327],
       [-0.929 , -0.4826, -0.0363,  1.0954],
       [ 0.9809, -0.5895,  1.5817, -0.5287]])

In [179]: arr.mean()
Out[179]: 0.19607051119998253

In [180]: np.mean(arr)
Out[180]: 0.19607051119998253

In [181]: arr.sum()
Out[181]: 3.9214102239996507
In [182]: arr.mean(axis=1)
Out[182]: array([ 1.022 ,  0.1875, -0.502 , -0.0881,  0.3611])

In [183]: arr.sum(axis=0)
Out[183]: array([ 3.1693, -2.6345,  2.2381,  1.1486])
In [184]: arr = np.array([0, 1, 2, 3, 4, 5, 6, 7])

In [185]: arr.cumsum()
Out[185]: array([ 0,  1,  3,  6, 10, 15, 21, 28])
In [186]: arr = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8]])
In [187]: arr
Out[187]: array([[0, 1, 2],       [3, 4, 5],       [6, 7, 8]])
In [188]: arr.cumsum(axis=0)Out[188]: array([[ 0,  1,  2],       [ 3,  5,  7],       [ 9, 12, 15]])
In [189]: arr.cumprod(axis=1)Out[189]: array([[  0,   0,   0],       [  3,  12,  60],       [  6,  42, 336]])
In [190]: arr = np.random.randn(100)
In [191]: (arr > 0).sum() # Number of positive values
Out[191]: 42In [192]: bools = np.array([False, False, True, False])
In [193]: bools.any()Out[193]: True
In [194]: bools.all()Out[194]: False
In [195]: arr = np.random.randn(6)
In [196]: arrOut[196]: array([ 0.6095, -0.4938,  1.24  , -0.1357,  1.43  , -0.8469])
In [197]: arr.sort()
In [198]: arr
Out[198]: array([-0.8469, -0.4938, -0.1357,  0.6095,  1.24  ,  1.43  ])
In [199]: arr = np.random.randn(5, 3)
In [200]: arr
Out[200]: array([[ 0.6033,  1.2636, -0.2555],       [-0.4457,  0.4684, -0.9616],       [-1.8245,  0.6254,  1.0229],       [ 1.1074,  0.0909, -0.3501],       [ 0.218 , -0.8948, -1.7415]])
In [201]: arr.sort(1)
In [202]: arr
Out[202]: array([[-0.2555,  0.6033,  1.2636],       [-0.9616, -0.4457,  0.4684],       [-1.8245,  0.6254,  1.0229],       [-0.3501,  0.0909,  1.1074],       [-1.7415, -0.8948,  0.218 ]])
In [203]: large_arr = np.random.randn(1000)
In [204]: large_arr.sort()
In [205]: large_arr[int(0.05 * len(large_arr))] # 5% quantile
Out[205]: -1.5311513550102103
In [206]: names = np.array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'])
In [207]: np.unique(names)
Out[207]: array(['Bob', 'Joe', 'Will'],      dtype='<U4')
In [208]: ints = np.array([3, 3, 3, 2, 2, 1, 1, 4, 4])
In [209]: np.unique(ints)
Out[209]: array([1, 2, 3, 4])
In [210]: sorted(set(names))
Out[210]: ['Bob', 'Joe', 'Will']
In [211]: values = np.array([6, 0, 0, 3, 2, 5, 6])
In [212]: np.in1d(values, [2, 3, 6])
Out[212]: array([ True, False, False,  True,  True, False,  True], dtype=bool)

用于数组的文件输入输出

np.save和np.load

In [213]: arr = np.arange(10)
In [214]: np.save('some_array', arr)
In [215]: np.load('some_array.npy')
Out[215]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [216]: np.savez('array_archive.npz', a=arr, b=arr)
In [217]: arch = np.load('array_archive.npz')
In [218]: arch['b']
Out[218]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
In [219]: np.savez_compressed('arrays_compressed.npz', a=arr, b=arr)

线性代数

dot函数
numpy.linalg

In [223]: x = np.array([[1., 2., 3.], [4., 5., 6.]])
In [224]: y = np.array([[6., 23.], [-1, 7], [8, 9]])

In [225]: x
Out[225]: 
array([[ 1.,  2.,  3.],
       [ 4.,  5.,  6.]])

In [226]: y
Out[226]: 
array([[  6.,  23.],
       [ -1.,   7.],
       [  8.,   9.]])

In [227]: x.dot(y)
Out[227]: 
array([[  28.,   64.],
       [  67.,  181.]])
In [228]: np.dot(x, y)
Out[228]: 
array([[  28.,   64.],
       [  67.,  181.]])
In [229]: np.dot(x, np.ones(3))
Out[229]: array([  6.,  15.])
In [231]: from numpy.linalg import inv, qr

In [232]: X = np.random.randn(5, 5)

In [233]: mat = X.T.dot(X)

In [234]: inv(mat)
Out[234]: 
array([[  933.1189,   871.8258, -1417.6902, -1460.4005,  1782.1391],
       [  871.8258,   815.3929, -1325.9965, -1365.9242,  1666.9347],
       [-1417.6902, -1325.9965,  2158.4424,  2222.0191, -2711.6822],
       [-1460.4005, -1365.9242,  2222.0191,  2289.0575, -2793.422 ],
       [ 1782.1391,  1666.9347, -2711.6822, -2793.422 ,  3409.5128]])

In [235]: mat.dot(inv(mat))
Out[235]: 
array([[ 1.,  0., -0., -0., -0.],
       [-0.,  1.,  0.,  0.,  0.],
       [ 0.,  0.,  1.,  0.,  0.],
       [-0.,  0.,  0.,  1., -0.],
       [-0.,  0.,  0.,  0.,  1.]])

In [236]: q, r = qr(mat)

In [237]: r
Out[237]: 
array([[-1.6914,  4.38  ,  0.1757,  0.4075, -0.7838],
       [ 0.    , -2.6436,  0.1939, -3.072 , -1.0702],
       [ 0.    ,  0.    , -0.8138,  1.5414,  0.6155],
       [ 0.    ,  0.    ,  0.    , -2.6445, -2.1669],
       [ 0.    ,  0.    ,  0.    ,  0.    ,  0.0002]])

伪随机数生成

numpy.random
1. normal：正态分布样本
2. np.random.seed：种子
3. 隔离随机数种子：np.random.RandomState

In [238]: samples = np.random.normal(size=(4, 4))

In [239]: samples
Out[239]: 
array([[ 0.5732,  0.1933,  0.4429,  1.2796],
       [ 0.575 ,  0.4339, -0.7658, -1.237 ],
       [-0.5367,  1.8545, -0.92  , -0.1082],
       [ 0.1525,  0.9435, -1.0953, -0.144 ]])
In [240]: from random import normalvariate

In [241]: N = 1000000

In [242]: %timeit samples = [normalvariate(0, 1) for _ in range(N)]
1.77 s +- 126 ms per loop (mean +- std. dev. of 7 runs, 1 loop each)

In [243]: %timeit np.random.normal(size=N)
61.7 ms +- 1.32 ms per loop (mean +- std. dev. of 7 runs, 10 loops each)
In [244]: np.random.seed(1234)
In [245]: rng = np.random.RandomState(1234)

In [246]: rng.randn(10)
Out[246]: 
array([ 0.4714, -1.191 ,  1.4327, -0.3127, -0.7206,  0.8872,  0.8596,
       -0.6365,  0.0157, -2.2427])