Python 数据加工 之 数组(numpy)

Python 数据加工 之 数组(numpy)

Python編程 从数据分析到数据科学 摘抄及拓展

随机数

• 生成一个随机数

import random
random.seed(3)
a = random.randint(1,100) #生成一个[1,100]的随机整数
print(a) #31
a = random.uniform(-10,10)#[-10,10]之间的随机浮点数
print(a) #1.852818212543312

• 生成一个随机数组
  *先下载numpy库,在cmd命令行中输入以下命令:

pip install numpy -i http://pypi.douban.com/simple/ --trusted-host pypi.douban.com

import numpy as np

rand = np.random.RandomState(32)
x = rand.randint(0, 10, (3, 6))
print(x)
'''
[[7 5 6 8 3 7]
 [9 3 5 9 4 1]
 [3 1 2 3 8 2]]
'''
----------------------------------------------------------------------------------------------
rand = np.random.RandomState(1)
x = rand.rand(5) *10
print(x)
'''
[4.17022005e+00 7.20324493e+00 1.14374817e-03 3.02332573e+00
 1.46755891e+00]
 '''
 #生成等距数列
 x = np.linspace(0,10,20)
print(x)
'''
[ 0.          0.52631579  1.05263158  1.57894737  2.10526316  2.63157895
  3.15789474  3.68421053  4.21052632  4.73684211  5.26315789  5.78947368
  6.31578947  6.84210526  7.36842105  7.89473684  8.42105263  8.94736842
  9.47368421 10.        ]
'''

数组

• 数组的创建
  如下5种方法

import numpy as np

myArray1 = np.arange(1, 20)
myArray2 = np.array([1, 2, 3, 4, 5, 6])
myArray3 = np.zeros((5, 5))#创建全0
myArray4 = np.full((3, 5), 2)#创建全为2
rand = np.random.RandomState(30)
myArray5 = rand.randint(1, 10, (3, 5))

print(myArray1)  # [ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19]
print(myArray2)  # [1 2 3 4 5 6]
print(myArray3)
'''
[[0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0.]]
 '''
print(myArray4)
'''
[[2 2 2 2 2]
 [2 2 2 2 2]
 [2 2 2 2 2]]
'''
print(myArray5)
'''
[[6 6 5 8 3]
 [6 2 4 8 8]
 [2 2 4 3 3]]
'''
print(type(myArray1))  # <class 'numpy.ndarray'>

• 特征
  • shape代表数组的形状
  • dtype代表数组的数据类型

import numpy as np

myArray1 = np.zeros(shape=(2,15),dtype=float)
print(myArray1)
'''
[[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]
'''

 * 切片的操作和列表类似

import numpy as np

myArray1 = np.arange(1, 20)
print(myArray1)  # [ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19]
print(myArray1[[1, 3, 5]])  # [2 4 6]
print(myArray1[myArray1 < 5])  # [1 2 3 4]
print(myArray1[::-1])  # [19 18 17 16 15 14 13 12 11 10  9  8  7  6  5  4  3  2  1]

• 规整化处理
  np.newaxis的操作,给数组加一个维度
  np.newaxis 与 None 等价

import numpy as np

myArray = np.arange(1, 10)
print(myArray.shape)  # (9,)
myArray1 = myArray[:, np.newaxis]
print(myArray1)
'''
[[1]
 [2]
 [3]
 [4]
 [5]
 [6]
 [7]
 [8]
 [9]]
 '''
print(myArray1.shape)  # (9, 1)
print(myArray1[:, np.newaxis].shape)  # (9, 1, 1)
print(myArray1[np.newaxis,:].shape)  # (1, 9, 1)

• 更改形状 reshape
• 切片

import numpy as np

myArray = np.arange(1, 21).reshape([5, 4])

print(myArray[[0, 2], 1])  # [ 2 10]

• 深浅拷贝(同列表中的深浅复制)需要用.copy()方法 深拷贝
• reshape 和 resize :
  • reshape 直接修改了数组本身,resize返回一个新的数组
• 矩阵的转置: swapaxes()和transpose() 交换维度
• 可以看下面的网页

https://blog.csdn.net/qq1483661204/article/details/70543952

• 多维数组转为一位数组 :flatten()
• 多维数组转化为嵌套列表: tolist()
• 重设数组元素数据类型: astype()
• 数组的三个常用属性,shape,ndim,size

ndarray的计算

• 乘法,相当于矩阵的乘法,对每个元素都乘对应的数
• 横向拆分 np.split()
• 纵向拆分 np.vsplit()

import numpy as np

myArray = np.arange(1, 21).reshape([5, 4])
print(myArray)
# 求矩阵的秩
print(myArray.ndim)  # 2
# 横向拆分
print(np.split(myArray.flatten(), [3, 6]))
# [array([1, 2, 3]), array([4, 5, 6]), array([ 7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20])]
# 纵向拆分
print(np.vsplit(myArray, [1, 2]))
'''
[array([[1, 2, 3, 4]]), array([[5, 6, 7, 8]]), array([[ 9, 10, 11, 12],
       [13, 14, 15, 16],
       [17, 18, 19, 20]])]
'''

• 合并
  axis=0 表示列
  axis=1 表示行

import numpy as np

myArray = np.array([[3, 4, 5], [6, 7, 8]])
myArray1 = np.array([[1, 2, 3], [2, 3, 4]])
print(myArray.shape)  # (2, 3)
print(myArray1.shape)  # (2, 3)
myArray2 = np.concatenate((myArray, myArray1), axis=0)
print(myArray2)
'''
[[3 4 5]
 [6 7 8]
 [1 2 3]
 [2 3 4]]
 '''
print(myArray2.shape)  # (4, 3)
myArray2 = np.concatenate((myArray, myArray1), axis=1)
print(myArray2)
'''
[[3 4 5 1 2 3]
 [6 7 8 2 3 4]]
 '''
print(myArray2.shape)  # (2, 6)

• 横向合并 vstack :列的个数相同
• 纵向合并 hstack :行的个数相同
  通俗的讲就是
  m * n 可以和m * k横向合并
  m * n 可以和 k * n 纵向合并
• numpy中的函数计算

from python 中的api
 >>> np.sum([0.5, 1.5])
    2.0
    >>> np.sum([0.5, 0.7, 0.2, 1.5], dtype=np.int32)
    1 #这里是1因为先把所以的数据线转成int32类型再相加
    >>> np.sum([[0, 1], [0, 5]])
    6 #什么都不写就是全加起来
    >>> np.sum([[0, 1], [0, 5]], axis=0)
    array([0, 6]) #axis = 0代表列相加
    >>> np.sum([[0, 1], [0, 5]], axis=1)
    array([1, 5]) #axis = 1代表行相加

• 元素类型

a1 = np.array([1, 2, 3, None])
print(a1)  # [1 2 3 None]
a1 = np.array([1, 2, 3, None, np.nan])
print(a1)  # [1 2 3 None nan]
print(a1.dtype)  # object

• 插入与删除
  np.delete 下标删除

import numpy as np

myArray1 = np.array([11, 12, 13, 14, 15, 16, 17])
a = np.delete(myArray1, 2)
print(a)  # [11 12 14 15 16 17]
print(myArray1)  # [11 12 13 14 15 16 17]

insert 下标插入

a = np.insert(myArray1,2,97)
print(a)  # [11 12 97 13 14 15 16 17]
print(myArray1)  # [11 12 13 14 15 16 17]

• 缺失值处理(np.nan)
  
  及 np.nan是float类型,可以参加算数运算,None缺不行
• 判断是否有缺失值 np.isnan()
• np.any() 是否至少有1个
• np.all() 是否全部

import numpy as np

a = np.array([1, np.nan, 2, 3, 4, 5, 6])
print(np.isnan(a))  # [False  True False False False False False]
print(np.any(np.isnan(a)))  # True
print(np.all(np.isnan(a)))  #False
print(np.nansum(a)) #21.0 可以求和操作,类型为float
print(np.sum(a)) #nan

• ndarray的广播规则
  如果列数一样但行数不一样,进行以行位单位的广播操作

import numpy as np

a = np.array([1, 2, 3, 4])
b = np.arange(1, 13).reshape(3, 4)
print(a + b)
'''
[[ 2  4  6  8]
 [ 6  8 10 12]
 [10 12 14 16]]
'''

• 排序

import numpy as np

a = np.array([1, 2, 3, 5, 7, 4, 6, 8, 10])
'排序返回结果'
print(np.sort(a))  # [ 1  2  3  4  5  6  7  8 10]
'排序返回下标'
print(np.argsort(a))  # [0 1 2 5 3 6 4 7 8]
'按行排序'
a = a.reshape(3, 3)
print(a)
'''
[[ 1  2  3]
 [ 5  7  4]
 [ 6  8 10]]
 '''
# 按行排序
print(np.sort(a, axis=1))
'''
[[ 1  2  3]
 [ 4  5  7]
 [ 6  8 10]]
 '''