利用python进入数据分析之Numpy基础知识
In [182]:
%matplotlib inline
from __future__ import division
from numpy.random import randn
import numpy as np
np.set_printoptions(precision=4, suppress=True)
In [2]:
data = randn(2, 3) # 两行三列的数组，
In [4]:
data
Out[4]:
In [5]:
data * 10
Out[5]:
In [6]:
data + data
Out[6]:
In [7]:
data.shape # 数据维度
Out[7]:
In [8]:
data.dtype # 数据类型
Out[8]:
In [9]:
data1 = [6, 7.5, 8, 0, 1]
arr1 = np.array(data1)
arr1
Out[9]:
In [11]:
data2 = [[1, 2, 3, 4], [5, 6, 7, 8]]
arr2 = np.array(data2)
arr2
Out[11]:
In [12]:
arr2.ndim
Out[12]:
In [13]:
arr2.shape
Out[13]:
In [14]:
arr1.dtype
Out[14]:
In [15]:
arr2.dtype
Out[15]:
In [16]:
np.zeros(10) # 创建全0数组
Out[16]:
In [17]:
np.zeros((3, 6))
Out[17]:
In [18]:
np.empty((2, 3, 2)) # 不安全的用法
Out[18]:
In [19]:
np.arange(15) #创建数组，按照顺序
Out[19]:
In [20]:
arr1 = np.array([1, 2, 3], dtype=np.float64)
arr2 = np.array([1, 2, 3], dtype=np.int32)
In [21]:
arr1.dtype
Out[21]:
In [22]:
arr2.dtype
Out[22]:
In [23]:
arr = np.array([1, 2, 3, 4, 5])
In [24]:
arr.dtype
Out[24]:
In [25]:
float_arr = arr.astype(np.float64) # 整数类型转换浮点数类型
In [26]:
float_arr.dtype
Out[26]:
In [27]:
arr = np.array([3.7, -1.2, -2.6, 0.5, 12.9, 10.1])
In [28]:
arr
Out[28]:
In [29]:
arr.astype(np.int32) #浮点数转换成整数
Out[29]:
In [30]:
numeric_strings = np.array(['1.25', '-9.6', '42'], dtype=np.string_)
numeric_strings.astype(float) # 字符串转换成浮点数
Out[30]:
In [31]:
int_array = np.arange(10)
calibers = np.array([.22, .270, .357, .380, .44, .50], dtype=np.float64)
int_array.astype(calibers.dtype)
Out[31]:
In [32]:
empty_uint32 = np.empty(8, dtype='u4')
empty_uint32
Out[32]:
In [33]:
arr = np.array([[1., 2., 3.], [4., 5., 6.]])
In [34]:
arr
Out[34]:
In [35]:
arr * arr
Out[35]:
In [36]:
arr - arr
Out[36]:
In [37]:
1 / arr
Out[37]:
In [38]:
arr ** 0.5
Out[38]:
In [39]:
arr = np.arange(10)
In [40]:
arr
Out[40]:
In [41]:
arr[5]
Out[41]:
In [42]:
arr[5:8]
Out[42]:
In [43]:
arr[5:8] = 12 # 将5到8赋值为12
In [44]:
arr
Out[44]:
In [45]:
arr_slice = arr[5:8]
In [46]:
arr_slice[1] = 12345
In [47]:
arr
Out[47]:
In [49]:
arr_slice[:] = 64
In [50]:
arr
Out[50]:
In [51]:
arr2d = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
arr2d[2]
Out[51]:
In [53]:
arr2d[0][2]
Out[53]:
In [54]:
arr2d[0, 2]
Out[54]:
In [55]:
arr3d = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]])
arr3d
Out[55]:
In [56]:
arr3d[0]
Out[56]:
In [57]:
old_values = arr3d[0].copy()
arr3d[0] = 42
arr3d
Out[57]:
In [58]:
arr3d[0] = old_values
arr3d
Out[58]:
In [59]:
arr3d[1, 0]
Out[59]:
In [60]:
arr[1:6]
Out[60]:
In [61]:
arr2d
Out[61]:
In [62]:
arr2d[:2]
Out[62]:
In [63]:
arr2d[:2, 1:]
Out[63]:
In [64]:
arr2d[1, :2]
Out[64]:
In [65]:
arr2d[2, :1]
Out[65]:
In [66]:
arr2d[:, :1]
Out[66]:
In [67]:
arr2d[:2, 1:] = 0
In [68]:
arr2d
Out[68]:
In [69]:
names = np.array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'])
data = randn(7, 4) # 生成正太分布的随机数据
In [70]:
names
Out[70]:
In [71]:
data
Out[71]:
In [72]:
names == 'Bob'
Out[72]:
In [81]:
data[names == 'Bob'] # 相当于只要第一行和第三行数据
Out[81]:
In [86]:
data[names == 'Bob', 2:]
Out[86]:
In [87]:
data[names == 'Bob', 3]
Out[87]:
In [88]:
names != 'Bob'
Out[88]:
In [89]:
data[-(names == 'Bob')]
Out[89]:
In [90]:
mask = (names == 'Bob') | (names == 'Will')
mask
data[mask]
Out[90]:
In [92]:
data[data < 0] = 0 # 将小于0的值，置0
data
Out[92]:
In [93]:
data[names != 'Joe'] = 7
data
Out[93]:
In [94]:
arr = np.empty((8, 4))
for i in range(8):
arr[i] = i
arr
Out[94]:
In [95]:
arr[[4, 3, 0, 6]]
Out[95]:
In [96]:
arr[[-3, -5, -7]]
Out[96]:
In [98]:
# more on reshape in Chapter 12
arr = np.arange(32).reshape((8, 4))
arr
Out[98]:
In [99]:
arr[[1, 5, 7, 2], [0, 3, 1, 2]] #行要1 5 7 2，列要0 3 1 2
Out[99]:
In [100]:
arr[[1, 5, 7, 2]][:, [0, 3, 1, 2]]
Out[100]:
In [101]:
arr[np.ix_([1, 5, 7, 2], [0, 3, 1, 2])]
Out[101]:
In [103]:
arr = np.arange(15).reshape((3, 5))
In [104]:
arr
Out[104]:
In [105]:
arr.T
Out[105]:
In [106]:
arr = np.random.randn(6, 3)
In [107]:
np.dot(arr.T, arr) # 计算矩阵内积
Out[107]:
In [108]:
arr = np.arange(16).reshape((2, 2, 4))
In [109]:
arr
Out[109]:
In [110]:
arr.transpose((1, 0, 2)) #轴转置
Out[110]:
In [111]:
arr
Out[111]:
In [112]:
arr.swapaxes(1, 2)
Out[112]:
In [115]:
arr = np.arange(10)
np.sqrt(arr) #每个元素求平方根
Out[115]:
In [116]:
np.exp(arr) #计算各元素的指数
Out[116]:
In [117]:
x = randn(8)
y = randn(8)
In [118]:
x
Out[118]:
In [119]:
y
Out[119]:
In [120]:
np.maximum(x, y) # element-wise maximum（元素级最大值）
Out[120]:
In [121]:
arr = randn(7) * 5
In [123]:
arr
Out[123]:
In [124]:
np.modf(arr) #将数组的小树和整数部分以两个独立的数组的形式返回
Out[124]:
In [125]:
points = np.arange(-5, 5, 0.01) # 1000个间隔相等的点
In [126]:
points
Out[126]:
In [127]:
xs, ys = np.meshgrid(points, points) #矢量化
In [128]:
xs
Out[128]:
In [129]:
ys
Out[129]:
In [130]:
from matplotlib.pyplot import imshow, title
In [137]:
import matplotlib.pyplot as plt
z = np.sqrt(xs ** 2 + ys ** 2)
In [138]:
z
Out[138]:
In [133]:
plt.imshow(z, cmap=plt.cm.gray); plt.colorbar()
Out[133]:
In [134]:
plt.title("Image plot of $\sqrt{x^2 + y^2}$ for a grid of values")
Out[134]:
In [136]:
plt.draw()
In [139]:
xarr = np.array([1.1, 1.2, 1.3, 1.4, 1.5])
yarr = np.array([2.1, 2.2, 2.3, 2.4, 2.5])
cond = np.array([True, False, True, True, False])
In [140]:
result = [(x if c else y)
for x, y, c in zip(xarr, yarr, cond)]
result
Out[140]:
In [141]:
result = np.where(cond, xarr, yarr)
result
Out[141]:
In [142]:
arr = randn(4, 4)
arr
Out[142]:
In [144]:
np.where(arr > 0, 2, -2)
Out[144]:
In [145]:
np.where(arr > 0, 2, arr) # set only positive values to 2
Out[145]:
In [150]:
arr = np.random.randn(5, 4) # 正态分布数据
arr.mean()
Out[150]:
In [151]:
np.mean(arr) #平均数
Out[151]:
In [154]:
arr.sum() # 求和
Out[154]:
In [156]:
arr
Out[156]:
In [157]:
arr.mean(axis=1)
Out[157]:
In [158]:
arr = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8]])
In [159]:
arr.cumsum(0) # 所有元素的累计和
Out[159]:
In [161]:
arr.cumprod(1)# 所有元素的累计积
Out[161]:
In [163]:
arr = randn(100)
In [167]:
arr
Out[167]:
In [166]:
(arr > 0).sum() # Number of positive values
Out[166]:
In [168]:
bools = np.array([False, False, True, False])
In [169]:
bools.any() # 测试数组中是否存在一个或多个TRUE
Out[169]:
In [170]:
bools.all()# 测试数组中的值是否都是TRUE
Out[170]:
In [171]:
arr = randn(8)
In [172]:
arr
Out[172]:
In [173]:
arr.sort()
In [174]:
arr
Out[174]:
In [175]:
arr = randn(5, 3)
In [176]:
arr
Out[176]:
In [177]:
arr.sort(1)
In [178]:
arr
Out[178]:
In [185]:
names = np.array(['Bob', 'Joe', 'Will', 'Bob', 'Will', 'Joe', 'Joe'])
np.unique(names) # 找出数组中的唯一值
Out[185]:
In [186]:
ints = np.array([3, 3, 3, 2, 2, 1, 1, 4, 4])
np.unique(ints)
Out[186]:
In [188]:
arr = np.arange(10)
np.save('some_array', arr)
In [189]:
np.load('some_array.npy')
Out[189]:
In [190]:
np.savez('array_archive.npz', a=arr, b=arr) # 用压缩文件的方式加载
In [191]:
arch = np.load('array_archive.npz') # 压缩文件的方式导入
arch['b']
Out[191]:
In [194]:
x = np.array([[1., 2., 3.], [4., 5., 6.]])
y = np.array([[6., 23.], [-1, 7], [8, 9]])
In [195]:
x
Out[195]:
In [196]:
y
Out[196]:
In [197]:
x.dot(y) # 计算内积
Out[197]:
In [198]:
np.dot(x, np.ones(3))
Out[198]:
In [200]:
samples = np.random.normal(size=(4, 4))
In [201]:
samples
Out[201]:
In [202]:
from random import normalvariate
N = 1000000
%timeit samples = [normalvariate(0, 1) for _ in xrange(N)]
In [204]:
%timeit np.random.normal(size=N) # np的random.normal模块要比python内置的random块
In [205]:
import random
position = 0
walk = [position]
steps = 1000
for i in xrange(steps):
step = 1 if random.randint(0, 1) else -1
position += step
walk.append(position)
In [236]:
nsteps = 1000
draws = np.random.randint(0, 2, size=nsteps) #随机生成0 1共1000个
steps = np.where(draws > 0, 1, -1)# 进行累加
walk = steps.cumsum()
In [237]:
walk.min()
Out[237]:
In [238]:
walk.max()
Out[238]:
In [239]:
(np.abs(walk) >= 10).argmax() #多久才能超过绝对值10
Out[239]:
In [242]:
nwalks = 5000
nsteps = 1000
draws = np.random.randint(0, 2, size=(nwalks, nsteps)) # 0 or 1
steps = np.where(draws > 0, 1, -1) #将0和1变成-1和1
walks = steps.cumsum(1)
walks
Out[242]:
In [244]:
walks.max()
Out[244]:
In [245]:
walks.min()
Out[245]:
In [246]:
hits30 = (np.abs(walks) >= 30).any(1)
hits30
Out[246]:
In [247]:
hits30.sum()
Out[247]:
In [248]:
crossing_times = (np.abs(walks[hits30]) >= 30).argmax(1)
crossing_times.mean()
Out[248]:
In [249]:
steps = np.random.normal(loc=0, scale=0.25,
size=(nwalks, nsteps))
In [250]:
steps
Out[250]: