数据分析 第三周 (numpy数组的处理 ， numpy数组的运用与画图的结合)笔记

numpy 从文件中读取数据 loadtxt

#np.loadtxt(frame = "路径" ， dtype = 数据类型 , delimiter = "间隔符" ， skiprows = x 间隔行 ， unpack = True / False 是否转置)

import numpy as np

us_file_path = "F:/All date/US_video_data_numbers.csv"
uk_file_path = "F:/All date/GB_video_data_numbers.csv"

t1 = np.loadtxt(us_file_path,delimiter=",",dtype="int")
t2 = np.loadtxt(uk_file_path,delimiter=",",dtype="int")

print(t1)
print("*"*100)
print(t2)

#矩阵的转置(t.T)

tt = np.array(range(12)).reshape(2,6)
print(tt)
print("*"*100)
ts = tt.T
print(ts)

数组的索引和切片

#数组的索引 和 切片
import numpy as np

us_file_path = "F:/All date/US_video_data_numbers.csv"
uk_file_path = "F:/All date/GB_video_data_numbers.csv"

t1 = np.loadtxt(us_file_path,delimiter=",",dtype="int")
t2 = np.loadtxt(uk_file_path,delimiter=",",dtype="int")


#取连续的数据用 “:”

#取第2 - 3行    
print(t2[ 1 : 3 , : ])
print("-" * 100)
#取第2 - 3列    
print(t2[ : , 1 : 3 ])
print("-" * 100)

#取不连续的数据用 “[ ]”

#取第 2 ， 3 ，5 行
print(t2[[1 , 2 , 4],  :  ])
print("-" * 100)
#取第 2 ， 3 列
print(t2[ : , [1 , 2] ])


print("-" * 100)
#取多个不相邻的点
#选出来的结果是（0，0） （2，1） （2，3）
print(t2[[0,2,2],[0,1,3]])

用索引和切片修改数组的值

#用索引和切片修改数组的值

import numpy as np


#用某个条件对数组进行修改

t1 = np.array(range(10)).reshape(2,5)
t1[t1 < 5] = 2
print(t1)


print("-" * 100)
#用切片和索引修改
t1[ : , [1] ] = 3

print(t1)

用索引交换行和列

#交换行和列

import numpy as np

t1 = np.array(range(10)).reshape(2,5)

print(t1)
print("-" * 100)

#用类似于交换两个数的方式交换行和列
t1[ [0,1] , ] = t1[ [1,0] , ]

print(t1)

数据的合并

#合并两组数据
# np.hstack  水平拼接     np.vstack 竖直拼接  里面放 tuple
# np.zeros 生成一个全是 0 的数组  np.ones 生成一个全是 1 的数组 里面放 tuple


#任务：把两个国家的数据打标记后拼接


import numpy as np

us_file_path = "F:/All date/US_video_data_numbers.csv"
uk_file_path = "F:/All date/GB_video_data_numbers.csv"

text1 = np.loadtxt(us_file_path , delimiter = ',' , dtype = "int")
text2 = np.loadtxt(uk_file_path , delimiter = ',' , dtype = "int")


# shape[0] 获取行数 shape 获取某一维的大小
num1 = text1.shape[0]
num2 = text2.shape[0]

f1 = np.zeros((num1 , 1))
f2 = np.ones((num2 , 1))

text1 = np.hstack((text1 , f1))
text2 = np.hstack((text2 , f2))

text3 = np.vstack((text1 , text2)) 

text3 = text3.astype("int")

print(text3)

求矩阵某一维最大值位置

# 创建一个单位矩阵 np.eye

def printf(x):
    print(x)
    print("-" * 60)
import numpy as np

n1 = np.eye(7)
printf(n1)

#求某一维的 最大 最小值 的位置 np.armax np.armin(数组 ， axis = 维度)

t1 = np.argmax(n1 , axis = 0)
printf(t1)
n1[n1 == 1] = -1
printf(n1)

t2 = np.argmin(n1 , axis = 1)
printf(t2)

numpy 随机数的运用 (seed and randint)

# seed 产生一个随机数种子保证每次的随机数都相等
# randint(low , high , (type)) 产生一个有范围有形状的随机数数组

def printf(x):
    print(x)
    print("-" * 60)

import numpy.random as npr
# import numpy as np

npr.seed(10)

t = npr.randint(0 , 30 , (3 , 4))

printf(t)

#深拷贝和浅拷贝

#赋值操作都是浅拷贝 ， 一个会影响另一个
#但是复制操作是深拷贝 ， 不会产生影响

a = t.copy()

a[a <= 40] = 0

printf(a)
printf(t)

b = a

a[a == 0] = 1

printf(b)

#深拷贝互不影响
#浅拷贝藕断丝连

numpy 数组 nan 的计数 np.count_nonzeros

# np.nan != np.nan 利用这个性质求出 nan 的个数
# np.count_zero()

#统计 nan 的个数

import numpy as np
import numpy.random as npr

def printf(x):
    print(x)
    print("-" * 60)
    
t = npr.randint(0, 30, (3, 5))
printf(t)

# printf(type(np.nan))

t = t.astype("float")

t[t < 10] = np.nan    

printf(t)


print(np.count_nonzero(t!=t))

应用：用列平均值 替换数组中 nan 值

#用列平均值替换数组所有 nan

import numpy as np
# import numpy.random as npr

def printf(x):
    print(x)
    print("-" * 60)
    
    
def change(t1):
    
    for i in range(t1.shape[1]):
        nl = t1[ : , i ]
        
        cnt = np.count_nonzero(nl != nl)
        
        if(cnt != 0):
            nl[nl != nl] = nl[nl == nl].mean()
        
#t.mean(axis = x) 求平均值
#t.max(axis = x)  求最大值
#t.sum(axis = x)  求和    
        
        
        
    
if __name__ == "__main__":
    
    t = np.array(range(15)).reshape(3,5).astype("float")
    
    t[ 2 , 3 : ] = np.nan
    
    printf(t)
    
    change(t)
    
    printf(t)

numpy 数组提取文件数据然后做正方图

import numpy as np
from matplotlib import pyplot as plt
import matplotlib as mtb


def printf(x):
    print(x)
    print("-" * 60)
    
mtb.rcParams['font.sans-serif'] = ["SimHei"]
mtb.rcParams["axes.unicode_minus"] = False
plt.figure(figsize = (15,5) , dpi = 80)


us_file_path = "F:/All date/US_video_data_numbers.csv"
uk_file_path = "F:/All date/GB_video_data_numbers.csv"

t1 = np.loadtxt(us_file_path,delimiter=",",dtype="int")
t2 = np.loadtxt(uk_file_path,delimiter=",",dtype="int")

us = t1[ : , -1]
uk = t2[ : , -1]
#一定要处理离群数据
us = us[us <= 10000]
uk = us[us <= 10000]

# printf(max(us))

d = 500

num = (max(us) - min(us)) // d



plt.hist(us , [min(us) + i * d for i in range(num + 2)], color = "#FF7F50")


plt.xticks(range(min(us) , max(us) + 2 * d , d))
plt.yticks(range(1,1000,50))


#设置标签
plt.xlabel("评论数量")
plt.ylabel("人数")
plt.title("美国油管前1000评论数量统计")


# 设置网格 alpha 是清晰度
plt.grid(alpha = 0.3 , color = "#000000")


#显示图像
plt.show()

numpy提取文件数据做散点图

import numpy as np
from matplotlib import pyplot as plt
import matplotlib as mtb


def printf(x):
    print(x)
    print("-" * 60)
    
mtb.rcParams['font.sans-serif'] = ["SimHei"]
mtb.rcParams["axes.unicode_minus"] = False
plt.figure(figsize = (15,5) , dpi = 80)


us_file_path = "F:/All date/US_video_data_numbers.csv"
uk_file_path = "F:/All date/GB_video_data_numbers.csv"

t1 = np.loadtxt(us_file_path,delimiter=",",dtype="int")
t2 = np.loadtxt(uk_file_path,delimiter=",",dtype="int")

us_p = t1[ : , -1]
uk_p = t2[ : , -1]
us_l = t1[ : , 1]
uk_l = t2[ : , 1]

plt.scatter(us_p , us_l , label = "美国" , color = "#4B0082"  )

plt.xticks(range(min(us_p) , max(us_p) , 30000))


#设置标签
plt.xlabel("评论数")
plt.ylabel("喜爱数")
plt.title("美国评论数与喜爱数的关系")

#设置网格 alpha 是清晰度
# plt.grid(alpha = 0.3 , color = "#006400")

#设置图例
#两步 ， 先在折线中设置标签 ， 再用legend函数显示
plt.legend(loc = 2)