numpy

之前学的：

# -*- coding: utf-8 -*-
"""
Created on Fri May 29 11:51:15 2020

@author: Administrator
"""

import numpy as np
import random
t1=np.array([2,3,4,5,6])
t2=np.array(range(10))
print(t1)
print(type(t1))
print(t2)

t3=np.arange(11,dtype="int8")
print(t3)
print(type(t3))
#输出数据类型
print(t3.dtype)
#调整数据类型
t3.astype("int16")

t4=np.array([random.random() for i in range(10)])
print(t4)
#都可以，取两位小数
t5=t4.round(2)
t5=np.round(t4,3)
print(t5)

#形状
t6=np.arange(30)
print(t6)
print(t6.shape)

t7=t6.reshape(3,5,2)
print(t7)
print(t7.flatten())
print(t7.ravel())
print(t7.shape)
#转化为一维数组
t8=t7.flatten()
print(t8)

#加法
print("*"*20)
t9=np.arange(10)+10
print(t9*2)
#ru如果行一样或者列一样都可以计算
#从右往左看，维度一样就可以计算
#print(np.arange(16).reshape(4,4)*np.arange(4).reshape(1,4))
print(np.arange(16).reshape(4,2,2)-np.arange(4).reshape(2,2))
#转置
t10=np.arange(1,17).reshape(4,4)
print(t10)
print(t10.T)
print(t10.transpose())
    #相同不转置，相反则转置
print(t10.swapaxes(0,0))

#洲 0=x 1=y 2=z   axis
#求x坐标轴的平均值   那就是其他的坐标都一样，然后 x坐标不同的值相加
t11=np.arange(1,17).reshape(2,2,4)
for x in range(2):
    for y in range(2):
        for z in range(4):
            print("({0[0]},{0[1]},{0[2]})={0[3]}".format([x,y,z,t11[x][y][z]]))

print(t11)
print(t11.sum(axis=0))
print(t11.mean(axis=0))
print(t11.min(axis=0))
print(t11.max(axis=0))



#读取数据
#unpack=True转置
ft=np.loadtxt("F:\\a.csv",delimiter=",",dtype="int")

#切片
t12=np.arange(1,17).reshape(4,4)
print(t12)
print(t12[[0,2]][:])
#条件
np.where(t12<10,0,16)
print(np.where(t12<10,0,16))
#小于3换成3，大于8换成8
print(t12.clip(3,8))
#赋值nan
t12=t12.astype(float)
t12[0][0]=np.nan
print(t12)
print(t12)


#数据的拼接
t13=np.arange(1,9).reshape(2,4)
t14=np.arange(9,17).reshape(2,4)
print(t13)
print(t14)
print(np.vstack((t13,t14)))
print(np.hstack((t13,t14)))
#行与行交换
t15=t14[[0,1],:]=t14[[1,0],:]
print(t14)
print(t15)
#列与列交换
t16=t14[:,[0,1]]=t14[:,[1,0]]
print(t14)
print(t16)
print(t14[:,[1,2,0]])

#全为0
np.zeros(2,2)
np.ones(2,2)
#对角线全为1的方阵
np.eye(3)
#获取最大值，最小值的位置
t17=np.arange(1,16).reshape(1,15)
print(t17)
np.argmax(t17)
np.argmin(t17)

#随机数
np.random.rand()#均匀分布
np.random.randn()#正态分布
np.random.randint(a,b,(4,5))
np.random.uniform(a,b,(4,5))
np.random.normal()
np.random.seed(s)#加上这个，效果一样，每次产生的随机数不变
#别用=号    a=b别用  切片也会相互影响
#用a=b.copy()这样不影响
#不为0的个数
#计算nan个数
np.count_nonzero(b!=b)
np.isnan(b)
np.count_nonzero(np.isnan(b))
#nan和啥计算都是nan

#把nan换成平均数
np.random.seed(1)
t18=np.random.randint(1,20,(4,4))
t18=t18.astype("float")
t18[1,2]=np.nan
#竟然可以这样写
print(t18[t18==t18])
t19=t18.copy()
t19=t19.reshape(16)
np.where(np.isnan(t19),0,t19)
n=np.count_nonzero(t18==t18)
my_mean=np.sum(np.where(np.isnan(t19),0,t19))/n
t18=np.where(np.isnan(t18),my_mean,t18)
print(t18)

#二替换nan为均值

np.random.seed(1)
t18=np.random.randint(1,20,(4,4))
t18=t18.astype("float")
t18[1,2]=np.nan
print(t18)
def tr_nan(arr):
    t18=arr
    #t19=t18[t18==t18].copy().reshape(t18.shape[0]*t18.shape[1])
    t19=t18[t18==t18].copy()
    
    my_mean=t19.sum()/t19.shape[0]
    t18[np.isnan(t18)]=my_mean
    print(t18)
    return t18

#判断提取数据用
print(t18[t18>10])
#判断且修改数据
np.where(t18==t18,t18,0)


import numpy as np

t20=np.arange(1,5)
print(t20.shape)
t21=t20.copy().reshape(4,1)
print(t21)
print(t20*t21)