pandas编程技巧2

1. pandas 的axis=1和0的区别

axis = 1 沿着横轴的方向（列操作，逐列）

axis = 0 沿着纵轴的方向 （行操作，逐行）

from pandas import DataFrame, Series
import pandas as pd

df = DataFrame([
    {"order_id":1001, "name":"wangwu"},
    {"order_id":1002, "name":"wangliu"},
    {"order_id":1003, "name":"zhangsan"},
    ])

df_2 = DataFrame([
    {"order_id":1004, "name":"wangwu"},
    {"order_id":1005, "name":"wangliu"},
    {"order_id":1006, "name":"zhangsan"},
    ])

df.index = [0,1,2]
print(df)
df_01 = df.drop("name", axis=1)
print("删除列 axis=1\n", df_01)
df_02 = df.drop(0, axis=0)
print("删除行 axis=0\n", df_02)
print("merge-----------，默认是列连接")
df_03 = df.merge(df_2, on="order_id", how = "left")
print(df_03)
print("concat----------默认是行连接")
df_04 = pd.concat([df, df_2])
print(df_04)
mean = df["order_id"].mean(axis = 0)
print(mean, type(mean))

-----------------------------result--------------------
   order_id      name
0      1001    wangwu
1      1002   wangliu
2      1003  zhangsan
删除列 axis=1
    order_id
0      1001
1      1002
2      1003
删除行 axis=0
    order_id      name
1      1002   wangliu
2      1003  zhangsan
merge-----------，默认是列连接
   order_id    name_x name_y
0      1001    wangwu    NaN
1      1002   wangliu    NaN
2      1003  zhangsan    NaN
concat----------默认是行连接
   order_id      name
0      1001    wangwu
1      1002   wangliu
2      1003  zhangsan
0      1004    wangwu
1      1005   wangliu
2      1006  zhangsan
1002.0 <class 'numpy.float64'>

2. astype()

   from pandas import DataFrame, Series
   
   df = DataFrame([
       {"order_id":1001, "name":"wangwu", "score":90},
       {"order_id":1002, "name":"wangliu", "score":89},
       {"order_id":1003, "name":"zhangsan", "score":70},
       ])
   print(df)
   print("*"*30)
   df["order_id"] = df["order_id"].astype(str)
   df["score"] = df["score"].astype(float)
   print(df)
   
   ------------------------result-------------------
      order_id      name  score
   0      1001    wangwu     90
   1      1002   wangliu     89
   2      1003  zhangsan     70
   ******************************
     order_id      name  score
   0     1001    wangwu   90.0
   1     1002   wangliu   89.0
   2     1003  zhangsan   70.0
   

3. dataframe筛选样本或特定列

   from pandas import DataFrame, Series
   import numpy as np
   df = DataFrame([
       {"order_id":1001, "name":"wangwu", "score":90, "address":np.nan},
       {"order_id":1002, "name":"wangliu", "score":89, "address":"nanjing"},
       {"order_id":1003, "name":"zhangsan", "score":70, "address":"beijing"},
       ])
   
   df_1 = df[["order_id", "score"]]
   print("根据列名选取指定的列：\n", df_1)
   
   df_2 = df[:2]
   print("选取指定的行:\n", df_2)
   
   df_3 = df[df["score"]>80]
   print("根据比较运算符筛选指定的行:\n", df_3)
   
   df_4 = df[df["name"]=="wangwu"]
   print(df_4)
   
   ------------------------result--------------------
   根据列名选取指定的列：
       order_id  score
   0      1001     90
   1      1002     89
   2      1003     70
   选取指定的行:
       order_id     name  score  address
   0      1001   wangwu     90      NaN
   1      1002  wangliu     89  nanjing
   根据比较运算符筛选指定的行:
       order_id     name  score  address
   0      1001   wangwu     90      NaN
   1      1002  wangliu     89  nanjing
   
      order_id    name  score address
   0      1001  wangwu     90     NaN
   
   

4. 缺失值的处理

   from pandas import DataFrame, Series
   import numpy as np
   
   df = DataFrame([
       {"order_id":1001, "name":"wangwu", "score":90, "address":np.nan},
       {"order_id":1002, "name":"wangliu", "score":89, "address":"nanjing"},
       {"order_id":1003, "name":"zhangsan", "score":70, "address":"beijing"},
       {"order_id":np.nan, "name":"zhangsan", "score":70, "address":np.nan},
       ])
   
   print("删除缺失值")
   df_1 = df.dropna(axis = 0, subset=["order_id"])
   print(df_1)
   print("留下特征值的数量>=n的行  (n是不为nan的值的数量)")
   df_2 = df.dropna(thresh=4)
   print(df_2)
   
   print("填充缺失值")
   print(df["address"].fillna("不详"))
   
   print("横向用缺失值前面的值替换缺失值")
   print(df.fillna(axis=1,method='ffill'))
   
   print("纵向用缺失值上面的值替换缺失值")
   print(df.fillna(axis=0,method='ffill'))
   
   print("使用默认值填充")
   values = {"order_id":1000, "address":"China"}
   print(df.fillna(value = values))
   
   #筛选出空值的index
   print(df[df["address"].isna()])
   
   
   #筛选非空值的index
   
   print(df[df["address"].notnull()])
   
   ---------------------result-----------------
   删除缺失值
      order_id      name  score  address
   0    1001.0    wangwu     90      NaN
   1    1002.0   wangliu     89  nanjing
   2    1003.0  zhangsan     70  beijing
   留下特征值的数量>=n的行  (n是不为nan的值的数量)
      order_id      name  score  address
   1    1002.0   wangliu     89  nanjing
   2    1003.0  zhangsan     70  beijing
   填充缺失值
   0         不详
   1    nanjing
   2    beijing
   3         不详
   Name: address, dtype: object
   横向用缺失值前面的值替换缺失值
     order_id      name score  address
   0     1001    wangwu    90       90
   1     1002   wangliu    89  nanjing
   2     1003  zhangsan    70  beijing
   3      NaN  zhangsan    70       70
   纵向用缺失值上面的值替换缺失值
      order_id      name  score  address
   0    1001.0    wangwu     90      NaN
   1    1002.0   wangliu     89  nanjing
   2    1003.0  zhangsan     70  beijing
   3    1003.0  zhangsan     70  beijing
   使用默认值填充
      order_id      name  score  address
   0    1001.0    wangwu     90    China
   1    1002.0   wangliu     89  nanjing
   2    1003.0  zhangsan     70  beijing
   3    1000.0  zhangsan     70    China
   
      order_id      name  score address
   0    1001.0    wangwu     90     NaN
   3       NaN  zhangsan     70     NaN
   
      order_id      name  score  address
   1    1002.0   wangliu     89  nanjing
   2    1003.0  zhangsan     70  beijing
   

5. 给dataframe添加列名和行序号

   from pandas import DataFrame, Series
   df = DataFrame([
       [100, "a", 30],
       [101, "b", 28],
       [102, "c", 26]
       ])
   
   df.columns = ["id", "name", "age"]
   df.index = [0,1,2]
   print(df)
   
   print("获取df的index数组", df.index.values)
   print("获取df的列名数组:", df.columns.values)
   -----------------------------result--------------------------------
       id name  age
   0  100    a   30
   1  101    b   28
   2  102    c   26
   获取df的index数组 ['user_id' 'age' 'age_new']
   获取df的列名数组: [2 1 0]
   

   6. 排序

      from pandas import DataFrame,Series
      
      import numpy as np
      
      df = DataFrame([
          {"order_id":1001, "name":"wangwu", "score":90, "address":np.nan},
          {"order_id":1002, "name":"wangliu", "score":89, "address":"nanjing"},
          {"order_id":1003, "name":"zhangsan", "score":70, "address":"beijing"},
          {"order_id":1004, "name":"zhangsan", "score":70, "address":np.nan},
          ])
      
      df = df.sort_values(by=["name", "order_id"], ascending=False)
      print(df)
      
      print(df.sort_values(by="score"))
      
      ------------------------------result-------------------
         order_id      name  score  address
      3      1004  zhangsan     70      NaN
      2      1003  zhangsan     70  beijing
      0      1001    wangwu     90      NaN
      1      1002   wangliu     89  nanjing
      
         order_id      name  score  address
      3      1004  zhangsan     70      NaN
      2      1003  zhangsan     70  beijing
      1      1002   wangliu     89  nanjing
      0      1001    wangwu     90      NaN