App下載

pandas 常規(guī)操作,你學(xué) ’廢‘ 了么?

猿友 2021-01-06 17:51:12 瀏覽數(shù) (2565)
反饋

一 聚合函數(shù)

1. numpy、pandas使用的統(tǒng)計(jì)方式
在數(shù)組中經(jīng)常使用的聚合方式
data[['counts', 'ches_name']].agg([np.mean, np.std])
agg({'xx':np.mean, 'xx2':[np.sum, np.std]})
2. 在pandas或者numpy中沒有現(xiàn)成的函數(shù)可以使用,可以使用transform自定義函數(shù)

如: 將指定列的全部數(shù)據(jù) * 2

方式一
data['counts'].transform(lambda x: x*2)
方式二:按照函數(shù)內(nèi)既定的規(guī)則,進(jìn)行指定數(shù)據(jù)的操作
def transform_func(values):
	"""自定義函數(shù),定義數(shù)據(jù)操作規(guī)則"""
	return values*2
data['counts'].transform(transform_func)   # 一維
data1 = data.groupby(by='品牌')['銷售額'].transform(tran_func)  # 分組之后自定義聚合

推薦好課:Python 自動(dòng)化辦公

二 透視表 - pivot_table

源碼參數(shù)分析
def pivot_table(
    data,             # Dataframe,對(duì)哪張表進(jìn)行操作
    values=None,      # 顯示的字段
    index=None,       # 行分組鍵,可以是數(shù)組,列表,如果是數(shù)組,必須有一樣的長(zhǎng)度
    columns=None,      # 列分組鍵
    aggfunc="mean",    # 聚合函數(shù), 默認(rèn)是mean
    fill_value=None,   # 填充空值, 將為Nan的值填充為對(duì)應(yīng)的值
    margins=False,     # 匯總開關(guān),默認(rèn)是False
    dropna=True, 
    margins_name="All", # 匯總的列或者行的bolumns,可以指定修改名稱
    observed=False,
1、index: 行分組鍵,分完組后,分組鍵的取值在行索引的位置上
pd.pivot_table(data, index=['order_id', 'dishes_name'], aggfunc=[np.mean, np.sum], values=['add_inprice', 'counts'])
                                mean                sum       
                         add_inprice counts add_inprice counts
order_id dishes_name                                          
137      農(nóng)夫山泉NFC果汁100%           0      1           0      1
         涼拌菠菜                      0      1           0      1
         番茄燉牛腩\r\n                 0      1           0      1
         白飯/小碗                     0      4           0      4
         西瓜胡蘿卜沙拉                   0      1           0      1
...                              ...    ...         ...    ...
1323     番茄燉秋葵                     0      1           0      1
         芝士燴波士頓龍蝦                  0      1           0      1
         芹黃鱔絲                      0      1           0      1
         蒜蓉生蠔                      0      1           0      1
         谷稻小莊                      0      1           0      1
[2778 rows x 4 columns]

2、columns: 列分組鍵,分完組后,分組鍵的取值在列索引上

pd.pivot_table(data, columns= ['order_id', 'amounts'], aggfunc=[np.mean, np.sum], values=['add_inprice', 'counts'])
# 列分組鍵,可以說是行分組鍵的轉(zhuǎn)置
            mean                                ...  sum                        
order_id    137                           165   ... 1323                        
amounts      1    6    26   27   35   99   9    ...  39  49  58  65  78  80  175
add_inprice  0.0  0.0  0.0  0.0  0.0  0.0  0.0  ...    0   0   0   0   0   0   0
counts       4.0  1.0  1.0  1.0  1.0  1.0  1.5  ...    1   1   1   1   1   1   1
[2 rows x 4956 columns]
3、結(jié)合使用
# aggfunc 聚合函數(shù)
# fill_value 為空的,怎么顯示,默認(rèn)為Nan
# margins 匯總,默認(rèn)是不匯總
# margins_name 匯總列或者行字段名稱,默認(rèn)為all
pd.pivot_table(data, index=['dishes_name'], columns='order_id', values='counts', aggfunc=np.sum, fill_value=0, margins=True, margins_name='總')
dishes_name   42度海之藍(lán)   北冰洋汽水   38度劍南春   50度古井貢酒  ...  黃油曲奇餅干  黃花菜炒木耳  黑米戀上葡萄     總
order_id                                         ...                              
137                0        0        0        0  ...       0       0       0     9
165                0        0        1        0  ...       0       1       0    21
166                0        0        0        0  ...       0       0       0     7
171                0        0        0        0  ...       0       0       0    10
177                0        0        0        0  ...       0       0       0     4
...              ...      ...      ...      ...  ...     ...     ...     ...   ...
1314               0        0        1        0  ...       0       0       0    12
1317               0        0        0        0  ...       0       0       0    18
1319               0        0        0        0  ...       0       0       0     9
1323               0        0        1        0  ...       0       0       0    15
總                  5       45        6        5  ...       5      15      18  3088

推薦好課:Python 自動(dòng)化管理

三 交叉表-crosstab

def crosstab(
    index,   # 行分組鍵
    columns,  # 列分組鍵
    values=None,   # 顯示的字段
    rownames=None,    # 行name
    colnames=None,     # 列name
    aggfunc=None,      # 聚合函數(shù)
    margins=False,      # 匯總
    margins_name: str = "All",   # 匯總列或者行的名稱
    dropna: bool = True,
    normalize=False,
基本語(yǔ)法
pd.crosstab(index = data['dishes_name'], columns=data['order_id'], values=data['counts'], aggfunc = np.sum)
dishes_name   42度海之藍(lán)   北冰洋汽水   38度劍南春   ...  黃油曲奇餅干  黃花菜炒木耳  黑米戀上葡萄
order_id                                ...                        
137              NaN      NaN      NaN  ...     NaN     NaN     NaN
165              NaN      NaN      1.0  ...     NaN     1.0     NaN
166              NaN      NaN      NaN  ...     NaN     NaN     NaN
171              NaN      NaN      NaN  ...     NaN     NaN     NaN
177              NaN      NaN      NaN  ...     NaN     NaN     NaN
...              ...      ...      ...  ...     ...     ...     ...
1309             NaN      NaN      NaN  ...     NaN     NaN     NaN
1314             NaN      NaN      1.0  ...     NaN     NaN     NaN
1317             NaN      NaN      NaN  ...     NaN     NaN     NaN
1319             NaN      NaN      NaN  ...     NaN     NaN     NaN
1323             NaN      NaN      1.0  ...     NaN     NaN     NaN
[278 rows x 156 columns]

四 表格合并

1、每個(gè)表的列都相同,pd.concat((df1, df2, df3 … ))

axis = 0 : 縱向合并axis = 1:橫向合并,索引對(duì)應(yīng)合并

函數(shù)源碼
def concat(
    objs: Union[Iterable["NDFrame"], Mapping[Label, "NDFrame"]], # 傳入的是Df格式
    axis=0,          # 進(jìn)行合并的方向
    join="outer",    # 默認(rèn)使用的外連接
    ignore_index: bool = False,  # 重置排序索引
    keys=None,
    levels=None,
    names=None,
    verify_integrity: bool = False,
    sort: bool = False,
    copy: bool = True,
left = pd.DataFrame({'key1': ['K0', 'K0', 'K1', 'K3'],
                     'key2': ['K0', 'K1', 'K0', 'K1'],
                         'A': ['A0', 'A1', 'A2', 'A3'],
                         'B': ['B0', 'B1', 'B2', 'B3']})right = pd.DataFrame({'key1': ['K0', 'K1', 'K1', 'K2'],                       'key2': ['K0', 'K0', 'K0', 'K0'],                          'C': ['C0', 'C1', 'C2', 'C3'],                          'D': ['D0', 'D1', 'D2', 'D3']}) pd.concat((left, right), axis = 0, join = 'inner') # 指定使用內(nèi)連接,進(jìn)行合并,默認(rèn)使用的是outer pd.concat((left, right), axis = 1, join = 'inner')

2、 表合并,解決行索引沒有意義情況下,數(shù)據(jù)行不匹配問題(解決concat橫向拼接問題)

def merge(
    left,                # 左表
    right,               # 右表
    how: str = "inner",    # 默認(rèn)是內(nèi)連接,
    on=None,               # 必須是兩張表中有公共的主鍵,才能作為主鍵
    left_on=None,          # 左表主鍵
    right_on=None,         # 右表主鍵
    left_index: bool = False,
    right_index: bool = False,
    sort: bool = False,
    suffixes=("_x", "_y"),
    copy: bool = True,
    indicator: bool = False,
    validate=None,

(1) 兩表中有相同的主鍵

on 連接的主鍵,兩表中共有的主鍵
how 連接的方式,默認(rèn)使用的是內(nèi)連接
outer外連接,返回全部     inner內(nèi)連接返回等值連接     left以左表為主     right以右表為主
pd.merge(left, right, on='key1', how='outer')
  key1 key2_x    A    B key2_y    C    D
0   K0     K0   A0   B0     K0   C0   D0
1   K0     K1   A1   B1     K0   C0   D0
2   K1     K0   A2   B2     K0   C1   D1
3   K1     K0   A2   B2     K0   C2   D2
4   K3     K1   A3   B3    NaN  NaN  NaN
5   K2    NaN  NaN  NaN     K0   C3   D3
多個(gè)相同主鍵連接
pd.merge(left, right, on=['key1', 'key2'], how='outer')
  key1 key2    A    B    C    D
0   K0   K0   A0   B0   C0   D0
1   K0   K1   A1   B1  NaN  NaN
2   K1   K0   A2   B2   C1   D1
3   K1   K0   A2   B2   C2   D2
4   K3   K1   A3   B3  NaN  NaN
5   K2   K0  NaN  NaN   C3   D3

(2) 兩表中沒有相同的主鍵

left_on   : 指定左表中的主鍵
right_on  : 指定右表中的主鍵
pd.merge(left, right, left_on = 'key1', right_on = 'key2', how='outer')
  key1_x key2_x   A   B key1_y key2_y    C    D
0     K0     K0  A0  B0     K0     K0   C0   D0
1     K0     K0  A0  B0     K1     K0   C1   D1
2     K0     K0  A0  B0     K1     K0   C2   D2
3     K0     K0  A0  B0     K2     K0   C3   D3
4     K0     K1  A1  B1     K0     K0   C0   D0
5     K0     K1  A1  B1     K1     K0   C1   D1
6     K0     K1  A1  B1     K1     K0   C2   D2
7     K0     K1  A1  B1     K2     K0   C3   D3
8     K1     K0  A2  B2    NaN    NaN  NaN  NaN
9     K3     K1  A3  B3    NaN    NaN  NaN  NaN

(3) 更改表格名稱的方法

left.rename(columns={'key1': 'key11111'}, inplace=True)
print(left)
  key11111 key2   A   B

(4) 重疊合并,目的是將殘缺的表,合并為完整的表df1.combine_first(df2)

主表.combine_first(附表)
dict1 = {'ID':[1,2,3,4,5,6,7,8,9],
    'System':['W10','w10',np.nan,'w10',np.nan,np.nan,'w7','w7','w8']}dict2 = {'ID':[1,2,3,4,5,6,7,8,9],     'System':[np.nan,np.nan,'w7','w7','w7','w7','w8',np.nan,np.nan]} df1 = pd.DataFrame(dict1) df2 = pd.DataFrame(dict2) print(df1,df2) # 誰(shuí)在前,為主表,主表中沒有的補(bǔ)全,有的值,不動(dòng) print(df1.combine_first(df2))     ID System 0 1 W10 1 2 w10 2 3 w7 3 4 w10 4 5 w7 5 6 w7 6 7 w7 7 8 w7 8 9 w8

?推薦好課:Python3進(jìn)階:數(shù)據(jù)分析及可視化

0 人點(diǎn)贊