按 1 月的第一天过滤 pandas 股票行情数据帧

Question

抱歉，我对 Python 很陌生。

我有当前代码：

# Put data into a dataframe
df = pd.DataFrame(ZACKSP_raw_data)

""" Reformat dataframe data """    
# Change exchange from NSDQ to NASDAQ
df['exchange'] = df['exchange'].str.replace('NSDQ','NASDAQ')

# Change date format to DD/MM/YYYY
df['date'] = df['date'].dt.strftime('%d/%m/%Y')

# Round closing share price to 2 digits
df['close'] = df['close'].round(2)

# Filter data for Jan 
ZACKSP_data_StartOfJanYearMinus1 = df[df['date'] == '05/01/%s' % CurrentYearMinus1]

# Test
print(ZACKSP_data_StartOfJanYearMinus1.head())

它以以下格式返回数据：

现在我希望数组只保留 1 月份记录的第一个收盘价和 12 月份记录的最后一个收盘价（对于每个股票代码）。我曾想过尝试在一天中使用通配符，然后使用诸如 head() 或 tail() 之类的东西来实现这一目标，但我很挣扎。有什么想法吗？

Answer 1

所有日期时间都已排序的解决方案：

我认为每个ticker 的第一行和最后一行都需要concat 和drop_duplicates。

还需要为 years 添加新列，用于每年的第一个和最后一个值与代码。

df['year'] = pd.to_datetime(df['date']).dt.year

df1 = pd.concat([df.drop_duplicates(['ticker', 'year']), 
                 df.drop_duplicates(['ticker', 'year'], keep='last')])  

未排序datetimes 的更通用解决方案：

c = ['ticker','exchange','date','close']
df = pd.DataFrame({'date':pd.to_datetime(['2017-01-04','2017-01-12',
                                          '2017-01-05',
                           '2018-01-02','2018-12-27','2017-12-27',
                           '2018-01-05','2018-01-12','2017-01-05',
                           '2017-01-12','2018-12-22','2017-12-22']),
                   'close':[4.56,5.45,4.32,5.67,5.23,4.78,7.43,8.67,
                            9.32,4.73,2.42,3.45],
                   'ticker':['BA','BA','BA','BA','BA','BA',
                             'AAPL','AAPL','AAPL','AAPL','AAPL','AAPL'],
                    'exchange':['NYSE'] * 6 + ['NSDQ'] * 6}, columns=c)

print (df)
   ticker exchange       date  close
0      BA     NYSE 2017-01-04   4.56
1      BA     NYSE 2017-01-12   5.45
2      BA     NYSE 2017-01-05   4.32
3      BA     NYSE 2018-01-02   5.67
4      BA     NYSE 2018-12-27   5.23
5      BA     NYSE 2017-12-27   4.78
6    AAPL     NSDQ 2018-01-05   7.43
7    AAPL     NSDQ 2018-01-12   8.67
8    AAPL     NSDQ 2017-01-05   9.32
9    AAPL     NSDQ 2017-01-12   4.73
10   AAPL     NSDQ 2018-12-22   2.42
11   AAPL     NSDQ 2017-12-22   3.45

---

""" Reformat dataframe data """    
# Change exchange from NSDQ to NASDAQ
df['exchange'] = df['exchange'].str.replace('NSDQ','NASDAQ')

# Round closing share price to 2 digits
df['close'] = df['close'].round(2)

#sorting dates for first date per ticker is first day in Jan and last day in Dec
df = df.sort_values('date')

#extract years from dates
df['year'] = pd.to_datetime(df['date']).dt.year

#get first rows per tickers and year
df1 = df.drop_duplicates(['ticker', 'year'])
print (df1)
  ticker exchange       date  close  year
0     BA     NYSE 2017-01-04   4.56  2017
8   AAPL   NASDAQ 2017-01-05   9.32  2017
3     BA     NYSE 2018-01-02   5.67  2018
6   AAPL   NASDAQ 2018-01-05   7.43  2018

#get last row per year and ticker
df2 = df.drop_duplicates(['ticker', 'year'], keep='last')
print (df2)
   ticker exchange       date  close  year
11   AAPL   NASDAQ 2017-12-22   3.45  2017
5      BA     NYSE 2017-12-27   4.78  2017
10   AAPL   NASDAQ 2018-12-22   2.42  2018
4      BA     NYSE 2018-12-27   5.23  2018

---

#join DataFrames together and sorting if necessary
df = pd.concat([df1, df2]).sort_values(['ticker','date'])
print (df)
   ticker exchange       date  close  year
8    AAPL   NASDAQ 2017-01-05   9.32  2017
11   AAPL   NASDAQ 2017-12-22   3.45  2017
6    AAPL   NASDAQ 2018-01-05   7.43  2018
10   AAPL   NASDAQ 2018-12-22   2.42  2018
0      BA     NYSE 2017-01-04   4.56  2017
5      BA     NYSE 2017-12-27   4.78  2017
3      BA     NYSE 2018-01-02   5.67  2018
4      BA     NYSE 2018-12-27   5.23  2018

通过聚合first 和last 具有不同数据输出的另一种解决方案：

""" Reformat dataframe data """    
# Change exchange from NSDQ to NASDAQ
df['exchange'] = df['exchange'].str.replace('NSDQ','NASDAQ')

# Round closing share price to 2 digits
df['close'] = df['close'].round(2)

#sorting dates for first date per ticker is first day in Jan and last day in Dec
df = df.sort_values('date')

#extract years from dates
df['year'] = pd.to_datetime(df['date']).dt.year

df = (df.groupby(['ticker','year'])['close']
       .agg(['first','last'])
       .reset_index())
print (df)
  ticker  year  first  last
0   AAPL  2017   9.32  3.45
1   AAPL  2018   7.43  2.42
2     BA  2017   4.56  4.78
3     BA  2018   5.67  5.23

Answer 2

你想df.groupby('ticker')，然后按月份分组，过滤月份=='Dec'并获取tail()，过滤月份=='Jan'并获取head()，然后ungroup() .

（如果您发布可重现的数据，我将发布执行此操作的代码。）

阅读 pandas 文档 Group By: split-apply-combine范式，数据科学的关键范式之一。有关 SO 的示例，请参阅标记 split-apply-combine。