将每小时数据与 15 分钟数据合并答案

【问题标题】：Merge hourly data with 15 minute data将每小时数据与 15 分钟数据合并
【发布时间】：2022-11-13 16:13:29
【问题描述】：

通过使用非常低效的字符串操作（将分钟替换为零，即“06:15:00”->“06:00:00”），我能够将每小时数据与 15 分钟数据合并。我想知道是否有更优雅的方式来合并数据。

提前致谢！

import ccxt
import pandas as pd

ex = ccxt.binance({'enableRateLimit': True})

df_15m = pd.DataFrame(ex.fetch_ohlcv(symbol='BTC/USDT', timeframe='15m', limit=9), columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
df_1h = pd.DataFrame(ex.fetch_ohlcv(symbol='BTC/USDT', timeframe='1h', limit=3), columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])

df_15m = df_15m.loc[:, ['timestamp', 'close']]
df_1h = df_1h.loc[:, ['timestamp', 'close']]

df_15m['timestamp'] = pd.to_datetime(df_15m['timestamp'], unit='ms')
df_1h['timestamp'] = pd.to_datetime(df_1h['timestamp'], unit='ms')

df_15m['timestamp_h'] = df_15m['timestamp'].astype("string").str[:14] + '00:00'
df_1h.rename(columns={"timestamp": "timestamp_h"}, inplace=True)
df_1h['timestamp_h'] = df_1h['timestamp_h'].astype("string")

df_15m.rename(columns={"close": "close_15m"}, inplace=True)
df_1h.rename(columns={"close": "close_h"}, inplace=True)

print('Hourly Data:\n', df_1h, '\n')
print('15m Data:\n', df_15m, '\n')

df_merged = pd.merge(left=df_15m, right=df_1h, how='left', on=['timestamp_h'])

print('Merged Data:\n', df_merged, '\n')

输出：

Hourly Data:
            timestamp_h   close_h
0  2022-11-13 05:00:00  16853.68
1  2022-11-13 06:00:00  16684.45
2  2022-11-13 07:00:00  16731.94 

15m Data:
             timestamp  close_15m          timestamp_h
0 2022-11-13 05:00:00   16857.53  2022-11-13 05:00:00
1 2022-11-13 05:15:00   16849.16  2022-11-13 05:00:00
2 2022-11-13 05:30:00   16856.41  2022-11-13 05:00:00
3 2022-11-13 05:45:00   16853.68  2022-11-13 05:00:00
4 2022-11-13 06:00:00   16862.98  2022-11-13 06:00:00
5 2022-11-13 06:15:00   16807.98  2022-11-13 06:00:00
6 2022-11-13 06:30:00   16806.79  2022-11-13 06:00:00
7 2022-11-13 06:45:00   16684.45  2022-11-13 06:00:00
8 2022-11-13 07:00:00   16731.94  2022-11-13 07:00:00 

Merged Data:
             timestamp  close_15m          timestamp_h   close_h
0 2022-11-13 05:00:00   16857.53  2022-11-13 05:00:00  16853.68
1 2022-11-13 05:15:00   16849.16  2022-11-13 05:00:00  16853.68
2 2022-11-13 05:30:00   16856.41  2022-11-13 05:00:00  16853.68
3 2022-11-13 05:45:00   16853.68  2022-11-13 05:00:00  16853.68
4 2022-11-13 06:00:00   16862.98  2022-11-13 06:00:00  16684.45
5 2022-11-13 06:15:00   16807.98  2022-11-13 06:00:00  16684.45
6 2022-11-13 06:30:00   16806.79  2022-11-13 06:00:00  16684.45
7 2022-11-13 06:45:00   16684.45  2022-11-13 06:00:00  16684.45
8 2022-11-13 07:00:00   16731.94  2022-11-13 07:00:00  16731.94

【问题讨论】：

标签： python pandas dataframe bitcoin ccxt

【解决方案1】：

例子

df

data = [['2022-11-13 05:00:00', 16853.68],
        ['2022-11-13 06:00:00', 16684.45],
        ['2022-11-13 07:00:00', 16731.94]]
df = pd.DataFrame(data, columns=['timestamp_h', 'close_h'])

    timestamp_h             close_h
0   2022-11-13  05:00:00    16853.6800
1   2022-11-13  06:00:00    16684.4500
2   2022-11-13  07:00:00    16731.9400

df1

data1 = [['2022-11-13 05:00:00', 16857.53],
         ['2022-11-13 05:15:00', 16849.16],
         ['2022-11-13 05:30:00', 16856.41],
         ['2022-11-13 05:45:00', 16853.68],
         ['2022-11-13 06:00:00', 16862.98],
         ['2022-11-13 06:15:00', 16807.98],
         ['2022-11-13 06:30:00', 16806.79],
         ['2022-11-13 06:45:00', 16684.45],
         ['2022-11-13 07:00:00', 16731.94]]
df1 = pd.DataFrame(data1, columns=['timestamp', 'close'])

    timestamp           close
0   2022-11-13 05:00:00 16857.5300
1   2022-11-13 05:15:00 16849.1600
2   2022-11-13 05:30:00 16856.4100
3   2022-11-13 05:45:00 16853.6800
4   2022-11-13 06:00:00 16862.9800
5   2022-11-13 06:15:00 16807.9800
6   2022-11-13 06:30:00 16806.7900
7   2022-11-13 06:45:00 16684.4500
8   2022-11-13 07:00:00 16731.9400

第一的

在df1中创建timestamp_h列（时间戳数据类型）

df1.assign(timestamp_h=pd.PeriodIndex(df1['timestamp'], freq='1H').to_timestamp())

输出：

    timestamp           close       timestamp_h
0   2022-11-13 05:00:00 16857.5300  2022-11-13 05:00:00
1   2022-11-13 05:15:00 16849.1600  2022-11-13 05:00:00
2   2022-11-13 05:30:00 16856.4100  2022-11-13 05:00:00
3   2022-11-13 05:45:00 16853.6800  2022-11-13 05:00:00
4   2022-11-13 06:00:00 16862.9800  2022-11-13 06:00:00
5   2022-11-13 06:15:00 16807.9800  2022-11-13 06:00:00
6   2022-11-13 06:30:00 16806.7900  2022-11-13 06:00:00
7   2022-11-13 06:45:00 16684.4500  2022-11-13 06:00:00
8   2022-11-13 07:00:00 16731.9400  2022-11-13 07:00:00

第二

使timstamp_h成为df中的时间戳dtype

df.assign(timestamp_h=pd.to_datetime(df['timestamp_h']))

最后的

merge（完整代码包括第一秒)

(df1
 .assign(timestamp_h=(pd.PeriodIndex(df1['timestamp'], freq='H').to_timestamp()))
 .merge(df.assign(timestamp_h=pd.to_datetime(df['timestamp_h']))))

输出：

    timestamp           close       timestamp_h         close_h
0   2022-11-13 05:00:00 16857.5300  2022-11-13 05:00:00 16853.6800
1   2022-11-13 05:15:00 16849.1600  2022-11-13 05:00:00 16853.6800
2   2022-11-13 05:30:00 16856.4100  2022-11-13 05:00:00 16853.6800
3   2022-11-13 05:45:00 16853.6800  2022-11-13 05:00:00 16853.6800
4   2022-11-13 06:00:00 16862.9800  2022-11-13 06:00:00 16684.4500
5   2022-11-13 06:15:00 16807.9800  2022-11-13 06:00:00 16684.4500
6   2022-11-13 06:30:00 16806.7900  2022-11-13 06:00:00 16684.4500
7   2022-11-13 06:45:00 16684.4500  2022-11-13 06:00:00 16684.4500
8   2022-11-13 07:00:00 16731.9400  2022-11-13 07:00:00 16731.9400

【讨论】：