在填写调查之前，计算15分钟内的平均温度传感器（匹配时间戳+添加新列）

Question

我试图在包含调查数据（=PIT_da.xlsx）的excel后面添加新列。在这些列中，应计算并添加填写调查前15、30和60分钟内的平均传感器值（如温度）。传感器数据位于excel文件“IEQ_da.xlsx”（包括时间戳）中

我是这样开始的：

#import raw file
import pandas as pd
import numpy as np
dfSD = pd.read_excel('IEQ_da.xlsx')
dfPIT = pd.read_excel('PIT_da.xlsx')

#main aim: add after each survey result row in PIT_da.xlsx columns for the average values of the indoor environmental quality parameters in 15/30/60 minutes before submitting the survey

#Step 0: set both timestamp and submitdate to right datetime object
dfSD['timestamp'] =  pd.to_datetime(dfSD['timestamp'], format='%d%b%Y:%H:%M:%S.%f')
dfPIT['submitdate'] =  pd.to_datetime(dfPIT['submitdate'], format='%d%b%Y:%H:%M:%S.%f')

#Step 1: introduce arrays and set to numpy
array1 = dfSD[['timestamp']].to_numpy().ravel()
array2 = dfPIT[['submitdate']].to_numpy().ravel()
data_sensorID = dfSD[['devid']].to_numpy().ravel()
survey_sensorID = dfPIT[['PIT5']].to_numpy().ravel()Each survey has a timestamp (=submitdate) and should be matched to the sensor data at that timestamp. 

将时间转换为数字，以便能够计算15分钟/30分钟/60分钟内的差异

#Step 2: set timestamps to number and define a match 
from datetime import datetime
def timestamps(x) : 
    Timestamps = np.empty(x.size)
    for i in range(x.size) : 
        date = x[i]
        dt64 = np.datetime64(date)
        timestamp = (dt64 - np.datetime64('1970-01-01T00:00:00Z')) / np.timedelta64(1, 's')
        Timestamps[i] = timestamp
    return Timestamps

array1TS = timestamps(array1)
array2TS = timestamps(array2)

接下来，对每个调查提交时间和传感器时间戳（已经四舍五入到最接近的5分钟）进行匹配，包括来自相同传感器设备ID（=devid）和PIT5的情况（调查中询问附近传感器ID的问题）

#Step 3: define match with conditions: must be same timestamp and must have same sensor ID, by means of a matrix
Match = np.empty([array1TS.size, array2TS.size])
for i in range(array1TS.size) : 
    for j in range(array2TS.size):
        if (data_sensorID[i] == survey_sensorID[j]):
            if (array1TS[i] == array2TS[j]):
                Match[i,j] = 1;
            else: 
                Match[i,j] = 0;

现在，通过此匹配，应在“PIT_da.xlsx”中添加一个新列，该列的平均值为de IEQ_da.xlsx文件中“SENtemp”列的匹配时间戳前15分钟（含温度值）

问题是： 1.如何从“匹配”到在匹配的时间戳之前15分钟从该时间戳中选择所有行。 2.如何计算这些选定行的平均值（忽略空单元格）并将其放置在PIT_da.xlsx中的新列中（该新列应命名为“SENtemp_15”，以表示填写调查前15分钟内的温度）

为了便于参考，使用了一些数据行：

IEQ_da.xlsx

    import pandas as pd

df = pd.DataFrame({'timestamp' : ['14/04/2020  00:18:00', '14/04/2020  00:18:05', '14/04/2020  00:17:55', '14/04/2020  00:17:50' , '14/04/2020  00:17:40', '14/04/2020  00:17:40', '14/04/2020  00:17:20', '14/04/2020  00:17:20'], 'devid' : ['4', '2', '4', '2', '4' , '2' , '4' , '2'], 
                       'SENtemp' : ['20,2', '18,8', '20,1', '19', '20,2', '18,8', '20,1', '18,9']})
df

PIT_da.xlsx

import pandas as pd

df = pd.DataFrame({'submitdate' : ['14/04/2020  00:18:00', '14/04/2020  00:18:05'], 'PIT5' : ['4', '2'],
                   })
df

我希望有人愿意帮助我

Answer 1

你最初的两个步骤是相当无用的。您可以直接在dfPIT上使用apply来构建新列。最困难的部分是SENtemp是一个字符串列，其小数点为逗号，不能直接转换为浮点。可能代码：

delta = [15, 30, 60]  # delta in minutes

columns = [f'Average{i}' for i in delta]  # column names per delta values

dfPIT[columns] = dfPIT.apply(axis=1, func=lambda x: pd.Series(
    [dfSD.loc[(dfSD['timestamp']>x['submitdate'] - pd.Timedelta(i, 'T'))
              &(dfSD['timestamp']<=x['submitdate']), 'SENtemp']
     .str.replace(',','.').astype('float').mean() for i in delta],
    index=columns))

根据您的示例数据，它提供：

           submitdate PIT5  Average15  Average30  Average60
0 2020-04-14 00:18:00    4  19.614286  19.614286  19.614286
1 2020-04-14 00:18:05    2  19.512500  19.512500  19.512500