こういうCSVデータを時系列データとして扱いたい。
Timeデータは（秒）を表す。
Time,Temperature,Humidity,CO2,"Soil Moisture" 22,20.7,79,380,80 44,20.7,80,380,87 66,20.7,80,396,88 88,20.6,79,399,87 110,20.6,79,399,87 132,20.6,79,399,87 154,20.6,79,396,87 176,20.6,79,399,87 198,20.6,79,399,88
任意時刻の取得は以下で行う。
stackoverflow.com
>>> from datetime import datetime >>> datetime.strptime("2012-may-31 19:00", "%Y-%b-%d %H:%M") datetime.datetime(2012, 5, 31, 19, 0)
任意時刻のPOSIX時間への変換は以下で行う。
www.geeksforgeeks.org
>>> import datetime >>> import time >>> date_time = datetime.datetime(2021, 7, 26, 21, 20) >>> print("unix_timestamp => ", (time.mktime(date_time.timetuple()))) 1627314600.0
POSIX秒から時刻への変換は以下で行う。
stackoverflow.com
>>> from datetime import datetime
>>> datetime.fromtimestamp(1485714600).strftime("%A, %B %d, %Y %I:%M:%S")
'Sunday, January 29, 2017 08:30:00'上記をまとめると、
from datetime import datetime import time origin = datetime.strptime("2021-oct-01 12:00", "%Y-%b-%d %H:%M") origin_unix = time.mktime(origin.timetuple()) times = [datetime.fromtimestamp(origin_unix+t).strftime("%A, %B %d, %Y %I:%M:%S") for t in df['Time'].tolist()] tf = pd.DataFrame(times,columns=["Time"])
これでTime（秒）が時刻に変換できた。
>>> tf.head() Time 0 Friday, October 01, 2021 12:00:22 1 Friday, October 01, 2021 12:00:44 2 Friday, October 01, 2021 12:01:06 3 Friday, October 01, 2021 12:01:28 4 Friday, October 01, 2021 12:01:50
あとは元データと連結すればいい。
filtered = df.iloc[:, 1:] # Rows other than time df2 = pd.concat([tf,filtered],axis=1)
df2.head() Time Temperature Humidity CO2 Soil Moisture 0 Friday, October 01, 2021 12:00:22 20.7 79.0 380.0 80.0 1 Friday, October 01, 2021 12:00:44 20.7 80.0 380.0 87.0 2 Friday, October 01, 2021 12:01:06 20.7 80.0 396.0 88.0 3 Friday, October 01, 2021 12:01:28 20.6 79.0 399.0 87.0 4 Friday, October 01, 2021 12:01:50 20.6 79.0 399.0 87.0
時刻を含むPandasデータのプロットにはindexを指定する。
stackoverflow.com
df2.set_index('Time').plot()
