Skip to content Skip to sidebar Skip to footer

Add Data For The Missing Dates Based On Previous Hour Data In Pandas

I have a dataframe like below :- Missing dates date is for below: 2021-01-23 2021-01-25 I want to fill the rows for 2021-01-23 and 2021-01-25 also with the previous dates. Example

Solution 1:

This is a continuation of fill values

  • generate a DF that is combination of sampled hours and instances (df2)
  • this generates 15 rows as there are 3 times for instanceA and 2 times for instanceB across 3 dates (2+3)*3
  • then use same technique to fill both CPULoad and synthesized memload
  • tested against pandas 1.0.1 as well as 1.2.0
import pandas as pd
import io
import datetime as dt
import numpy as np
df = pd.read_csv(io.StringIO("""id  creTimestamp    CPULoad instnceId
0   2021-01-22 18:00:00 22.0    instanceA
1   2021-01-22 19:00:00 22.0    instanceA
2   2021-01-22 20:00:00 23.0    instanceB
3   2021-01-23 18:00:00 24.0    instanceA
4   2021-01-23 20:00:00 22.0    instanceA
5   2021-01-24 18:00:00 23.0    instanceB
6   2021-01-24 20:00:00 23.5    instanceA
"""), sep="\t", index_col=0)

df.creTimestamp = pd.to_datetime(df.creTimestamp)
df["memload"] = np.random.random(len(df))

# generate a DF for each time in instance in each date
df2 = (pd.merge(
    # for each time in instance
    df.assign(timestamp=df.creTimestamp.dt.time)
        .loc[:,["instnceId","timestamp"]]
        .drop_duplicates()
        .assign(foo=1),
    # for each date
    df.creTimestamp.dt.date.drop_duplicates().to_frame().assign(foo=1),
    on="foo"
).assign(creTimestamp=lambda dfa: dfa.apply(lambda r: dt.datetime.combine(r["creTimestamp"], r["timestamp"]), axis=1))
 .drop(columns="foo")
       # merge values back..
 .merge(df, on=["creTimestamp", "instnceId"], how="left")
)

# now get values to fill NaN
df2 = (df2.merge(df2.dropna().drop_duplicates(subset=["instnceId","timestamp"], keep="last"),
         on=["timestamp","instnceId"], suffixes=("","_pre"))
 .assign(CPULoad=lambda dfa: dfa.CPULoad.fillna(dfa.CPULoad_pre))
 .assign(memload=lambda dfa: dfa.memload.fillna(dfa.memload_pre))

)

output

    instnceId timestamp        creTimestamp  CPULoad    creTimestamp_pre  CPULoad_pre
0   instanceA  18:00:00 2021-01-22 18:00:00     22.0 2021-01-23 18:00:00         24.0
1   instanceA  18:00:00 2021-01-23 18:00:00     24.0 2021-01-23 18:00:00         24.0
2   instanceA  18:00:00 2021-01-24 18:00:00     24.0 2021-01-23 18:00:00         24.0
3   instanceA  19:00:00 2021-01-22 19:00:00     22.0 2021-01-22 19:00:00         22.0
4   instanceA  19:00:00 2021-01-23 19:00:00     22.0 2021-01-22 19:00:00         22.0
5   instanceA  19:00:00 2021-01-24 19:00:00     22.0 2021-01-22 19:00:00         22.0
6   instanceB  20:00:00 2021-01-22 20:00:00     23.0 2021-01-22 20:00:00         23.0
7   instanceB  20:00:00 2021-01-23 20:00:00     23.0 2021-01-22 20:00:00         23.0
8   instanceB  20:00:00 2021-01-24 20:00:00     23.0 2021-01-22 20:00:00         23.0
9   instanceA  20:00:00 2021-01-22 20:00:00     23.5 2021-01-24 20:00:00         23.5
10  instanceA  20:00:00 2021-01-23 20:00:00     22.0 2021-01-24 20:00:00         23.5
11  instanceA  20:00:00 2021-01-24 20:00:00     23.5 2021-01-24 20:00:00         23.5
12  instanceB  18:00:00 2021-01-22 18:00:00     23.0 2021-01-24 18:00:00         23.0
13  instanceB  18:00:00 2021-01-23 18:00:00     23.0 2021-01-24 18:00:00         23.0
14  instanceB  18:00:00 2021-01-24 18:00:00     23.0 2021-01-24 18:00:00         23.0

Post a Comment for "Add Data For The Missing Dates Based On Previous Hour Data In Pandas"