How to custom sort datetime column in pandas?

I have the following dataset.

import random
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import matplotlib.ticker as ticker
random.seed(0)

df = pd.DataFrame({'DATE_TIME': pd.date_range('2022-11-01', '2022-11-06 23:00:00', freq='20min'),
                   'ID': [random.randrange(1, 3) for n in range(430)]})
df['VALUE1'] = [random.uniform(110, 160) for n in range(430)]
df['VALUE2'] = [random.uniform(50, 80) for n in range(430)]
df['INSPECTION'] = df['DATE_TIME'].dt.day
# df['INSPECTION'] = df['INSPECTION'].replace(6, 1)
# df['INSPECTION'] = df['INSPECTION'].replace(3, 1)

df['MODE'] = np.select([df['INSPECTION'] == 1, df['INSPECTION'].isin([2, 3])], ['A', 'B'], 'C')
df['TIME'] = df['DATE_TIME'].dt.time
df['TIME'] = df['TIME'].astype('str')

df['TIMEINTERVAL'] = df.DATE_TIME.diff().astype('timedelta64[m]')
df['TIMEINTERVAL'] = df['TIMEINTERVAL'].fillna(0)


def to_day_period(s):
    bins = ['0', '06:00:00', '13:00:00', '18:00:00', '23:00:00', '24:00:00']
    labels = ['Nighttime', 'Daytime', 'Daytime', 'Nighttime', 'Nighttime']

    return pd.cut(
        pd.to_timedelta(s),
        bins=list(map(pd.Timedelta, bins)),
        labels=labels, right=False, ordered=False
    )


df['TIME_OF_DAY'] = to_day_period(df['TIME'])
df_monthly = df

I would like to sort the DATE_TIME column in a way that it starts from 12:00:00 and ends at 11:59:59. How can I achieve this? I tried shifting DATE_TIME object for 12 hours and applying np.argsort on VALUE1 column, but his did not work.
For ID=1 and INSPECTION=1, suppose I have the following data:

What I expect:

Use numpy.lexsort with dt.normalize and the 12h subtraction that you suggested (only keeping the time):

import numpy as np

order = np.lexsort([df['DATE_TIME'].sub(pd.Timedelta('12h')).dt.time,
                    df['DATE_TIME'].dt.normalize()])

out = df.iloc[order]

Using a function to benefit from sort_values' key parameter:

def custom_date_sorter(s):
    s = pd.to_datetime(s)
    return np.argsort(np.lexsort(
展开收缩
))

out = df.sort_values(by='DATE_TIME', key=custom_date_sorter)

Output:

            DATE_TIME  VALUE1
2 2022-11-01 12:00:00      40
3 2022-11-01 22:00:00      35
4 2022-11-01 23:00:00      70
0 2022-11-01 00:00:00      55
1 2022-11-01 11:00:00      45

Alternatively:

(df.assign(date=df['DATE_TIME'].dt.normalize(),
           time=df['DATE_TIME'].sub(pd.Timedelta('12h')).dt.time)
   .sort_values(by=['date', 'time'])
   .drop(columns=['date', 'time'])
)

this is a small example how you can solve your problem:

import pandas as pd
import numpy as np
from datetime import datetime

df = pd.DataFrame({
    "VALUE1": [1, 2, 3]
})

df.index = [datetime(2023, 3, 6, 8), datetime(2023, 3, 6, 13), datetime(2023, 3, 6, 14), ]

df = df.iloc[
    np.argsort(list(map(lambda x: f"{(x.hour - 12) % 12}-{x.minute}-{x.second}", df.index)))
]

input df looks like this

	VALUE1
2023-03-06 08:00:00	1
2023-03-06 13:00:00	2
2023-03-06 14:00:00	3

output like this

2023-03-06 13:00:00	2
2023-03-06 14:00:00	3
2023-03-06 08:00:00	1