Turning a time into hours and minutes

A clock time such as 22:20 becomes two numeric features, the hour and the minute, because the text itself cannot be compared or added.

Last updated: 12 Oct, 2026 · pandas 3.0 · scikit-learn 1.9 · openpyxl 3.1

Splitting a date into columns turned one date into three numbers. The two time columns need the same treatment and one extra step, because some of the arrival times carry a date on the end: 01:10 22 Mar means the flight landed the next day.

Working on fields of the same kind together

Dep_Time and Arrival_Time hold the same kind of value and need the same recipe, so they are handled one after the other rather than in the order the columns happen to sit in. That habit is what keeps a long feature pass readable.

import pandas as pd

BASE = ("https://raw.githubusercontent.com/krishnaik06/5-Days-Live-EDA-and-Featur"
        "e-Engineering/main/Flight%20Prediction/")
train = pd.read_excel(BASE + "Data_Train.xlsx")   # 10683 rows, Price included
test = pd.read_excel(BASE + "Test_set.xlsx")      # 2671 rows, no Price
flights = pd.concat([train, test], ignore_index=True)

Cutting the trailing date off an arrival time

Splitting on a space and keeping the first piece leaves 01:10 from 01:10 22 Mar, and leaves 13:15 untouched, because a value with no space splits into a list of one.

flights['Arrival_Time'] = flights['Arrival_Time'].str.split(' ').str[0]

Splitting the clock time on the colon

The hour is before the colon and the minute is after it. Both pieces are cast, because '05' and '5' are different strings and the same number.

clock = flights['Arrival_Time'].str.split(':')
flights['Arrival_hour'] = clock.str[0].astype(int)
flights['Arrival_min'] = clock.str[1].astype(int)
flights = flights.drop(columns='Arrival_Time')

Building four columns out of two time columns

print("raw arrival times:", flights['Arrival_Time'].head(4).tolist())
carry_a_date = flights['Arrival_Time'].str.contains(' ')
print("rows carrying a date:", int(carry_a_date.sum()), "of", len(flights))

flights['Arrival_Time'] = flights['Arrival_Time'].str.split(' ').str[0]
for source, prefix in [('Arrival_Time', 'Arrival'), ('Dep_Time', 'Dept')]:
    clock = flights[source].str.split(':')
    flights[prefix + '_hour'] = clock.str[0].astype(int)
    flights[prefix + '_min'] = clock.str[1].astype(int)
    flights = flights.drop(columns=source)

print(flights[['Arrival_hour', 'Arrival_min', 'Dept_hour', 'Dept_min']].head(4).to_string())
print("\nhour range:", flights["Dept_hour"].min(), "to", flights["Dept_hour"].max())
print("minute values:", sorted(flights["Dept_min"].unique().tolist()))

raw arrival times: ['01:10 22 Mar', '13:15', '04:25 10 Jun', '23:30']
rows carrying a date: 5401 of 13354
   Arrival_hour  Arrival_min  Dept_hour  Dept_min
0             1           10         22        20
1            13           15          5        50
2             4           25          9        25
3            23           30         18         5

hour range: 0 to 23
minute values: [0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55]

• 5401 of the 13354 arrival times carry a date. Those are the overnight flights. Without the first split, '01:10 22 Mar'.split(':') would hand '10 22 Mar' to the cast and it would raise.
• The loop does the same four lines twice. Two columns of the same kind get one recipe, which is the habit the chapter is built on.
• Hours run 0 to 23 and minutes take twelve values, all multiples of five. Airlines publish times on a five minute grid, and a model sees that as twelve levels rather than sixty.

The same split with to_datetime

A clock time is a time, so pandas can parse it and hand back the parts. The result is identical, and the parse also refuses a value that is not a time instead of producing a wrong number in silence.

clock = pd.to_datetime(flights['Dep_Time'], format='%H:%M')
flights['Dept_hour'] = clock.dt.hour
flights['Dept_min'] = clock.dt.minute

import matplotlib.pyplot as plt

clock = pd.to_datetime(flights['Dep_Time'], format='%H:%M')
by_split = flights['Dep_Time'].str.split(':').str[0].astype(int)
print("the two routes agree:", bool((clock.dt.hour == by_split).all()))

per_hour = clock.dt.hour.value_counts().sort_index()
print("busiest hour:", int(per_hour.idxmax()), "with", int(per_hour.max()), "departures")
print("quietest hour:", int(per_hour.idxmin()), "with", int(per_hour.min()), "departures")

fig, ax = plt.subplots(figsize=(7.5, 3.4))
ax.bar(per_hour.index, per_hour.values, color='#3a6fd8')
ax.set_xticks(range(0, 24, 2))
ax.set_xlabel('departure hour'); ax.set_ylabel('flights')
ax.set_title('Departures by hour of the day')
fig.tight_layout()

the two routes agree: True
busiest hour: 9 with 1152 departures
quietest hour: 3 with 30 departures

• The hour from the split and the hour from the parse match on every row.
• The counts carry a pattern. The quietest hour, three in the morning, has 30 departures and the busiest, nine, has 1152, so the hour is information a model can use.
• The chart also shows why the next lesson exists. Hour 23 and hour 0 are next to each other on a clock and at opposite ends of this axis.

Splitting the string compared with parsing the time

Where hour and minute features are used

• Fare and demand models. A six in the morning departure and a six in the evening departure sell at different prices.
• Server and traffic logs, where the hour of the request is often the strongest single feature in the table.
• Shift and staffing data, where the minute rarely matters and the hour always does.

Related

• Previous: Splitting a date into columns
• Next: Turning a duration string into minutes
• See also: Cyclical encoding for hours and weekdays
• Reference: Time and date components in the pandas docs