Time Series Data

Pandas Basics

3 min read

Published Sep 29 2025, updated Aug 17 2026

21
0
0
0
PandasPython

What is Time Series Data?

  • A sequence of data points indexed (or labeled) by time.
  • Examples:
    • Daily stock prices
    • Hourly temperature readings
    • Monthly sales numbers
  • In Pandas, time series data is usually represented with a DatetimeIndex.




Creating DatetimeIndex

pd.date_range() creates a sequence of datetime values to use as an index.

Example:

import pandas as pd​index = pd.date_range(start='27/10/2019 00:00:00', periods=24, freq='h')print(index)

Output:

DatetimeIndex(['2019-10-27 00:00:00', '2019-10-27 01:00:00',               '2019-10-27 02:00:00', ..., '2019-10-27 23:00:00'],              dtype='datetime64[ns]', freq='h')

It generates 24 hourly timestamps starting at midnight on Oct 27, 2019. So, it’s basically one full day, hour by hour.


Parameters Used:

  • start='27/10/2019 00:00:00' → the first datetime in the range.
  • periods=24 → number of timestamps to generate.
  • freq='h' → frequency = hourly.

Other Common Parameters:

  • end → specify an end date instead of periods.
  • freq → frequency of steps. Examples:
    • 'B' — Business day frequency
    • 'D' — Calendar day frequency
    • 'h' — Hourly frequency
    • 'min' — Minutely frequency
    • 's' — Secondly frequency
    • 'ms' — Millisecond frequency
    • 'us' — Microsecond frequency
    • 'ns' — Nanosecond frequency
    • 'ME' — Month end frequency
    • 'MS' — Month start frequency
    • 'QE' — Quarter end frequency
    • 'QS' — Quarter start frequency
    • 'YE' — Year end frequency
    • 'YS' — Year start frequency
    • 'bh' — Business hour frequency
    • '2D' — Every 2 days
    • '3W' — Every 3 weeks
    • '1H30T' — Every 1 hour and 30 minutes
  • tz → set a timezone.

Creating DataFrame example:

import numpy as npimport pandas as pd​np.random.seed(3)df = pd.DataFrame(data={'Col1':np.random.randint(low=20,high=35,size=24) },                     index=pd.date_range(start='27/10/2019 00:00:00', periods=24,freq='h'))


Output:

                     Col12019-10-27 00:00:00    302019-10-27 01:00:00    282019-10-27 02:00:00    292019-10-27 03:00:00    232019-10-27 04:00:00    282019-10-27 05:00:00    282019-10-27 06:00:00    202019-10-27 07:00:00    252019-10-27 08:00:00    332019-10-27 09:00:00    232019-10-27 10:00:00    302019-10-27 11:00:00    312019-10-27 12:00:00    292019-10-27 13:00:00    292019-10-27 14:00:00    302019-10-27 15:00:00    252019-10-27 16:00:00    272019-10-27 17:00:00    262019-10-27 18:00:00    202019-10-27 19:00:00    242019-10-27 20:00:00    322019-10-27 21:00:00    332019-10-27 22:00:00    272019-10-27 23:00:00    34

You can then access the different parts such as hour, year etc of the index. Here we add them as extra columns:

df['Hour'] = df.index.hourdf['Minute'] = df.index.minutedf['Seconds'] = df.index.seconddf['Day'] = df.index.daydf['WeekDay_number'] = df.index.weekdaydf['WeekDay_text'] = df.index.day_name()df['Month_number'] = df.index.monthdf['Month_short'] = df.index.month_name()df['Year'] = df.index.yeardf['Year-Month'] = df.index.to_period('M')​# Print just the top 5 rowsprint(df.head())

Output:

                     Col1  Hour  Minute  Seconds  Day  WeekDay_number  \2019-10-27 00:00:00    30     0       0        0   27               6   2019-10-27 01:00:00    28     1       0        0   27               6   2019-10-27 02:00:00    29     2       0        0   27               6   2019-10-27 03:00:00    23     3       0        0   27               6   2019-10-27 04:00:00    28     4       0        0   27               6   ​                    WeekDay_text  Month_number Month_short  Year Year-Month  2019-10-27 00:00:00       Sunday            10     October  2019    2019-10  2019-10-27 01:00:00       Sunday            10     October  2019    2019-10  2019-10-27 02:00:00       Sunday            10     October  2019    2019-10  2019-10-27 03:00:00       Sunday            10     October  2019    2019-10  2019-10-27 04:00:00       Sunday            10     October  2019    2019-10





Accessing date parts from non-indexed columns:

df_date = pd.DataFrame(data={'Date':pd.date_range(start='27/10/2019 00:00:00', periods=24,freq='h'),                             'Col1':np.random.randint(low=20,high=35,size=24)})​print(df_date.head())

Creates a table with a date range that is just a standard column.



Output:

                 Date  Col10 2019-10-27 00:00:00    311 2019-10-27 01:00:00    202 2019-10-27 02:00:00    253 2019-10-27 03:00:00    244 2019-10-27 04:00:00    34

Access the date parts of the column:

df_date['Hour'] = df_date['Date'].dt.hourdf_date['Minute'] = df_date['Date'].dt.minutedf_date['Year'] = df_date['Date'].dt.year​print(df_date.head())

Add .dt before the date information when indicating the datetime data is a column.


Output:

                 Date  Col1  Hour  Minute  Year0 2019-10-27 00:00:00    31     0       0  20191 2019-10-27 01:00:00    20     1       0  20192 2019-10-27 02:00:00    25     2       0  20193 2019-10-27 03:00:00    24     3       0  20194 2019-10-27 04:00:00    34     4       0  2019





Converting to datetime

import pandas as pd​df = pd.DataFrame({    "Date": ["2025-01-01", "2025-01-02", "2025-01-03"],    "Sales": [100, 150, 200]})​df["Date"] = pd.to_datetime(df["Date"])

  • Converts strings to datetime objects.
  • Benefits: allows arithmetic, slicing, and resampling.

Setting datetime as index:

df.set_index("Date", inplace=True)print(df)

Output:

            SalesDate             2025-01-01    1002025-01-02    1502025-01-03    200

Now df.index is a DatetimeIndex.






Accessing Time Series Data

By specific date:

df.loc["2025-01-02"]

By date range:

df.loc["2025-01-01":"2025-01-02"]

By year, month, or day (via string slicing):

# all data in January 2025df.loc["2025-01"]​# all data in 2025df.loc["2025"]





Frequency and Resampling

Pandas supports regular time intervals (daily, monthly, yearly, etc.).

# Resample daily to monthly sumdf.resample("ME").sum()​# Resample monthly to daily meandf.resample("D").mean()

Common frequency strings:

  • Time-based
    • "s" → seconds
    • "min" → minutes
    • "h" → hours
    • "D" → calendar day
    • "B" → business day
  • Week-based
    • "W" → weekly (default Sunday end)
    • "W-MON" → weekly anchored to Monday
  • Month/Quarter/Year
    • "ME" → month end
    • "MS" → month start
    • "QE" → quarter end
    • "QS" → quarter start
    • "YE" → year end
    • "YS" → year start
  • Custom
    • "2h" → every 2 hours
    • "15min" → every 15 minutes
    • "7D" → every 7 days





Shifting and Lagging

Shift data forward/backward in time:

# lag by 1 daydf["Prev_Sales"] = df["Sales"].shift(1)​# lead by 1 daydf["Next_Sales"] = df["Sales"].shift(-1)

Useful for calculating differences or returns:

df["Sales_Change"] = df["Sales"] - df["Sales"].shift(1)​# Alternative way to do the same using diff function:df["Sales_Change_Alternative"] = df["Sales"].diff(periods=1) ​# Or as a percentage changedf["Sales_Percent_Change"] = df["Sales"].pct_change(periods=1)





Rolling and Moving Windows

Compute statistics over a rolling window:

# 3-day rolling meandf["Sales_RollingMean"] = df["Sales"].rolling(window=3).mean()​# 3-day rolling sumdf["Sales_RollingSum"] = df["Sales"].rolling(3).sum()





Time Series Arithmetic

Pandas allows operations on dates:

# add 1 daydf.index + pd.Timedelta(days=1)​# difference from first datedf.index - df.index[0]

Supports resampling, shifting, and alignment.

© 2025 SimpleSteps.guide
AboutFAQPoliciesContact
Pandas Basics | Time Series Data | SimpleSteps.guide