🪴LYC

🪴LYC

Resampling Time-Series Data

Dec 03, 2025

Resampling in Pandas allows you to change the frequency of your time-series data. This is particularly useful when you want to:

  • Aggregate data to a lower frequency (e.g., daily to monthly or yearly).
  • Upsample data to a higher frequency (e.g., monthly to daily).

It uses a combination of the resample() method and an aggregation function (e.g., mean, sum, count) to group and summarize data.

1. Basic Syntax

data.resample(rule).agg(function)
  • rule: A string specifying the new frequency (e.g., 'M' for monthly, 'A' for yearly).
  • agg(function): An aggregation method like mean, sum, or count.

2. Time-Series Offset Aliases

Common frequency strings for resampling:

  • 'D' : Daily
  • 'W' : Weekly
  • 'M' : Month-end
  • 'MS': Month-start
  • 'A' : Year-end
  • 'H' : Hourly

A full list is available in the Pandas documentation.

3. Example: Resampling in Action

Let’s start with a simple example:

Example Data

import pandas as pd
import numpy as np
 
# Create a sample DataFrame with a datetime index
date_rng = pd.date_range(start='2023-01-01', end='2023-01-10', freq='D')
data = pd.DataFrame(date_rng, columns=['Date'])
data['Sales'] = np.random.randint(100, 500, size=(len(date_rng)))
data = data.set_index('Date')
 
print(data)

Sample Output:

            Sales
Date             
2023-01-01   239
2023-01-02   455
2023-01-03   341
2023-01-04   408
2023-01-05   181
2023-01-06   134
2023-01-07   274
2023-01-08   120
2023-01-09   467
2023-01-10   331

4. Downsampling (Reduce Frequency)

Example: Resample to Weekly Frequency

# Resample to weekly frequency and calculate the sum of sales for each week
weekly_sales = data.resample('W').sum()
print(weekly_sales)

Sample Output:

            Sales
Date             
2023-01-01   239
2023-01-08  1988
2023-01-15   798

In this example:

  • The first row (2023-01-01) contains only the sales for that day.
  • The second row (2023-01-08) aggregates sales from 2023-01-02 to 2023-01-08.

Example: Resample to Monthly Frequency

monthly_sales = data.resample('M').mean()
print(monthly_sales)

5. Upsampling (Increase Frequency)

Upsampling is used to create higher-frequency data by filling in missing values.

Example: Resample to Daily Frequency

# Resample to hourly frequency (Upsample)
hourly_data = data.resample('H').asfreq()
print(hourly_data)

By default, the new rows will have NaN values for missing data.

Fill Missing Values

You can handle missing values during upsampling:

# Forward-fill missing data
hourly_data_filled = data.resample('H').ffill()
print(hourly_data_filled)

6. Custom Aggregation

You can pass custom aggregation functions like sum, mean, or even your own lambda functions:

Example: Weekly Aggregation with Multiple Metrics

# Aggregate with multiple functions
weekly_stats = data.resample('W').agg(['mean', 'sum'])
print(weekly_stats)

Output:

            Sales      
             mean   sum
Date                    
2023-01-01  239.0   239
2023-01-08  283.0  1988
2023-01-15  399.0   798

7. Plotting Resampled Data

Resampling is commonly used to prepare time-series data for visualization:

import matplotlib.pyplot as plt
 
# Plot original and resampled data
data['Sales'].plot(label='Daily Sales')
weekly_sales['Sales'].plot(label='Weekly Sales', linestyle='--')
plt.legend()
plt.show()

8. .dt for Datetime Operations

After resampling, you can use .dt to access specific attributes of a datetime object, such as month or year:

data['Month'] = data.index.month
data['Year'] = data.index.year

Graph View

Backlinks