In time series analysis, a moving average is simply the average value of a certain number of previous periods.

An exponential moving average is a type of moving average that gives more weight to recent observations, which means it’s able to capture recent trends more quickly.

This tutorial explains how to calculate an exponential moving average for a column of values in a pandas DataFrame.

Example: Exponential Moving Average in Pandas

Suppose we have the following pandas DataFrame:

import pandas as pd#create DataFramedf = pd.DataFrame({'period': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
                   'sales': [25, 20, 14, 16, 27, 20, 12, 15, 14, 19]})#view DataFrame df

        period	sales
0	1	25
1	2	20
2	3	14
3	4	16
4	5	27
5	6	20
6	7	12
7	8	15
8	9	14
9	10	19

We can use the pandas.DataFrame.ewm() function to calculate the exponentially weighted moving average for a certain number of previous periods.

For example, here’s how to calculate the exponentially weighted moving average using the four previous periods:

#create new column to hold 4-day exponentially weighted moving averagedf['4dayEWM'] = df['sales'].ewm(span=4, adjust=False).mean()#view DataFrame df

        period	sales	4dayEWM
0	1	25	25.000000
1	2	20	23.000000
2	3	14	19.400000
3	4	16	18.040000
4	5	27	21.624000
5	6	20	20.974400
6	7	12	17.384640
7	8	15	16.430784
8	9	14	15.458470
9	10	19	16.875082

We can also use the matplotlib library to visualize the sales compared to the 4-day exponentially weighted moving average:

import matplotlib.pyplot as plt#plot sales and 4-day exponentially weighted moving average plt.plot(df['sales'], label='Sales')
plt.plot(df['4dayEWM'], label='4-day EWM')#add legend to plotplt.legend(loc=2)

Additional Resources