Tutorial

Main differences between matplotlib, seaborn, and plotly

Compare Python's top visualization libraries: Matplotlib, Seaborn, and Plotly. Learn when to use each library, their strengths, limitations, and best practices for different visualization needs.

Which Python libraries can you use for data visualization? In which cases do you use each of them?

  1. Matplotlib for highly customizable plots.
  2. Seaborn for automated plots based on Matplotlib.
  3. Plotly for interactive plots.

How do you import these libraries into the Python environment?

You must import a sub-library containing the main plotting functions for some of them.

import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px

Enough theory! Let’s look at some examples using the same plot for each library.

Data

Using the famous gapminder dataset, we will plot data from countries in 2007.

import pandas as pd
df = pd.read_excel('gapminder.xlsx')
Basic raw data plot of gapminder dataset using Python visualization libraries
F1. Basic raw data plot from gapminder dataset

Scatter plot

Matplotlib

Matplotlib creates a simple scatter without labels on the axes.

plt.scatter(x='gdpPercap', y='lifeExp', data=df)
Simple scatter plot created with Matplotlib showing GDP per capita vs life expectancy
F2. Matplotlib scatter plot of GDP per capita versus life expectancy

Seaborn

On the other hand, seaborn creates a scatter plot with labels on the axes. And even adds some borders to the points.

sns.scatterplot(x='gdpPercap', y='lifeExp', data=df)
Scatter plot with axes labels and styled points, created using Seaborn
F3. Seaborn scatter plot with labeled axes and styled points

Now, what if you wanted to color the points by continent?

With matplotlib, you get a fatal error because it expects a column with names of colors, not a column with names of continents.

plt.scatter(x='gdpPercap', y='lifeExp', data=df, c='continent')

# ValueError: Invalid RGBA argument: 'Asia'

On the other hand, seaborn automates creating a scatter plot with a color-coded legend for the continent column, using the hue parameter.

sns.scatterplot(x='gdpPercap', y='lifeExp', data=df, hue='continent')
Color-coded scatter plot by continent using Seaborn's hue parameter
F4. Seaborn scatter plot color-coded by continent

Now, how can you identify the country represented by each point?

This process is not automatic when using matplotlib, nor is it when using seaborn.

You’d need to create a horrible for loop to add the country names to the points, ending up with a cluttered plot.

sns.scatterplot(x='gdpPercap', y='lifeExp', data=df, hue='continent')

for idx, data in df.iterrows():
    plt.text(x=data['gdpPercap'], y=data['lifeExp'], s=data['country'])
Seaborn scatter plot with overlaid country names creating a cluttered appearance
F5. Cluttered scatter plot with country names using Seaborn and Matplotlib

Yes, you can combine seaborn with matplotlib to add new elements to the plot because seaborn is built on top of matplotlib.

Also, you can add new elements to the plot using matplotlib functions in a new line of code—for example, a title, labels, or a legend.

sns.scatterplot(x='gdpPercap', y='lifeExp', data=df, hue='continent')

for idx, data in df.iterrows():
    plt.text(x=data['gdpPercap'], y=data['lifeExp'], s=data['country'])

plt.title('Cluttered plot')
Highly cluttered scatter plot with country names and a title, showcasing the drawbacks of excessive labeling
F6. Highly cluttered plot with country names and added title

How can we declutter the mess of country names?

Plotly

Instead of adding the country names to the plot, you can hover over the points to see the country names.

px.scatter(df, x='gdpPercap', y='lifeExp', color='continent', hover_name='country')
Interactive Plotly scatter plot with hoverable points showing country names
F7. Interactive Plotly scatter plot with hoverable data points

I love plotly because it’s interactive, and you can zoom in, zoom out, click on the legend to hide some countries, and even save the plot as an image.

But don’t get me wrong, plotly is not perfect.

If I wanted a highly customized plot, I’d use matplotlib because it’s more flexible, and you can control every plot detail.

These three libraries contain the necessary functions to create any other type of plot. Just choose the one that best suits your needs.

To learn more about these libraries, visit the official documentation:

Visit their examples gallery to see what you can do with them and replicate the plots in your projects with your data.

Conclusions

  1. If you are a beginner, use plotly for simple plots.
  2. If you are an intermediate user, use seaborn with matplotlib for more complex plots.
  3. If you are an advanced user, use matplotlib for highly customizable plots.