Altair is a declarative statistical visualization library in Python, designed to create interactive and customizable visualizations with ease.

Built on the powerful Vega and Vega-Lite visualization grammars, Altair provides an intuitive syntax for building complex data visualizations, making it particularly useful for data science projects.

This guide explores Altair’s features with examples that can help you understand how to integrate it effectively into your data analysis workflow.

Getting Started with Altair

To use Altair, start by installing it with the following command:

pip install altair

Once installed, you can start building visualizations. Altair uses data frames to create its visualizations, and it works seamlessly with Pandas.

import altair as alt
import pandas as pd

Basic Altair Concepts and Syntax

Altair is declarative, meaning you define what you want rather than how to draw it. The syntax focuses on connecting a chart type to a data frame, specifying encodings, and adding additional chart properties.

chart = alt.Chart(data).mark_type().encode(
    x='x_column:Q',
    y='y_column:Q',
    color='color_column:N'
)

• mark_type(): Defines the chart type (e.g., mark_bar, mark_line, etc.).
• encode(): Defines data mappings to visual properties like x, y, color, etc.
• Q: Quantitative data.
• N: Nominal (categorical) data.
• O: Ordinal (ordered categorical) data.
• T: Temporal data.

Example 1: Creating a Simple Bar Chart

Let’s start with a simple bar chart that visualizes car counts per category. We’ll use the popular Iris dataset to create a bar chart showing counts of each species.

import altair as alt
import pandas as pd
from vega_datasets import data

# Load the dataset
iris = data.iris()

# Create a bar chart
chart = alt.Chart(iris).mark_bar().encode(
    x='species:N',
    y='count():Q',
    color='species:N'
).properties(
    title='Count of Each Species in the Iris Dataset'
)

# Save the chart as an HTML file
chart.save('chart.html')

In this example:

• mark_bar() specifies a bar chart.
• x='species:N' defines the x-axis with the categorical species data.
• y='count():Q' counts occurrences on the y-axis.
• color='species:N' colors each bar by species.

Output:

Example 2: Scatter Plot with Tooltips and Interactivity

Altair supports interactive visualizations. Here, we’ll create a scatter plot with tooltips for data points, comparing sepal length and sepal width.

import altair as alt
import pandas as pd
from vega_datasets import data

# Load the Iris dataset
iris = data.iris()

# Create a scatter plot
chart = alt.Chart(iris).mark_circle(size=60).encode(
    x='sepalLength:Q',
    y='sepalWidth:Q',
    color='species:N',
    tooltip=['species:N', 'sepalLength:Q', 'sepalWidth:Q']
).properties(
    title='Sepal Length vs Sepal Width with Tooltips'
).interactive()

# Save the chart as an HTML file
chart.save('chart1.html')

Output:

Explanation:

• mark_circle(size=60): Specifies a scatter plot with circles.
• color='species:N': Colors each point by species.
• tooltip: Displays sepal length, width, and species on hover.
• .interactive(): Enables zooming and panning for the plot.

Example 3: Line Chart for Time Series Data

To illustrate time-series data, let’s plot stock prices using Altair’s mark_line().

import altair as alt
import pandas as pd
from vega_datasets import data

stocks = data.stocks()

chart = alt.Chart(stocks).mark_line().encode(
    x='date:T',
    y='price:Q',
    color='symbol:N'
).properties(
    title='Stock Prices Over Time'
)


# Save the chart as an HTML file
chart.save('chart2.html')

Output:

Explanation:

• mark_line() creates a line plot.
• x='date:T' encodes the x-axis with temporal data (dates).
• y='price:Q' and color='symbol:N' plot prices with lines colored by symbol.

Example 4: Layered and Combined Charts

Altair allows for chart layering and concatenation. Here, we’ll layer a bar chart and line chart.

import altair as alt
import pandas as pd
from vega_datasets import data

stocks = data.stocks()


# Bar chart for monthly stock volume
bars = alt.Chart(stocks).mark_bar().encode(
    x='month(date):O',
    y='sum(volume):Q',
    color='symbol:N'
)

# Line chart for monthly stock price
line = alt.Chart(stocks).mark_line(color='red').encode(
    x='month(date):O',
    y='mean(price):Q',
    detail='symbol:N'
)

# Layering both charts
combined_chart = bars + line


# Save the chart as an HTML file
combined_chart.save('chart3.html')

Output:

Explanation:

• mark_bar() and mark_line() define chart types.
• bars + line layers the charts to overlay the bar and line chart.

Advanced Features of Altair

1. Adding Selection and Filtering

Selections let users filter data based on interactions. Here’s a bar chart where users can click on bars to update another chart.

import altair as alt
import pandas as pd
from vega_datasets import data

# Load the Iris dataset
iris = data.iris()

# Define a selection for the species using selection_point
species_selection = alt.selection_point(fields=['species'], bind='legend')

# Base scatter plot with selection
scatter = alt.Chart(iris).mark_circle(size=60).encode(
    x='sepalLength:Q',
    y='petalLength:Q',
    color=alt.condition(species_selection, 'species:N', alt.value('lightgray')),
    tooltip=['species:N', 'sepalLength:Q', 'petalLength:Q']
).add_params(
    species_selection
)

# Save the chart as an HTML file
scatter.save('chart4.html')

Output:

2. Faceting for Small Multiples

Faceting allows displaying multiple subsets of data side-by-side. Here’s an example faceting the Iris dataset by species.

import altair as alt
import pandas as pd
from vega_datasets import data

# Load the Iris dataset
iris = data.iris()

# Faceted histogram by species
chart = alt.Chart(iris).mark_bar().encode(
    x=alt.X('sepalWidth:Q', bin=True),
    y='count()',
    color='species:N'
).facet(
    'species:N',
    columns=3
).properties(
    title='Distribution of Sepal Width by Species'
)

chart.save('chart5.html')

Output:

Conclusion

Altair provides a powerful yet intuitive approach to data visualization for data science in Python. With its declarative syntax, Altair allows you to focus on what you want to visualize rather than the intricate details of how the plot should be rendered. This makes it ideal for data scientists who want to quickly explore data patterns, create clean visuals, and incorporate interactivity with minimal code.

By integrating with Pandas data frames, Altair enables seamless handling of complex data sets, making it a suitable choice for exploratory data analysis. Its reliance on Vega-Lite, a high-level grammar of interactive graphics, also ensures that Altair visualizations are not only aesthetically pleasing but also highly customizable and capable of supporting complex data insights. Additionally, Altair’s support for layered and faceted charts, as well as advanced features like tooltips and selections, allows users to design interactive and multi-dimensional visualizations without extensive coding.

Altair’s compatibility with Jupyter Notebook, JupyterLab, Google Colab, and options to export charts as HTML or images, make it a flexible and accessible choice for anyone working in a Python data science environment. Its straightforward API allows for reproducibility and easy sharing, which is crucial for collaborative data science and reporting.

In conclusion, Altair is an excellent choice for Python users who need an efficient way to create professional, interactive data visualizations. Its focus on simplicity, combined with powerful interactivity and customization features, makes it a valuable tool for modern data analysis and visualization workflows. Whether you’re building quick exploratory plots or detailed visualizations for presentations, Altair provides the tools and flexibility to bring your data to life effectively.