Double heatmap with custom annotations

Heatmaps can be a great way to visualize evolution through time. This post shows how to create a double heatmap, one normal and one normalized, with annotations using matplotlib and seaborn.

We will create a step-by-step guide to build the plot, starting with a very simple heatmap and ending with a fully customized one.

About

Heatmap is a graphical representation of data where the individual values contained in a matrix are represented as colors. This is very useful to display the concentration of values between two dimensions, without looking at the individual values.

This chart has been created by Joseph Barbier. Thanks to him for accepting sharing its work here!

Libraries

First, we need to install the following libraries:

matplotlib: for plot customization
seaborn: for creating the plot
pandas: for data manipulation
highlight_text: for annotations

import matplotlib.pyplot as plt
from matplotlib.patches import FancyArrowPatch
import matplotlib.colors as mcolors
import seaborn as sns
import pandas as pd
from highlight_text import ax_text

Dataset

The type of data needed when creating a heatmap is a matrix. Each individual cell in the matrix represents a value that will be visualized using a color scale.

In this post we need 2 datasets:

original heatmap
normalized heatmap

path = 'https://raw.githubusercontent.com/holtzy/The-Python-Graph-Gallery/master/static/data/heatmap_data.csv'
heatmap_data = pd.read_csv(path, index_col=0)

path = 'https://raw.githubusercontent.com/holtzy/The-Python-Graph-Gallery/master/static/data/heatmap_data_norm.csv'
heatmap_data_norm = pd.read_csv(path, index_col=0)

Simple double heatmap

We start by creating a figure with 2 subplots.

The first subplot will contain the original heatmap, and the second subplot will contain the normalized heatmap.

We use the heatmap() function from seaborn to create the heatmaps.

fig, axs = plt.subplots(ncols=2, figsize=(16, 10))

# iterate over the datasets
for i, data in enumerate([heatmap_data, heatmap_data_norm]):
    
    # plot the heatmap
    sns.heatmap(
        data,
        ax=axs[i],
        cbar=False
    )

# save and show
plt.savefig('../../static/graph/web-heatmap-comparison-1.png', bbox_inches='tight')
plt.show()

preview

Custom color map and remove axis

The next step is to customize the color map and remove the axis.

the colormap is created using the LinearSegmentedColormap.from_list() function, and then we pass it as an argument to the cmap parameter of the heatmap() function
the axis is removed using the set_axis_off() function

# create a custom colormap
cmap = mcolors.LinearSegmentedColormap.from_list("", ["#2a9d8f", "#e9c46a", "#e76f51"])

fig, axs = plt.subplots(ncols=2, figsize=(16, 10))

# iterate over the datasets
for i, data in enumerate([heatmap_data, heatmap_data_norm]):
    
    # plot the heatmap
    sns.heatmap(
        data,
        ax=axs[i],
        cmap=cmap,
        cbar=False
    )

    # remove the axis
    axs[i].set_axis_off()

# save and show
plt.savefig('../../static/graph/web-heatmap-comparison-2.png', bbox_inches='tight')
plt.show()

preview

Add region labels

The next step is to add region labels to the heatmaps.

For this, we iterate over the index of the dataset (where the region names are stored) and add the text to the plot using the text() function. We have to specify:

the x and y coordinates
the text to display
the horizontal alignment
the vertical alignment
the font size
the font weight

Warning: make sure both of the heatmaps have the same region names in the same order, otherwise the labels will not match the regions.

# create a custom colormap
cmap = mcolors.LinearSegmentedColormap.from_list("", ["#2a9d8f", "#e9c46a", "#e76f51"])

fig, axs = plt.subplots(ncols=2, figsize=(16, 10))

# iterate over the datasets
for i, data in enumerate([heatmap_data, heatmap_data_norm]):
    
    # plot the heatmap
    sns.heatmap(
        data,
        ax=axs[i],
        cmap=cmap,
        cbar=False
    )

    # remove the axis
    axs[i].set_axis_off()

    # add the region names
    for j,region in enumerate(data.index):
        axs[i].text(
            0.4, # x axis position
            j+0.5, # y axis position
            f"{region}", # text
            ha='left',
            va='center',
            fontsize=17,
            fontweight='light',
        )

# save and show
plt.savefig('../../static/graph/web-heatmap-comparison-3.png', bbox_inches='tight')
plt.show()

preview

Add legend

Now let's add a legend to the heatmaps so that it's more understandable.

We add both Consommation d'énergie and Consommation d'énergie par habitant to the legend using the text() function.

# create a custom colormap
cmap = mcolors.LinearSegmentedColormap.from_list("", ["#2a9d8f", "#e9c46a", "#e76f51"])

fig, axs = plt.subplots(ncols=2, figsize=(16, 10))

# iterate over the datasets
for i, data in enumerate([heatmap_data, heatmap_data_norm]):
    
    # plot the heatmap
    sns.heatmap(
        data,
        ax=axs[i],
        cmap=cmap,
        cbar=False
    )

    # remove the axis
    axs[i].set_axis_off()

    # add the region names
    for j,region in enumerate(data.index):
        axs[i].text(
            0.4, # x axis position
            j+0.5, # y axis position
            f"{region}", # text
            ha='left',
            va='center',
            fontsize=17,
            fontweight='light',
        )

    # description of each heatmap
    if i==0: # first heatmap
        text = "Consommation d'énergie"
    else: # second heatmap
        text = "Consommation d'énergie par habitant"
    ax_text(
        0, 12.4,
        f"<{text}>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )

    # date for reference
    ax_text(
        0, -0.2,
        "<2011>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )
    ax_text(
        10, -0.2,
        "<2021>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )

# save and show
plt.savefig('../../static/graph/web-heatmap-comparison-4.png', bbox_inches='tight')
plt.show()

preview

We have the main components of the plot! It only misses a title and a bunch of annotations.

Final chart with annotations

All annotations relies on the ax_text() function from the highlight_text library. In this post we explain how to use it in detail.

The arrows are created using the FancyArrowPatch() function from matplotlib.patches. You can learn more about them in this post.

# create a custom colormap
cmap = mcolors.LinearSegmentedColormap.from_list("", ["#2a9d8f", "#e9c46a", "#e76f51"])

fig, axs = plt.subplots(ncols=2, figsize=(16, 10))

# iterate over the datasets
for i, data in enumerate([heatmap_data, heatmap_data_norm]):
    
    # plot the heatmap
    sns.heatmap(
        data,
        ax=axs[i],
        cmap=cmap,
        cbar=False
    )

    # remove the axis
    axs[i].set_axis_off()

    # add the region names
    for j,region in enumerate(data.index):
        axs[i].text(
            0.4, # x axis position
            j+0.5, # y axis position
            f"{region}", # text
            ha='left',
            va='center',
            fontsize=17,
            fontweight='light',
        )

    # description of each heatmap
    if i==0: # first heatmap
        text = "Consommation d'énergie"
    else: # second heatmap
        text = "Consommation d'énergie par habitant"
    ax_text(
        0, 12.4,
        f"<{text}>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )

    # date for reference
    ax_text(
        0, -0.2,
        "<2011>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )
    ax_text(
        10, -0.2,
        "<2021>",
        ha='left', va='center',
        fontsize=12, fontweight='light',
        color='black',
        highlight_textprops=[
            {"style": "italic"}
        ],
        ax=axs[i]
    )

# title
text = "L'île de France pollue t'elle vraiment <plus> ?"
ax_text(
    0, -4,
    text,
    ha='left', va='center',
    fontsize=24, fontweight='bold',
    color='black',
    highlight_textprops=[
        {"color": "#e76f51"}
    ],
    ax=axs[0]
)

# subtitle
text = "Consommation d'énergie et consommation d'énergie par habitant, de 2011 à 2021, par région (or Corse et territoires d'outre-mer)"
ax_text(
    0, -3.5,
    text,
    ha='left', va='center',
    fontsize=14, 
    color='grey',
    ax=axs[0]
)

# credit source
text = """
<Design>: Joseph Barbier   <Données>: data.gouv.fr
Consommation de gaz et d'électricité en France, mesuré en MWh. Les données sont agrégées par région et par année.
La population moyenne de 2008 et 2024 est utilisée pour normaliser la consommation d'énergie par habitant.
Améliorer ce graphique: <https://github.com/JosephBARBIERDARNAL/Small-Projects/tree/main/conso-france>
"""
ax_text(
    0, 13.9,
    text,
    ha='left', va='center',
    fontsize=10, fontweight='light',
    color='black',
    highlight_textprops=[
        {"fontweight": "bold"},
        {"fontweight": "bold"},
        {"fontweight": "bold"}
    ],
    ax=axs[0]
)

# left explanation
text = "Dans l'absolu, <l'Île-de-France>\nest la région avec la plus\n<haute consommation d'énergie>"
ax_text(
    4, -1.7,
    text,
    ha='left', va='center',
    fontsize=15, fontweight='light',
    color='black',
    highlight_textprops=[
        {"fontweight": "bold"},
        {"color": "#e76f51", "fontweight": "bold"}
    ],
    ax=axs[0]
)

# right explanation
text = "Relativement à son nombre d'habitants,\n<l'Île-de-France> est une région avec\n<consommation d'énergie par habitant>\n<plutôt faible>"
ax_text(
    12, -1.7,
    text,
    ha='left', va='center',
    fontsize=15, fontweight='light',
    color='black',
    highlight_textprops=[
        {"fontweight": "bold"},
        {"color": "#2a9d8f", "fontweight": "bold"},
        {"color": "#2a9d8f", "fontweight": "bold"}
    ],
    ax=axs[0]
)

# arrows
def draw_arrow(tail_position, head_position, invert=False):
    kw = dict(arrowstyle="Simple, tail_width=0.5, head_width=4, head_length=8", color="k")
    if invert:
        connectionstyle = "arc3,rad=-.5"
    else:
        connectionstyle = "arc3,rad=.5"
    a = FancyArrowPatch(tail_position, head_position,
                                connectionstyle=connectionstyle,
                                transform=fig.transFigure,
                                **kw)
    fig.patches.append(a)

draw_arrow((0.25, 1.01), (0.17, 0.91))
draw_arrow((0.77, 1.01), (0.82, 0.91), invert=True)

# save and show
plt.savefig('../../static/graph/web-heatmap-comparison-5.png', bbox_inches='tight')
plt.show()

preview

Going further

This post shows how to create a double heatmap with annotations using matplotlib and seaborn.

For more examples of advanced customization, check out how to create a beautiful legend and this really nice heatmap for time series.

Scatterplot

Heatmap

Correlogram

Bubble

Connected Scatter

2D Density

Contact & Edit

👋 This document is a work by Yan Holtz. You can contribute on github, send me a feedback on twitter or subscribe to the newsletter to know when new examples are published! 🔥

This page is just a jupyter notebook, you can edit it here. Please help me making this website better 🙏!

Double heatmap with custom annotations

About

Libraries

Dataset

Simple double heatmap

Custom color map and remove axis

Add region labels

Add legend

Final chart with annotations

Going further

Correlation

Contact & Edit