Pandas df.plot set color proportional to a dataframe column

Question

Question:

I have a pandas dataframe df with 6 rows representing different variables.
I want to plot the two first columns in a bar graph and I would like to use the third column score to colorize the bars based on an existing colormap, for example the matplotlib viridis one.

This dataframe looks like this:

    ID  count   score
0    4      1       3
1    5      4       4
2    5      8       5
3    3      7       2
4    1      5       5
5    1      3       4

What my code looks like for the moment:

df.plot('ID', 'count', kind='bar', figsize=(15,5), ax=plt.subplot(122), colors=cm.viridis.colors)
plt.show()

But the color of all bar is a unique purple.
I want, let say, to sample the viridis colormap proportionally to values in my score column.
Is there a simple way to achieve that?

Edit [2019-10-31]:

Using seaborn as proposed in the first answer gives me an error:

my_palette = sns.color_palette(plt.get_cmap('viridis'), 5)
sns.barplot(x="New ID", y="count", hue="Note", data=df, palette=my_palette)

The error returned:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-48-1943cafa0d79> in <module>
----> 1 my_palette = sns.color_palette(plt.get_cmap('viridis'), 5)

/usr/local/lib/python3.6/dist-packages/seaborn/palettes.py in color_palette(palette, n_colors, desat)
    238 
    239     # Always return as many colors as we asked for
--> 240     pal_cycle = cycle(palette)
    241     palette = [next(pal_cycle) for _ in range(n_colors)]
    242 

TypeError: 'ListedColormap' object is not iterable

Edit [2019-11-05]:

Using the proposed code lead the resulting graph to have misaligned bars with the x-ticks (which I expected to be centered under the bar on the x-axis), especially when there is only one bar per category as the next figure shows:

In addition, I would like to be able to take into consideration 'NaN' values in the df['count'] column. Currently, if there is one 'NaN' value, it raises a:
TypeError: unsupported operand type(s) for /: 'str' and 'int'
So, I have to set this value equal to 0 even if it is NOT truly a zero.

But if there is 'NaN' in my count attribute, there is actually no bar. So, the color doesn't matter. — swiss_knight
– swiss_knight, Commented Nov 6, 2019 at 15:42

Cowflu · Accepted Answer · 2019-10-31 12:57:30Z

2

The simplest is to use seaborn.barplot this way:

import seaborn as sns
ax = sns.barplot(x="ID", y="count", hue="score", data=df, palette=your_palette)

You can create a palette this way:

your_palette = sns.color_palette('inferno', n_point)

where n_point are the number of different colors in your custome palette.

Hope it helps.

edited Oct 31, 2019 at 12:57

answered Oct 31, 2019 at 11:26

Cowflu

3322 silver badges13 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

swiss_knight Over a year ago

Thanks for your answer, unfortunately this gave me an error. My version of seaborn is 0.9.0 (downloaded from PyPi using pip).

Cowflu Over a year ago

I changed the code. Sorry for the error. color_palette is a method that accepts the name of a colormap and the number of point. I tested it again and it works now.

swiss_knight Over a year ago

I can see that the bars are randomly aligned with the x-ticks, this is annoying and I've search the documentation without finding anything to fix it.

Robert King Over a year ago

I don't think they are 'randomly' aligned. The sample data you provided has two entries with id=1. How would you like these to be plotted?

Klaifer Garcia · Accepted Answer · 2024-04-11 12:00:35Z

1

You also can do this with matplotlib

import pandas as pd
import matplotlib.pyplot as plt

df = pd.DataFrame({
    'ID': [4, 5, 5, 3, 1, 1],
    'count': [1, 4, 5, 6, 7, 3],
    'score': [3, 4, 5, 2, 5, 4],
})

cm = plt.get_cmap('viridis')
maxval = max(df.score)

df.plot('ID', 'count', kind='bar', 
        figsize=(15,5), ax=plt.subplot(122), 
        color=cm(df.score/maxval))
plt.show()

answered Apr 11, 2024 at 12:00

Klaifer Garcia

4205 silver badges14 bronze badges

Collectives™ on Stack Overflow

Pandas df.plot set color proportional to a dataframe column

Question:

Edit [2019-10-31]:

Edit [2019-11-05]:

2 Answers 2

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

Question:

Edit [2019-10-31]:

Edit [2019-11-05]:

2 Answers 2

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related