Display totals and percentage in stacked bar chart using DataFrame.plot

Question

My data frame looks like below:

Airport	ATA Cost	Destination Handling	Custom	Total Cost
PRG	599222	11095	20174	630491
LXU	364715	11598	11595	387908
AMS	401382	23562	16680	441623
PRG	599222	11095	20174	630491

Using below codes it gives a stacked bar chart:

import pandas as pd

# sample dataframe
data = {'Airport': ['PRG', 'LXU', 'AMS', 'PRG'],
        'ATA Cost': [599222, 364715, 401382, 599222],
        'Destination Handling': [11095, 11598, 23562, 11095],
        'Custom': [20174, 11595, 16680, 20174],
        'Total Cost': [630491, 387908, 441623, 630491]}
df = pd.DataFrame(data)

# plot columns without Total Cost
df.iloc[:, :-1].plot(x='Airport', kind='barh', stacked=True, title='Breakdown of Costs', mark_right=True)    

How to add the totals (separated by thousands 1,000) over each stacked bar chart? How to add % for each segments in the stacked bar chart?

Novellanovello

Answer 1

You can use plt.text to place the information at the positions according to your data.

However, if you have very small bars, it might need some tweaking to look perfect.

df_total = df['Total Cost']
df = df.iloc[:, 0:4]
df.plot(x = 'Airport', kind='barh',stacked = True, title = 'Breakdown of Costs', mark_right = True)

df_rel = df[df.columns[1:]].div(df_total, 0)*100

for n in df_rel:
    for i, (cs, ab, pc, tot) in enumerate(zip(df.iloc[:, 1:].cumsum(1)[n], df[n], df_rel[n], df_total)):
        plt.text(tot, i, str(tot), va='center')
        plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va='center', ha='center')

EDIT: Some arbitrary ideas for better readability:

shift the total values to the right, use 45° rotated text:

    plt.text(tot+10000, i, str(tot), va='center')
    plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va='center', ha='center', rotation=45)

switch between top- and bottom-aligned text:

va = ['top', 'bottom']
va_idx = 0
for n in df_rel:
    va_idx = 1 - va_idx
    for i, (cs, ab, pc, tot) in enumerate(zip(df.iloc[:, 1:].cumsum(1)[n], df[n], df_rel[n], df_total)):
        plt.text(tot+10000, i, str(tot), va='center')
        plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va=va[va_idx], ha='center')

label only bars with 10% or more:

if pc >= 10:
    plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va='center', ha='center')

...or still print them, but vertical:

if pc >= 10:
    plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va='center', ha='center')
else:
    plt.text(cs - ab/2, i, str(np.round(pc, 1)) + '%', va='center', ha='center', rotation=90)

Roswald

thank you. it works very well. is there a way to avoid text overlapping each other? – Novellanovello

I don't know of an automatism, but perhaps rotating by 45 degree might help as a first approach? – Roswald

if I do not plot the number < 0.1, how to adjust the code? – Novellanovello

What do you mean - do you want to reduce the digits of precision right to the decimal sign to only print the integer values? – Roswald

I mean i want to hide those text overlapping each other. it's also good to make the % with 0 decimal place – Novellanovello

you could change str(np.round(pc, 1)) to str(int(np.round(pc, 0))) – Roswald

the more important thing for me here is to hide those small % overlapping each other. – Novellanovello

honestly: the easiest and most effective way would be simply to plot a (vertical) bar-chart instead of a hbar-chart – Roswald

I have to plot horizontally because i have a big chart. can we conditional plot the text (i,e. plot if the % is greater than 10%) – Novellanovello

You could even still label smaller values too, but these vertical rotated, so that they fit better in... – Roswald

Answer 2

• This is more easily implemented with matplotlib.pyplot.bar_label, which is thoroughly described in How to add value labels on a bar chart.

Data and Imports

import pandas as pd

# load the dataframe from the OP and set the x-axis column as the index
df = df.set_index('Airport')

# calculate the percent for each row
per = df.iloc[:, :-1].div(df['Total Cost'], axis=0).mul(100).round(2)

Horizontal Bars

# plot
ax = df.iloc[:, :-1].plot(kind='barh', stacked=True, figsize=(10, 6))

ax.legend(bbox_to_anchor=(1, 0.5), loc='center left', frameon=False)

# iterate through the containers
for c in ax.containers:
    
    # get the current segment label (a string); corresponds to column / legend
    label = c.get_label()
    
    # create custom labels with percent
    labels = per[label].astype(str) + '%'
    
    # add the annotation
    ax.bar_label(c, labels=labels, label_type='center', rotation=-90, fontsize=7)

# annotate the top of the bar with the full count
_ = ax.bar_label(ax.containers[-1], label_type='edge', rotation=-90)

Vertical Bars

ax = df.iloc[:, :-1].plot(kind='bar', stacked=True, figsize=(10, 6), rot=0)

ax.legend(bbox_to_anchor=(1, 0.5), loc='center left', frameon=False)

# iterate through the containers
for c in ax.containers:
    
    # get the current segment label (a string); corresponds to column / legend
    label = c.get_label()
    
    # create custom labels with percent
    labels = per[label].astype(str) + '%'
    
    # add the annotation
    ax.bar_label(c, labels=labels, label_type='center', fontsize=7)

# annotate the top of the bar with the full count
_ = ax.bar_label(ax.containers[-1], label_type='edge')

Vertical Bars - Not Stacked

• It is typically better to not stack, because the relative sizes of segments are easier to compare.

# plot
ax = df.iloc[:, :-1].plot(kind='bar', figsize=(10, 6), rot=0, width=0.85)

ax.legend(bbox_to_anchor=(1, 0.5), loc='center left', frameon=False)

# iterate through the containers
for c in ax.containers:
    
    # get the current segment label (a string); corresponds to column / legend
    label = c.get_label()
    
    # create custom labels with percent
    labels = per[label].astype(str) + '%'
    
    # add the percent
    ax.bar_label(c, labels=labels, label_type='center', fontsize=7)

    # add the count to the top of the bar
    _ = ax.bar_label(c, label_type='edge')

Gooch