制作从 seaborn 到 matplotlib 的堆叠条 plot

Question

I need some help making a set of stacked bar charts in python with matlibplot.

Formally, my dataframe looks like this

plt.figure(figsize=(10, 14))
fig= plt.figure()


ax = sns.countplot(x="airlines",hue='typecode', data=trafic,
              order=trafic.airlines.value_counts(ascending=False).iloc[:5].index,
              hue_order=trafic.typecode.value_counts(ascending=False).iloc[:5].index,
                )
ax.set(xlabel="Airlines code", ylabel='Count')

As written in order and hue_order, I want to isolate the 5 most present airlines and aircraft types in my database

I was advised to make a stacked bar plot to make a more presentable graph, only I don't see any functionality with Seaborn to make one, and I can't manage with matplotlib to plot it while respecting this idea of isolating the 5 airlines/aircraft types most present in my database

Thanks for your help!

Answer 1

The following code uses seaborn's countplot with dodge=False . This places all bars belonging to the same airline one on top of the other. In a next step, all bars are moved up to stack them:

import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import seaborn as sns

sns.set()
np.random.seed(123)
trafic = pd.DataFrame({'airlines': np.random.choice([*'abcdefghij'], 500),
                       'typecode': np.random.choice([*'qrstuvwxyz'], 500)})

fig = plt.figure(figsize=(10, 5))
ax = sns.countplot(x="airlines", hue='typecode', palette='rocket', dodge=False, data=trafic,
                   order=trafic.airlines.value_counts(ascending=False).iloc[:5].index,
                   hue_order=trafic.typecode.value_counts(ascending=False).iloc[:5].index)
ax.set(xlabel="Airlines code", ylabel='Count')
bottoms = {}
for bars in ax.containers:
    for bar in bars:
        x, y = bar.get_xy()
        h = bar.get_height()
        if x in bottoms:
            bar.set_y(bottoms[x])
            bottoms[x] += h
        else:
            bottoms[x] = h
ax.relim()  # the plot limits need to be updated with the moved bars
ax.autoscale()
plt.show()

Note that the airlines are sorted on their total airplanes, not on their total for the 5 overall most frequent airplane types.

PS: In the question's code, plt.figure() is called twice. That first creates an empty figure with the given figsize, and then a new figure with a default figsize.