Iterate through columns to generate barplots while using groupby

Question

This is my dataframe:

data = {'machine': ['machine_a', 'machine_a', 'machine a', 'machine a', 'machine a', 'machine a', 'machine_b', 'machine_b', 'machine_b', 'machine_b', 'machine_b', 'machine_b', 'machine_c', 'machine_c', 'machine_c', 'machine_c', 'machine_c', 'machine_c'], 'bin': ['(0, 200]', '(200, 400]', '(400, 600]', '(600, 800]', '(800, 1000]', '(1000, 1200]', '(0, 200]', '(200, 400]', '(400, 600]', '(600, 800]', '(800, 1000]', '(1000, 1200]', '(0, 200]', '(200, 400]', '(400, 600]', '(600, 800]', '(800, 1000]', '(1000, 1200]'], 'speed': [10, 0, 20, 0, 20, 10, 5, 0, 40, 10, 20, 10, 5, 25, 0, 10, 5, 10], 'Temp': [0, 0, 0, 20, 20, 0, 35, 0, 0, 20, 0, 70, 30, 0, 0, 5, 0, 25]}
df = pd.DataFrame(data)

      machine           bin  speed  Temp
0   machine_a      (0, 200]     10     0
1   machine_a    (200, 400]      0     0
2   machine a    (400, 600]     20     0
3   machine a    (600, 800]      0    20
4   machine a   (800, 1000]     20    20
5   machine a  (1000, 1200]     10     0
6   machine_b      (0, 200]      5    35
7   machine_b    (200, 400]      0     0
8   machine_b    (400, 600]     40     0
9   machine_b    (600, 800]     10    20
10  machine_b   (800, 1000]     20     0
11  machine_b  (1000, 1200]     10    70
12  machine_c      (0, 200]      5    30
13  machine_c    (200, 400]     25     0
14  machine_c    (400, 600]      0     0
15  machine_c    (600, 800]     10     5
16  machine_c   (800, 1000]      5     0
17  machine_c  (1000, 1200]     10    25

I want to create different bar plots for columns speed and temp, where x axis is the bins column. I want to do this for every different machine.

so far I've created a for loop to iterate over the last two columns

import seaborn as sns
import matplotlib.pyplot as plt

for column in df.columns[2:]: 
    sns.set()
    fig, ax = plt.subplots()
    sns.set(style="ticks")
    sns.barplot(x = df.bin, y=column, data=df)
    sns.despine(offset=10, trim=True) 
    fig.set_size_inches(22,14)

This creates 2 bar plots. 1 for speed and 1 for temp. How do I make so I get back 6 bar plots (2 for each machine)? Essentially how do I use groupby in this case?

Complete Synthetic Data

50 machines, each with 109 bins

import pandas as pd
import numpy as np

bins = [f'({n*200}, {(n+1)*200}]' for _ in range(50) for n in range(109)]
machines = [f'machine_{n}' for n in range(50) for _ in range(109)]
np.random.seed(365)
speed = np.random.randint(0, 40, size=len(machines))
temp = np.random.choice([0, 30, 70], size=len(machines))

df = pd.DataFrame({'machine': machines, 'bin': bins, 'speed': speed, 'Temp': temp})

df.head()
     machine          bin  speed  Temp
0  machine_0     (0, 200]     18    30
1  machine_0   (200, 400]     33    70
2  machine_0   (400, 600]     27    30
3  machine_0   (600, 800]      5    30
4  machine_0  (800, 1000]     34    30

df.tail()
         machine             bin  speed  Temp
5445  machine_49  (20800, 21000]      6     0
5446  machine_49  (21000, 21200]     20    30
5447  machine_49  (21200, 21400]     14     0
5448  machine_49  (21400, 21600]     38    30
5449  machine_49  (21600, 21800]     24    70

mozway · Answer 1 · 2021-09-07T13:47:37.113

2

You can use seaborn.catplot directly. You just need to melt the y-columns first:

import seaborn as sns

sns.catplot(data=df.melt(id_vars=['machine', 'bin']),
            col='machine',
            x='bin', y='value', hue='variable',
            kind='bar'
           )

output:

Or to have the variables as rows:

import seaborn as sns

sns.catplot(data=df.melt(id_vars=['machine', 'bin']),
            col='machine',
            row='variable',
            x='bin', y='value',
            kind='bar',
            color='k',
            )

output:

Or to split the variables:

import seaborn as sns

sns.catplot(data=df.melt(id_vars=['machine', 'bin']),
            hue='machine',
            x='bin', y='value', col='variable',
            kind='bar'
           )

edited Sep 07 '21 at 13:47

answered Sep 07 '21 at 13:36

mozway

194,879
13
39
75

Neither of these are what i want in return. I would want the X value to be the bins column and the Y to be either the speed or temp (not both on the same plot). – Sam Sep 07 '21 at 13:43
this options works to an extent. The reason I want a for loop is because my actual dataframe contains many more bins. Using sns.catplot makes the plot way to small to even see. – Sam Sep 07 '21 at 13:52
I have 109 bins for each machine and over 50 machines... The only way to do it where one can still see the information is with an individual plot for each machine and variable. This is why I started off with a for loop. – Sam Sep 07 '21 at 15:28

Trenton McKinney · Accepted Answer · 2021-09-07T19:59:22.700

Given the existing data as described in the comments and with the updated synthetic data set.
Using seaborn.barplot. seaborn is a high-level API for matplotlib

import matplotlib.pyplot as plt
import seaborn as sns

# melt the dataframe into a long form and group by the machine
dfmg = df.melt(id_vars=['machine', 'bin']).groupby('machine')

Option 1

Create a single plot figure for each machine, which is saved with the machine name
- this saves 50 plots, 1 for each machine

for mach, data in dfmg:
    plt.figure(figsize=(20, 5))
    sns.barplot(data=data, x='bin', y='value', hue='variable')
    plt.xticks(rotation=90, ha='center')
    plt.title(mach)
    plt.legend(loc='upper left')
    plt.tight_layout()
    plt.savefig(f'{mach}.png')
    plt.show()

Separate plots for 'Temp' and 'speed'

for mach, data in dfmg:
    fig, (ax1, ax2) = plt.subplots(nrows=2, figsize=(20, 5), sharex=True)
    
    # select data
    temp = data[data.variable.eq('Temp')]
    speed = data[data.variable.eq('speed')]
    
    # plot data
    sns.barplot(data=temp, x='bin', y='value', ax=ax1)  # ax1.bar(data=temp, x='bin', y='value') - without seaborn
    sns.barplot(data=speed, x='bin', y='value', ax=ax2)  # ax2.bar(data=speed, x='bin', y='value') - without seaborn
    
    ax1.set(xlabel=None, ylabel='Temp')
    ax1.tick_params(bottom=False)
    ax2.set(ylabel='Speed')
    ax2.set_xticklabels(ax2.get_xticklabels(), rotation=90, ha='center')  # ax2.tick_params('x', rotation=90) - without seaborn
    
    fig.suptitle(mach)
    fig.tight_layout()
    fig.savefig(f'{mach}.png')
    plt.show()

Option 2

Alternatively, create a single figure will all plots on the figure
- this saves 1 figure with all 50 machines
The plot of each machine on the figure in this section, looks the same as the previous plots with orange and blue hue.

fig, axes = plt.subplots(nrows=len(df.machine.unique()), ncols=1, figsize=(20, 250))
for i, (mach, data) in enumerate(df.melt(id_vars=['machine', 'bin']).groupby('machine')):
    sns.barplot(data=data, x='bin', y='value', hue='variable', ax=axes[i])
    axes[i].set_xticklabels(axes[i].get_xticklabels(), rotation=90, ha='center')
    axes[i].set(title=mach)
    axes[i].legend(loc='upper left')
fig.tight_layout()
fig.savefig('machines_all.png')
plt.show()

seaborn.catplot can create an equivalent, single figure-level plot with one line

import seaborn as sns

# melt the dataframe into a long form
dfm = df.melt(id_vars=['machine', 'bin'])

# plot
p = sns.catplot(data=dfm, col='machine', x='bin', y='value', hue='variable', kind='bar', col_wrap=1, height=4, aspect=5)
p.set_xticklabels(rotation=90) 
p.savefig("facet_plot.png")

score 0 · Answer 3 · answered Sep 07 '21 at 14:05

Here is a version using groupby:

df2 = df.melt(id_vars=['machine', 'bin'])

COLS = df2['machine'].nunique()
ROWS = df2['variable'].nunique()

fig, axes = plt.subplots(ncols=COLS,
                         nrows=ROWS,
                        )
i = 0
for group, d in df2.groupby(['machine', 'variable']):
    ax = axes[i%ROWS][i//ROWS]
    ax.bar(d['bin'], d['value'])
    if not i//ROWS:
        ax.set_ylabel(group[1])
    if i%ROWS == ROWS-1:
        ax.set_xlabel(group[0])
    i+=1

output:

Iterate through columns to generate barplots while using groupby

Complete Synthetic Data

3 Answers3

Option 1

Option 2

Linked