CODEWITHSUNDEEP
python-3.xmatplotlibpandas-groupby
Giuseppe Minardi
Giuseppe Minardi

Reputation: 411

Stacked bar plot with matplotlib

I have a dataframe extracted from WhatsApp with columns: Date&Time, msg, name, msg_len. Date&Time is a DateTime object that represents when the message has been sent, msg is the actual message, name is who sent the message and msg_len is the actual length of the message. I'm trying to build a stacked bar plot using this dataframe: on the X-axis the date (e.g. 2019-02), on the y-axis, the mean length or the number of messages sent in that month and each bar is divided by each person. So far my function looks like this:

def BarPlotMonth(Data):
    """
    This function plots a barplot for the number of messages sent for each month and the mean length of the messages for each month
    """

    fig,axes = plt.subplots(2,1,
            figsize=(18,10),
            sharex = True)


    GroupedByMonth = Data.groupby(Data['Date&Time'].dt.strftime('%Y-%m'))['msg_len']

    Mean = GroupedByMonth.mean()
    Count = GroupedByMonth.count()
    Std = GroupedByMonth.std()

    axes[0].bar(Count.index, Count, color = 'lightblue')
    axes[0].set_title('Number of text per month')
    axes[0].set_ylabel('Count')

    axes[1].bar(Mean.index, Mean, color = 'lightblue', yerr = Std)
    axes[1].set_title('Mean lenght of a message per month')
    axes[1].set_ylabel('Mean lenght')
    axes[1].set_xlabel('Year-Month')

    plt.xticks(rotation=45)
    axes[1].legend()

    plt.savefig('WhatsApp_conversations.png')
    plt.show()

But I can't divide each bar. How can I solve this?

Upvotes: 2

Views: 6206

Answers (1)

mfitzp
mfitzp

Reputation: 15545

You will need to restructure your DataFrame a bit to be able to use df.plot(kind='bar', stacked=True).

group_by_month_per_user = df.groupby(
    [
        df['Date&Time'].dt.strftime('%Y-%m'),
        'name'
    ]   
).mean().unstack()

group_by_month_per_user

This produces a table with the following structure.

             msg_len                                 
name           alice        bob   giuseppe     martin
Date&Time                                            
2019-01    48.870968  42.315789  56.391304  49.586207
2019-02    51.099174  48.777778  56.173913  51.895652
2019-03    52.336364  49.626168  47.021898  46.626263

Note that the columns is a multindex with msg_len over all columns, we need to remove this to keep the legend tidy (can simply select the entire column). Then the resulting DataFrame can be passed to .plot.

group_by_month_per_user['msg_len'].plot(kind='bar', stacked=True, legend=['name'])

This produces the following plot.

output showing stacked bar for SMS length

The following code was used to generate a random dataset.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime, timedelta
from random import randint, choice
import string

ts = datetime.now()
data = []
names = ['bob', 'alice', 'martin', 'giuseppe']

for n in range(1000):
    msg_len = randint(0, 100)
    row = [
        ts - timedelta(days=randint(-30,30)),
        ''.join(random.choice(string.ascii_lowercase) for _ in range(msg_len)),
        choice(names),
        msg_len
    ]

    data.append(row)

df = pd.DataFrame(data, columns = ['Date&Time', 'msg', 'name', 'msg_len'])

Upvotes: 4

Related Questions