Bubble Chart

Source: Data Viz Catalogue

• Multi-variable graph (a cross between a Scatterplot and a Proportional Area Chart)
• Up to 4 variables:
  • X axis (quantitative variable)
  • Y axis (quantitative variable)
  • circle size (quantitative variable)
  • circle color (qualitative variable)
• They are used to compare and show the relationships between categorised circles, by the use of positioning and proportions.
• Limitations:
  • data size capacity: too many bubbles can make the chart hard to read
    • This can be somewhat remedied by interactivity
    • Transparent circles
• Important: the sizes of the circles need to be drawn based on the circle’s area, not its radius or diameter.

$Circle\_Area = \pi + {Radius}^{2}$

$Circle\_Diameter = 2 \sqrt{\frac{Area}{\pi}}$

import math
import pandas as pd
from pandas.api.types import CategoricalDtype

from bokeh.io import output_notebook, show
from bokeh.plotting import figure
from bokeh.transform import transform
from bokeh.models import ColumnDataSource
from bokeh.models.mappers import CategoricalColorMapper
output_notebook()

Loading BokehJS ...

Dataset

Dataset comes from Kaggle: Titanic - Suited for binary logistic regression

import kaggle

kaggle.api.authenticate()
kaggle.api.dataset_download_files('heptapod/titanic', path='.', unzip=True)

Data reading and preproccesing

df = pd.read_csv('train_and_test2.csv', sep=',')

df.rename(columns={'2urvived': 'survived'}, inplace=True)

df['Sex'] = df['Sex'].replace({0: 'Male', 1: 'Female'}).astype('category')
df['Pclass'] = df['Pclass'].replace({1: '1st', 2: '2nd', 3: '3th'}).astype('category')
df['Age'] = pd.cut(df['Age'].round(), 5)

df[['Age', 'Pclass', 'sibsp', 'survived']].sample(10)

	Age	Pclass	sibsp	survived
516	(32.0, 48.0]	2nd	0	1
29	(16.0, 32.0]	3th	0	0
523	(32.0, 48.0]	1st	0	1
1250	(16.0, 32.0]	3th	1	0
911	(48.0, 64.0]	1st	1	0
161	(32.0, 48.0]	2nd	0	1
723	(48.0, 64.0]	2nd	0	0
972	(64.0, 80.0]	1st	1	0
383	(32.0, 48.0]	1st	1	1
1244	(48.0, 64.0]	2nd	1	0

aux = df[['Age', 'Pclass', 'sibsp', 'survived']].groupby(by=['Age', 'Pclass']).agg(['sum', 'count']).copy()
aux.reset_index(inplace=True)
aux['Age'] = aux['Age'].astype(str)
aux['percent'] = aux['survived']['sum'] / aux['survived']['count'] * 100
aux['sibsp_area'] = 2 * ((aux['sibsp']['sum'] / aux['sibsp']['count']) / math.pi).pow(1./2)
aux = aux[aux['percent'] > 0]
aux

	Age	Pclass	sibsp		survived		percent	sibsp_area
			sum	count	sum	count
0	(-0.08, 16.0]	1st	6	11	8	11	72.727273	0.833363
1	(-0.08, 16.0]	2nd	21	30	19	30	63.333333	0.944070
2	(-0.08, 16.0]	3th	177	93	28	93	30.107527	1.556683
3	(16.0, 32.0]	1st	51	127	57	127	44.881890	0.715053
4	(16.0, 32.0]	2nd	61	157	41	157	26.114650	0.703348
5	(16.0, 32.0]	3th	192	507	83	507	16.370809	0.694387
6	(32.0, 48.0]	1st	43	106	47	106	44.339623	0.718681
7	(32.0, 48.0]	2nd	20	64	21	64	32.812500	0.630783
8	(32.0, 48.0]	3th	33	95	7	95	7.368421	0.665044
9	(48.0, 64.0]	1st	38	71	23	71	32.394366	0.825501
10	(48.0, 64.0]	2nd	7	24	6	24	25.000000	0.609394
11	(48.0, 64.0]	3th	1	11	1	11	9.090909	0.340219
12	(64.0, 80.0]	1st	3	8	1	8	12.500000	0.690988

aux['sibsp_area'] *= 10

Data visualization

source = ColumnDataSource(data=aux)

p = figure(
    title='Survival chance on The Titanic',
    plot_width=600, 
    plot_height=400,
    x_range=aux['Age'].unique().tolist(),
)

color_mapper = CategoricalColorMapper(palette=["green", "blue", "red"], factors=["1st", "2nd", "3th"])

p.circle(
    x='Age_', 
    y='percent_', 
    size='sibsp_area_', 
    fill_color=transform('Pclass_', color_mapper), 
    line_color=None,
    fill_alpha=0.6,
    legend_group='Pclass_',
    source=source,
)

p.yaxis.axis_label = 'Survival percentage (%)'
p.xaxis.axis_label = 'Age ranges'

show(p)