Exploratory visualization in pandas

For exploring data, pandas actually has pretty decent visualization capabilities.

%matplotlib inline
import matplotlib.pyplot as plt

import pandas as pd

import seaborn as sns

df = sns.load_dataset('iris')

df.head()

	sepal_length	sepal_width	petal_length	petal_width	species
0	5.1	3.5	1.4	0.2	setosa
1	4.9	3.0	1.4	0.2	setosa
2	4.7	3.2	1.3	0.2	setosa
3	4.6	3.1	1.5	0.2	setosa
4	5.0	3.6	1.4	0.2	setosa

pd.options.plotting.backend = 'matplotlib'

df.plot.kde(layout = (2,2), subplots = True, sharey=True)
pass

from pandas.plotting import boxplot_frame_groupby

boxplot_frame_groupby(df.groupby('species'), layout=(1,3), grid=False, rot=45);

pd.options.plotting.backend = 'plotly'

df.plot.scatter(
    x='sepal_length',
    y='petal_length',
    color='species',
    marginal_y="violin",
    marginal_x="box",
    trendline="ols"
)

Using pandas-bokeh

pd.options.plotting.backend = 'pandas_bokeh'
import pandas_bokeh
from bokeh.io import output_notebook

output_notebook()

Loading BokehJS ...

Example from official docs: pandas-bokeh

df_mapplot = pd.read_csv(r"https://bit.ly/325W5Yy")
df_mapplot["size"] = df_mapplot["pop_max"] / 1000000
df_mapplot.plot_bokeh.map(
    x="longitude",
    y="latitude",
    hovertool_string="<h2> @{name} </h2> <h3> Population: @{pop_max} </h3>",
    tile_provider='STAMEN_TERRAIN_RETINA',
    size="size",
    figsize=(900, 600),
    title="World cities with more than 1.000.000 inhabitants")

Figure(

id = '1002', …)

above = [],

align = 'start',

aspect_ratio = None,

aspect_scale = 1,

background = None,

background_fill_alpha = 1.0,

background_fill_color = '#ffffff',

below = [MercatorAxis(id='1013', ...)],

border_fill_alpha = 1.0,

border_fill_color = '#ffffff',

center = [Grid(id='1020', ...), Grid(id='1028', ...), Legend(id='1062', ...)],

css_classes = [],

disabled = False,

extra_x_ranges = {},

extra_y_ranges = {},

frame_height = None,

frame_width = None,

height = None,

height_policy = 'auto',

hidpi = True,

js_event_callbacks = {},

js_property_callbacks = {},

left = [MercatorAxis(id='1021', ...)],

lod_factor = 10,

lod_interval = 300,

lod_threshold = 2000,

lod_timeout = 500,

margin = (0, 0, 0, 0),

match_aspect = False,

max_height = None,

max_width = None,

min_border = 5,

min_border_bottom = None,

min_border_left = None,

min_border_right = None,

min_border_top = None,

min_height = None,

min_width = None,

name = None,

outline_line_alpha = 1.0,

outline_line_cap = 'butt',

outline_line_color = '#e5e5e5',

outline_line_dash = [],

outline_line_dash_offset = 0,

outline_line_join = 'bevel',

outline_line_width = 1,

output_backend = 'webgl',

plot_height = 600,

plot_width = 900,

renderers = [TileRenderer(id='1045', ...), GlyphRenderer(id='1052', ...)],

reset_policy = 'standard',

right = [],

sizing_mode = None,

subscribed_events = [],

tags = [],

title = Title(id='1003', ...),

title_location = 'above',

toolbar = Toolbar(id='1036', ...),

toolbar_location = 'right',

toolbar_sticky = True,

visible = True,

width = None,

width_policy = 'auto',

x_range = DataRange1d(id='1005', ...),

x_scale = LinearScale(id='1009', ...),

y_range = DataRange1d(id='1007', ...),

y_scale = LinearScale(id='1011', ...))

More controlled visualizations

Grammar of graphics in Python

If you love ggplot2 and just want to stick with it.

import warnings
from plotnine import *
from plotnine.exceptions import PlotnineWarning
from plotnine.data import meat

warnings.simplefilter('ignore', FutureWarning)
warnings.simplefilter('ignore', PlotnineWarning)

meat.sample(3)

	date	beef	veal	pork	lamb_and_mutton	broilers	other_chicken	turkey
321	1970-10-01	1913.0	49.0	1278.0	48.0	633.4	NaN	276.9
720	2004-01-01	1926.0	16.0	1758.0	15.5	2823.6	39.2	440.0
566	1991-03-01	1720.0	25.0	1300.0	36.0	1530.9	NaN	329.7

df = pd.melt(meat, id_vars=['date'],
             var_name='meat',
             value_name='price')

p = (
    ggplot(df, aes(x='date', y='price', color='meat')) +
    geom_line() +
    theme_xkcd() +
    labs(title="Lorem ipsum bacon")
)

p.draw();

p.save('meat.png')

from IPython.display import Image

Image('meat.png')

Similar plot in seaborn

with plt.xkcd():
    g = sns.lineplot(data=df, x='date', y='price', hue='meat')
    g.set_title('Lorem ipsum bacon')
    plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

Show as heatmap

(
    sns.heatmap(
        meat.select_dtypes('number').
        apply(lambda x: (x-x.mean())/x.std(), axis=0))
)
pass