Data visualization with Pandas¶

Pandas implements some high level plotting functions using matplotlib. Note If you have seaborn imported, pandas will relay the plotting through seaborn and you get better looking plots for the same data and commands.

In [1]:

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

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

Read some data¶

In [2]:

df1 = pd.read_csv('/Users/atma6951/Documents/code/pychakras/pychakras/udemy_ml_bootcamp/Python-for-Data-Visualization/Pandas Built-in Data Viz/df1', index_col=0)
df2 = pd.read_csv('/Users/atma6951/Documents/code/pychakras/pychakras/udemy_ml_bootcamp/Python-for-Data-Visualization/Pandas Built-in Data Viz/df2')

df1 = pd.read_csv('/Users/atma6951/Documents/code/pychakras/pychakras/udemy_ml_bootcamp/Python-for-Data-Visualization/Pandas Built-in Data Viz/df1', index_col=0) df2 = pd.read_csv('/Users/atma6951/Documents/code/pychakras/pychakras/udemy_ml_bootcamp/Python-for-Data-Visualization/Pandas Built-in Data Viz/df2')

In [3]:

df1.head()

df1.head()

Out[3]:

	A	B	C	D
2000-01-01	1.339091	-0.163643	-0.646443	1.041233
2000-01-02	-0.774984	0.137034	-0.882716	-2.253382
2000-01-03	-0.921037	-0.482943	-0.417100	0.478638
2000-01-04	-1.738808	-0.072973	0.056517	0.015085
2000-01-05	-0.905980	1.778576	0.381918	0.291436

In [4]:

df2.head()

df2.head()

Out[4]:

	a	b	c	d
0	0.039762	0.218517	0.103423	0.957904
1	0.937288	0.041567	0.899125	0.977680
2	0.780504	0.008948	0.557808	0.797510
3	0.672717	0.247870	0.264071	0.444358
4	0.053829	0.520124	0.552264	0.190008

3 ways of calling plot from a DataFrame¶

• df.plot() and specify the plot type, the X and Y columns etc
• df.plot.hist() calling plot in OO fashion. Only specify teh X and Y and color or size columns
• df['column'].plot.plotname() - calling plot on a series

Types of plot that can be called: area, bar, line, scatter, box, hexbin, kde etc.

Ways of plotting histogram¶

In [8]:

df1.plot(x='A', kind='hist')

df1.plot(x='A', kind='hist')

Out[8]:

<matplotlib.axes._subplots.AxesSubplot at 0x1142ece80>

In [10]:

df1['A'].plot.hist(bins=30)

df1['A'].plot.hist(bins=30)

Out[10]:

<matplotlib.axes._subplots.AxesSubplot at 0x1169717f0>

Plotting a histogram of all numeric columns in the dataframe:¶

In [7]:

df1.hist()

df1.hist()

Out[7]:

array([[<matplotlib.axes._subplots.AxesSubplot object at 0x11a872eb8>,
        <matplotlib.axes._subplots.AxesSubplot object at 0x11abd27f0>],
       [<matplotlib.axes._subplots.AxesSubplot object at 0x11abf7e80>,
        <matplotlib.axes._subplots.AxesSubplot object at 0x11ac27550>]],
      dtype=object)

In reality, you have a lot more columns. You can prettify the above by creating a layout and figsize:

In [8]:

ax_list = df1.hist(bins=25, layout=(2,2), figsize=(7,7))
plt.tight_layout()

ax_list = df1.hist(bins=25, layout=(2,2), figsize=(7,7)) plt.tight_layout()

Plotting histogram of all columns and sharing axes¶

The chart above might make more sense if you shared the X as well as Y axes for different columns. This helps in comparing the distribution of values visually.

In [14]:

ax_list = df1.hist(bins=25, sharex=True, sharey=True, layout=(1,4), figsize=(15,4))

ax_list = df1.hist(bins=25, sharex=True, sharey=True, layout=(1,4), figsize=(15,4))

In [15]:

ax_list = df1.hist(bins=25, sharex=True, sharey=True, layout=(2,2), figsize=(8,8))

ax_list = df1.hist(bins=25, sharex=True, sharey=True, layout=(2,2), figsize=(8,8))

Backgrounds¶

You can specify dark or white background and style info to the matplotlib that is used behind the scenes.

Area plot¶

In [13]:

plt.style.use('dark_background')
df2.plot.area()

plt.style.use('dark_background') df2.plot.area()

Out[13]:

<matplotlib.axes._subplots.AxesSubplot at 0x116b35f98>

Bar chart¶

Another style is fivethirtyeight

In [14]:

plt.style.use('fivethirtyeight')
df2.plot.bar()

plt.style.use('fivethirtyeight') df2.plot.bar()

Out[14]:

<matplotlib.axes._subplots.AxesSubplot at 0x116c92cc0>

Line plot¶

This is suited for time series data

In [18]:

#reset the style
plt.style.use('default')

# pass figsize to the matplotlib backend engine and `lw` is line width
df1.plot.line(x=df1.index, y='A', figsize=(12,2), lw=1)

#reset the style plt.style.use('default') # pass figsize to the matplotlib backend engine and `lw` is line width df1.plot.line(x=df1.index, y='A', figsize=(12,2), lw=1)

Out[18]:

<matplotlib.axes._subplots.AxesSubplot at 0x116f61f98>

Scatter plot¶

Use colormap or size to bring in a visualize a 3rd variable in your scatter

In [21]:

df1.plot.scatter(x='A', y='B',c='C', cmap='coolwarm')

df1.plot.scatter(x='A', y='B',c='C', cmap='coolwarm')

Out[21]:

<matplotlib.axes._subplots.AxesSubplot at 0x1176e09b0>

In [22]:

# you could specify size s='c' however the points come out tiny.
# had to scale it by 100, hence using actual series data and not the column name
df2.plot.scatter(x='a',y='b', s=df2['c']*100)

# you could specify size s='c' however the points come out tiny. # had to scale it by 100, hence using actual series data and not the column name df2.plot.scatter(x='a',y='b', s=df2['c']*100)

Out[22]:

<matplotlib.axes._subplots.AxesSubplot at 0x117440f60>