Article Outline
Example Python program pareto.py
Modules
- from operator import itemgetter
- import matplotlib.pyplot as plt
Methods
- def pareto(data, labels=[], cumplot=True, axes=None, limit=1.0,
- def test():
Code
Python example
from operator import itemgetter
import matplotlib.pyplot as plt
def pareto(data, labels=[], cumplot=True, axes=None, limit=1.0,
data_args=(), data_kw={}, line_args=(), line_kw={},
limit_kw={}):
"""
Plots a `pareto chart`_ of input categorical data. NOTE: The matplotlib
command ``show()`` will need to be called separately. The default chart
uses the following styles:
- bars:
- color = blue
- align = center
- width = 0.9
- cumulative line:
- color = blue
- linestyle = solid
- markers = None
- limit line:
- color = red
- linestyle = dashed
Parameters
----------
data : array-like
The categorical data to be plotted (not necessary to put in descending
order prior).
Optional
--------
labels : list
A list of strings of the same length as ``data`` that provide labels
to the categorical data. If none provided, a simple integer value is
used to label the data, relating to the original order as given. If
a list is provided, but not the same length as ``data``, then it will
be treated as if no labels had been input at all.
cumplot : bool
If ``True``, a cumulative percentage line plot is included in the chart
(Default: True) and a second axis indicating the percentage is returned.
axes : axis object(s)
If valid matplotlib axis object(s) are given, the chart and cumulative
line plot of placed on the given axis. Otherwise, a new figure is
created.
limit : scalar
The cumulative percentage value at which the input data should be
"chopped off" (should be a value between zero and one).
data_args : tuple
Any valid ``matplotlib.pyplot.bar`` non-keyword arguments to apply to
the bar chart.
data_kw : dict
Any valid ``matplotlib.pyplot.bar`` keyword arguments to apply to
the bar chart.
line_args : tuple
Any valid ``matplotlib.pyplot.plot`` non-keyword arguments to apply to
the cumulative line chart.
line_kw : dict
Any valid ``matplotlib.pyplot.plot`` keyword arguments to apply to
the cumulative line chart.
limit_kw : dict
Any valid ``matplotlib.axes.axhline`` keyword arguments to apply to
the limit line.
Returns
-------
fig : matplotlib.figure
The parent figure object of the chart.
ax1 : matplotlib.axis
The axis for the categorical data.
ax2 : matplotlib.axis
The axis for the cumulative line plot (not returned if
``cumplot=False``).
Examples
--------
The following code is the same test code if the ``paretoplot.py`` file is
run with the command-line call ``$ python paretoplot.py``::
# plot data using the indices as labels
data = [21, 2, 10, 4, 16]
# define labels
labels = ['tom', 'betty', 'alyson', 'john', 'bob']
# create a grid of subplots
fig,axes = plt.subplots(2, 2)
# plot first with just data
pareto(data, axes=axes[0, 0])
plt.title('Basic chart without labels', fontsize=10)
# plot data and associate with labels
pareto(data, labels, axes=axes[0, 1], limit=0.75, line_args=('g',))
plt.title('Data with labels, green cum. line, limit=0.75', fontsize=10)
# plot data and labels, but remove lineplot
pareto(data, labels, cumplot=False, axes=axes[1, 0],
data_kw={'width': 0.5, 'color': 'g'})
plt.title('Data without cum. line, bar width=0.5', fontsize=10)
# plot data cut off at 95%
pareto(data, labels, limit=0.95, axes=axes[1, 1], limit_kw={'color': 'y'})
plt.title('Data trimmed at 95%, yellow limit line', fontsize=10)
# format the figure and show
fig.canvas.set_window_title('Pareto Plot Test Figure')
plt.show()
.. _pareto chart: http://en.wikipedia.org/wiki/Pareto_chart
"""
# re-order the data in descending order
data = list(data)
n = len(data)
if n!=len(labels):
labels = range(n)
ordered = sorted(zip(data, labels), key=itemgetter(0), reverse=True)
ordered_data = [dat for dat, lab in ordered]
ordered_labels = [lab for dat, lab in ordered]
# allow trimming of data (e.g. 'limit=0.95' keeps top 95%)
assert 0.0 <= limit <= 1.0, 'limit must be a positive scalar between 0.0 and 1.0'
# create the cumulative line data
line_data = [0.0]*n
total_data = float(sum(ordered_data))
for i, dat in enumerate(ordered_data):
if i==0:
line_data[i] = dat/total_data
else:
line_data[i] = sum(ordered_data[:i+1])/total_data
# determine where the data will be trimmed based on the limit
ltcount = 0
for ld in line_data:
if ld<limit:
ltcount += 1
limit_loc = range(ltcount+1)
limited_data = [ordered_data[i] for i in limit_loc]
limited_labels = [ordered_labels[i] for i in limit_loc]
limited_line = [line_data[i] for i in limit_loc]
# if axes is specified, grab it and focus on its parent figure; otherwise,
# create a new figure
if axes:
plt.sca(axes)
ax1 = axes
fig = plt.gcf()
else:
fig = plt.gcf()
ax1 = plt.gca()
# create the second axis
if cumplot:
ax2 = ax1.twinx()
# plotting
if not data_kw.has_key('align'):
data_kw['align'] = 'center'
if not data_kw.has_key('width'):
data_kw['width'] = 0.9
ax1.bar(limit_loc, limited_data, *data_args, **data_kw)
if cumplot:
ax2.plot(limit_loc, [ld*100 for ld in limited_line], *line_args,
**line_kw)
ax1.set_xticks(limit_loc)
ax1.set_xlim(-0.5,len(limit_loc)-0.5)
# formatting
if cumplot:
# since the sum-total value is not likely to be one of the tick marks,
# let's make it the top-most one, regardless of label closeness
ax1.set_ylim(0, total_data)
loc = ax1.get_yticks()
newloc = [loc[i] for i in range(len(loc)) if loc[i]<=total_data]
newloc += [total_data]
ax1.set_yticks(newloc)
ax2.set_ylim(0, 100)
if limit<1.0:
xmin,xmax = ax1.get_xlim()
if not limit_kw.has_key('linestyle'):
limit_kw['linestyle'] = '--'
if not limit_kw.has_key('color'):
limit_kw['color'] = 'r'
ax2.axhline(limit*100, xmin-1, xmax-1, **limit_kw)
# set the x-axis labels
ax1.set_xticklabels(limited_labels)
# adjust the second axis if cumplot=True
if cumplot:
yt = [str(int(it))+r'%' for it in ax2.get_yticks()]
ax2.set_yticklabels(yt)
if cumplot:
return fig,ax1,ax2
else:
return fig,ax1
def test():
# plot data using the indices as labels
data = [21, 2, 10, 4, 16]
# define labels
labels = ['tom', 'betty', 'alyson', 'john', 'bob']
# create a grid of subplots
fig,axes = plt.subplots(2, 2)
# plot first with just data
pareto(data, axes=axes[0, 0])
plt.title('Basic chart without labels', fontsize=10)
# plot data and associate with labels
pareto(data, labels, axes=axes[0, 1], limit=0.75, line_args=('g',))
plt.title('Data with labels, green cum. line, limit=0.75', fontsize=10)
# plot data and labels, but remove lineplot
pareto(data, labels, cumplot=False, axes=axes[1, 0],
data_kw={'width': 0.5, 'color': 'g'})
plt.title('Data without cum. line, bar width=0.5', fontsize=10)
# plot data cut off at 95%
pareto(data, labels, limit=0.95, axes=axes[1, 1], limit_kw={'color': 'y'})
plt.title('Data trimmed at 95%, yellow limit line', fontsize=10)
# format the figure and show
fig.canvas.set_window_title('Pareto Plot Test Figure')
plt.show()
if __name__=='__main__':
test()Useful Links
- Articles: https://python-commandments.org/
- Python shell: https://bsdnerds.org/learn-python/
- Tutorial: https://pythonprogramminglanguage.com/