Introduction

In companies, institutes or at home we generate large amounts types of data. A success of our business or research may depend on a proper handling of the data. We can extract meaningful patterns by applying data reduction and analysis techniques, but eventually we must present the data graphically. Many software packages allow to create a simple line plots or bar charts, but creating data-dense visualisations without distortions is still more of an art than science. The goal of the workshop is to arm you with practice-oriented tips that will help you to avoid clutter and increase the data density of your graphs.

matplotlib is a de facto standard in 2D plotting with Python in active development since 2003. The large number of visualisation types is not matched by any other plotting library available for Python. It can be used to create interactive visualisations, hard copy plots or standalone applications.

The tutorial will introduce the basic theory of data visualisation and put it in use through matplotlib. To unleash the full power of matplotlib, we will reach under the hood and discover some hidden gems in terms of customisation and working with visual primitives. The participants will be encouraged to practice their visualisations skills trough a series of examples. They will learn how to build complex data visualisations from ground up and spice them with a bit of interactivity.

Schedule

0) Introduction

• introduction to data visualisation
• matplotlib and its history

1) Visualising patterns over time

• quick interactive data exploration
• pyplot interface
• matplotlib backends
• time series plotting and smoothing
• working with dates

2) Visualising proportions

• matplotlib object-oriented model
• creating visual style
• bar plots

3) Visualising distributions

• matplotlib transforms
• histograms, boxplots and kernel density estimators

4) Visualising correlations

5) Finding patterns

6) Making maps

Materials

Slides

Notebooks

Handouts

See also slides, examples and exercises from my previous tutorial

Prerequisites

Required software:

• python 2.7 or 3.x
• matplotlib
• numpy
• pandas

Optional software:

• scipy
• cartopy
• scikit-learn
• statsmodels
• lxml

All of the libraries are available in standard linux distributions, via pip or in scientific python distributions (such as anaconda).

You can also install them with pip (for example inside virtual environment):

mkdir ~/.virtualenv
virtualenv ~/.virtualenv/ep2014_mpl          #crete virtualenv
source ~/.virtualenv/ep2014_mpl/bin/activate #activate it (in bash)
pip install -U numpy scipy matplotlib        #absoulte basics 
pip install -U ipython pyzmq jinja2          #for ipython notebook
pip install -U pandas                        #used in many examples
pip install -U scikit-learn lxml patsy       #extras
pip install -U statsmodels 

About the trainer

Bartosz Telenczuk has been active Python user since 2005. He is creator of svgutils and he has contributed to many open source Python libraries including numpy and matplotlib. He is also a Python advocate and co-organiser of advanced Python schools for scientists. Currently he is a researcher in France, developing methods to interpret the electrical activity of the brain.