Thursday October 28 12:30 PM – Thursday October 28 1:00 PM in Talks I

Why *Interactive* Data Visualization Matters for Data Science in Python

Nicolas Kruchten

Prior knowledge:
No previous knowledge expected

Summary

Data visualization is a powerful tool for data analysis, although interactivity is often perceived as frivolous or necessary only to communicate with non-data-scientists. This talk will challenge that notion by showing multiple examples of how Python makes it easy to build interactive data visualization that can lead to deeper insights faster than relying only on code and static visualizations.

Description

  • Interactivity in data visualization is sometimes taken by data scientists working in Python as unnecessary, frivolous or mostly useful for communicating with stakeholders when the real work is done.
  • My main claim in this talk is that interactive data visualization is a powerful tool for data scientists throughout their data analysis work, and I want to unpack specific examples of how interactivity can help you get to deeper insights, faster, without having to write a lot more code.
  • I will do this by examining interactive data visualizations as a progression through a series of levels, with examples.
  • Level 0 is the interaction loop of modifying a notebook cell to produce a visual, then modifying that cell repeatedly to modify the figure in response to what the last one showed.
  • Level 1 is the use of widgets like dropdowns and sliders around a visual to do the same kinds of things as in level 0, but faster and more smoothly, with a mouse.
  • Level 2 involves direct interaction with a figure: panning, zooming, selecting legend items, hovering and selecting, and I will show how this type of interaction is a very powerful and useful way to get insights about a data set, and far faster and more intuitive than Levels 1 or 0.
  • Level 3 involves linking multiple visual representations of data together to build a tailor-made user interface to solve a data problem. I will demonstrate the common crossfiltering pattern and show how it can reveal patterns.