Skip to content

GSoC_2017_applications_elections

Jump to bottom Edit New page
Lea Goetz edited this page Feb 8, 2017 · 17 revisions

Example Data Project - Elections

The idea of the data project is to use ML in general - and of course Shogun in particular - on “real world” data and demonstrate that it its useful beyond the recognising handwritten digits or iris species.

Mentors

Description

2016 saw some spectacular popular votes, such as the UK “Brexit” referendum and the US presidential elections. Importantly, the outcomes of these votes were deemed surprising. For example, in the election of Donald Trump as 45th president of the United States major news coverage - up until very late during election night - predicted a win for Hillary Clinton (see below). Why should it be so difficult to predict understand and predict voters’s behaviour - even if there are only 2 options - when on the other hand whole personalities can be reconstructed from someone’s facebook preferences? Europe has a year of crucial elections ahead of itself, for example France presidential elections, Germany parliamentary elections. The challenge of this GSoC Data Project will be to use Shogun tools in addition to classical data analysis of voter characteristics and understand the election results. Depending on your interest, we can put the focus on mapping voter characteristics or predicting/explaining election outcomes (see for example the Digital Times analysis of the UK 2015 elections which explains their method and even provides code).

Waypoints and initial work

  • Identify datasets on voter demographics, such as census data, GIP, education levels, as well as surveys on voting intentions for both previous elections most current data
  • Clean data and map voter features
  • Find predictors for voting behaviour in each constituency
  • Use ALGORITHM GOES HERE
  • Predict an upcoming election/ explain a previous election
  • Visualize the process and results

Why this is cool

With a real world data project we can - finally - use the power of Shogun on data that really matter. In particular, understanding voter behaviour and predicting election outcomes is not just (one) holy grail of predicition; the actual events have a huge impact on a lot of people's lifes. So making the right prediction or providing a good explanation for the result really matters!

Useful resources

Add a custom footer
Clone this wiki locally