Tuesday, 9 April 2013

Can We Predict Presidential Elections?


I came across this article last week as I was reading some material about machine learning. Chris Wilson holds that computers cannot predict the results of the presidential election. I believe the opposite but we probably cannot decide yet. Machine learning is a mathematical model that can perform tasks that are impossible to humans (coursera link).

This is the case with machine learning. They are used to finding some structure in data. Programmers design these machines to find common patterns and similar features within a huge amount of data, that a human obviously wouldn't be able to process. Unsupervised machines are utilised in many different fields; here are some examples.
Coming back to the initial issue I would like to recommend reading this article. The researchers in this study use twitter posts to analyse people's sentiments and opinions. Why could we not do the same thing to predict the presidential elections, or at least, to predict them in a social networking environment?

4 comments:

  1. Here is a BBC article in which Nate Silver of the New York Times explains the science of presidential predictions

    http://www.bbc.co.uk/news/magazine-20246741

    ReplyDelete
  2. There was a study done by a secondary school student in Dublin a few years ago using Twitter to map the mood of a nation. It won a prize in the BT Young Scientist and Technology Exhibition. This is the link: http://thevibesofireland.com/ Some of the graphs are very interesting, especially the one showing reaction to Ireland agreeing to financial support from the EU.

    ReplyDelete
  3. Game theory is often used for strategic decision making and predicting behavior. Here is an interesting article that applies game theory for predicting political behavior: http://www.jstor.org/stable/pdfplus/2132065.pdf?acceptTC=true .

    ReplyDelete