What is Predikon?
Predikon is a website born out of an academic project by the Information and Network Dynamics Laboratory (INDY Lab) at the Swiss Federal Institute for Technology in Lausanne (EPFL). Its main goal is to provide predictions about votes in Switzerland, as soon as municipal results are published by Swiss cantons starting at 12:00pm on the day of a vote.
As an example, we acurately predicted (a difference of less than 1%) the final result of the Law against sexual discrimination one minute after noon, when only 9% of the citizen ballots were counted. We are actively working on extending the functionalities available on the web site, notably to improve the accuracy of our predictions.
How do your predictions work?
We designed a predictive model of the municipal results (in percentage of "yes") for Swiss referendum votes. Based on past results, our model automatically learns representations of municipalities. On the day of a vote, and as soon as some municipalities publish their results (their percentage of "yes"), our model uses the representations to predict the unpublished results of other municipalities. The prediction of the national vote result is obtained by aggregating published results and predicted results.
Where do the data come from?
We obtain both the past and the live data from the new API developed by the Swiss Federal Office for Statistics.
Where can I find the former version of Predikon?
It is availbale here. The former version of Predikon contains interesting visualizations of voting patterns in Switzerland.
Who are you?
We are a team of researchers from the INDY Lab at EPFL.
How can I contact you?
Acknowledgements
We are grateful to the Swiss Federal Statistical Office for their transparency and their help in understanding their data. We thank Ragnor Comerford for helping us setting up the new version of this web site. We thank Brunella Spinelli and Young-Jun Ko for helping us translating the web site into Italian and German. We give an honorable mention to Vincent Etter and Julien Herzen who initiated this project in 2014 and who provided valuable feedback all along its continuous development.
Relevant Publications
A. Immer, V. Kristof, M. Grossglauser, P. Thiran, Sub-Matrix Factorization for Real-Time Vote Prediction, Knowledge Discovery and Data Mining (KDD), 2020.
V. Etter, E. Khan, M. Grossglauser, P. Thiran, Online Collaborative Prediction of Regional Vote Results, Conference on Data Science and Advanced Analytics (DSAA), 2016.
V. Etter, J. Herzen, M. Grossglauser, P. Thiran, Mining Democracy, Conference on Online Social Networks (COSN), 2014.