Applying Hidden Markov Models to Voting Advice Applications
Cyprus University of Technology, 30, Arch. Kyprianos str., Limassol, 3036, Cyprus
* e-mail: firstname.lastname@example.org
Accepted: 17 November 2016
Published online: 13 December 2016
In recent times, a phenomenon that threatens the representative democracy of many developed countries is the low voter turnout. Voting Advice Applications (VAAs) are used to inform citizens about the political stances of the parties that involved in the upcoming elections, in an effort to facilitate their decision making process and increase their participation in this democratic process. VAA is a Web application that calls the users and parties to state their position in a set of policy statements, which are based on the current affairs of their country and then it recommends to each user the party that better fits their political views. SVAAs, a social recommendation approach of VAAs, on the other hand, base their recommendation on the VAA community of users in a collaborative filtering manner. In this paper we resort to Hidden Markov Models (HMMs) in an attempt to improve the effectiveness of SVAAs. In particular, we try to model party-supporters using HMMs and then use these models to recommend each VAA user the party whose model best fits his/her answer sequence of the VAA policy statements. HMMs proved to be effective machine learning tools for sequential and correlated data and this is the main rationale behind this study. VAA policy statements are usually correlated and grouped into categories such as external policy, economy, society, etc. As a result, opting from the various answer choices in each policy statement might be related with selections in previous and subsequent policy statements. Given that the order of policy statements is kept fixed within each VAA one can assume that (a) answer patterns (sequences of choices for all policy statements included in the VAA) can be found that characterise ‘typical’ voters of particular parties, and (b) the answer choice in each policy statement can be ‘predicted’ from previous answer choices. For our experiments we use three datasets based on the 2014 elections to the European Parliament (http://www.euvox2014.eu/), which are publicly available through the Preference Matcher website (http://www.preferencematcher.org/?page_id=18).
Key words: Hidden Markov Models / Voting Advice Applications / collaborative filtering / expectation maximisation / recommender systems
© The Author(s), 2016