engleski

Machine learning based prediction of Croatian 2017. local elections

Internet development enabled intensive applications of new ways of communication. In today’s digital age, social media became fundamental mean of communication. Political parties are increasingly using social media for the purpose of advertising and voters` mobilization. An extensive literature review has identified multiple benefits of social media usage, such as gaining publicity, spreading messages and mobilizing voters, but also the need to monitor the content that is published, and moreover to analyze the impact of that content on potential voters. This paper examines the influence of political candidates ; activities on social media on elections outcome. By means of decision tree methodology, predictive model of election outcome is developed based on dataset consisting of candidates` characteristics, but also data of the candidates` activity on the social network Facebook. The predictive model is developed on dataset consisting of candidates in the local elections at large Croatian cities held in Croatia in 2017. The research identified the most important factors of political communication on the social network Facebook for election outcome and provides guidelines for the effective use of Facebook in political campaigns.

machine learning ; predictive model ; social network data ; political elections

nije evidentirano