Supervised Learning Analysis (Prediction): Vote Margin per District – Sragen Local Election 2024

Abstract

This study examines whether the difference in votes between candidate pairs in the 2024 Sragen Regional Election can be predicted using basic administrative data at the sub-district level. The analysis was carried out quantitatively with analysis units of 20 sub-districts in Sragen Regency. For each sub-district, the composition of voters (number of male voters, number of female voters, total Permanent Voter List), voter participation (participation rate, calculated as the number of valid votes divided by the total registered voters), and electoral results (difference in votes, defined as the votes of Candidate 1 minus the votes of Candidate 2). The purpose of this study is to explain why some sub-districts produce a very large difference in votes for one candidate (strong base), while other sub-districts are very tight and only differ by hundreds of votes (swing areas). The results of the study show three main patterns. First, sub-districts with large numbers of voters and high participation tend to produce wide vote differences for dominant candidates. Second, sub-districts with a difference of only hundreds of votes can be identified as competitive areas that can easily change direction if there is a slight change in voter mobilization. Third, the multiple linear regression model with variables of total registered voters, participation rate, and proportion of male voters was able to explain about 38% of the variation in vote difference between sub-districts. These findings show that the difference in votes is not just the final result of the count, but is related to the ability to mobilize voters in densely populated areas as well as to the demographic structure of voters in those areas.