Determining and Clustering Potential Legislative Candidate in West Java District Using K-Nearest Neighbors Algorithm

Abstract

Indonesia held its first general election in 1955 to elect legislatures from all provinces. The latest was held in 2014, which elected 560 members to the People's Representative Council (Dewan Perwakilan Rakyat, DPR) and 128 to the Regional Representative Council (Dewan Perwakilan Daerah, DPD). The PRC was elected by proportional representation from multi-candidate constituencies/districts. Currently, there are 77 constituencies in Indonesia, each of which returns 3-10 Members of Parliament based on population. Under Indonesia's new multi-party system, no party has been able to secure an outright victory; hence, selecting the right candidate for the right constituencies has been a major effort for all participating parties. Many combinations have been tried; popularities, intelligence, public figures, ‘putera daerah’ are all variables that can only show a fraction of winning pattern where no general conclusion can be drawn. This research used data mining techniques to create an unfound pattern, and to suggest which particular legislative candidate is most suitable for which constituency. Using 11 West Java constituencies (11 clusters), K-Nearest Neighbors (K-NN) algorithms, we found out that an 83.33% accuracy using data from 2014 general election.