Optimasi Self-Organizing Map Menggunakan Particle Swarm Optimization untuk Mengelompokkan Desa/Kelurahan Tertinggal di Kabupaten Kutai Kartanegara Provinsi Kalimantan Timur

Studi Kasus : Data Potensi Desa Tahun 2018

Abstract

Self-Organizing Maps (SOM) is an efficient cluster analysis in handling high dimensional and large dataset. Particle Swarm Optimization (PSO) is an effective in nonlinear optimization problems and easy to implement. A clustering process occurs if all data can clustered into 1 cluster, however if one or two data did not join then the data have a deviant behavior called outliers or noise.  PSO is used to evolve the weights for SOM to improve the clustering result and to cluster some social aspect in society, for example is poverty. Development strategies are prioritized to regions with largest population lived in poverty. Kutai Kartanegara regency  (Kukar) are recorded as the biggest contributor on population lived in poverty at East Kalimantan in 2017. Development of underdeveloped villages is requires Village Potential data, which focus on visualizing the situation in the regions. This study aims to determine the number of clusters formed and to find the value of Davies Bouldin Index (DBI) from clustering underdeveloped villages in Kukar region using PODES 2018 data. This study uses 9 particle which are the final weight of the SOMs algorithm with different learning rate each particle. Based on the analysis, the optimal number of clusters is 2 clusters with DBI value of 0.7803, where cluster 1 consists of 82 underdeveloped villages and the cluster 2 consist of underdeveloped villages.