Stasiun Pengisian Kendaraan Listrik Umum (SPKLU) Model Using GIS and Machine Learning

Abstract

The adoption of electric vehicles in Indonesia is a strategic initiative supporting the national “Go Green” agenda and the Net Zero Emission target by 2060. As electric vehicle usage continues to grow, especially in West Java, well-distributed Stasiun Pengisian Kendaraan Listrik Umum (SPKLU) locations are increasingly important. Inefficient placement may lead to operational issues, reduced user convenience, and financial losses. This research develops such a model by integrating Geographic Information System (GIS) techniques with machine learning algorithms, specifically Multi-Layer Perceptron (MLP) and Support Vector Machine (SVM). Data preparation includes collecting spatial datasets such as road networks, housing, apartments, parking areas, public facilities from OSM, administrative boundaries from Geofabric, and existing SPKLU points from OpenData West Java. Proximity analysis is used to measure distances to key features, enabling classification of potential locations into Shared-Residential, Enroute, and Destination categories. These outputs are combined with socio-economic variables, including population density, income levels, vehicle ownership, household characteristics, education levels, and age distribution processed using Kernel Density Estimation (KDE). Results show that MLP performs best with 92.8% accuracy. The most influential variable is the productive-age population, minority population, unemployment, and total population. Overall, demographic factors play a dominant role in determining optimal SPKLU locations.