PERBANDINGAN MODEL ANDREASSEN DAN MODEL ARTIFICIAL NEURAL NETWORK UNTUK PREDIKSI FATALITAS KORBAN KECELAKAAN LALULINTAS

Abstract

In ASEAN, Indonesia has largest area and population, longest road infrastructure, and largest number of motor vehicles, but road victims’ fatality is low. This indicates under-reporting. Existing model of road victims’ fatality is Andreassen models which use population numbers and motor vehicles numbers as
variables to prediction numbers of fatality. This study aimed to obtaining the best predictive model of road victims’ fatality which suits Indonesia’s conditions. Three models were compared are Andreassen model, Artificial Neural Network with two variables (ANN2) and four variables (ANN4), with driving license holder
and road length as two additional variables. Model validation was performed on three cities in West Java with different categories population densities. The results reveal that ANN4 is the best fatality prediction model. In addition, predictions of road victim numbers in Indonesia are not only influenced by population
and vehicles number, but also by driving license holder numbers and road length.

Keywords: fatality, model comparison, Andreassen model, Artificial Neural Network model