Authors : Supratman Agus
Abstract: The artificial neural network model with 3 variables and 1 Hidden Layer (ANN3-1HL) is a fatality prediction model developed in 2014 for big cities in Indonesia. The equation for this model is F = 100+680.F; in which F = 2.077, -2.17.H (1:1)-0.115.H(1:2). The ANN3-1HL Model is developed from the multivariable Artificial Neural Network (ANN) Model based on regional characteristics and road infrastructure in Indonesia. It is based on an analysis of eight variables on a sample of big city area. The present study is intended to field-test the implementation of ANN3-1HL Model in 8 big cities of West Java Province in Indonesia with input variables of Population (P), Number of Vehicle (V) and Accessibility (A) in 2007-2013. The findings of the study are as follow. The ANN3-1HL Model can predict the actual fatality rate in 8 big cities up to 175.68% greater than the prediction data from the National Police Force of Republic of Indonesia with the number of prediction increasing 185.63% each year. The implementation of ANN3-1HL Model for fatality prediction in big cities is influenced by the sequence and consistency of data input from the National Police Force. Fatality prediction is less reliable if the sequence of data input is inconsistent and vice versa. The findings of the study show that there are 175.68% fatality data yet to be reported by the National Police Force (under reporting). It is recommended that road safety researchers implement the ANN3-1HL Model to predict the number of actual fatality in Indonesia.
Supratman Agus , 2018. Prediction of Actual Fatality in Big Cities of West Java Province based on ANN3-1HL Model Developed from the Multivariable ANN Model for Fatality Prediction in Indonesia. Journal of Engineering and Applied Sciences, 13: 3463-3471.