FATALITY PREDICTION STUDY OF ROAD VICTIMS’ WITH MULTI VARIABLE BASED ON POPULATION DENSITY IN WEST JAVA PROVINCE

Abstract

n Indonesia, data of fatality is not known with certainty. In ASEAN, Indonesia has the largest of area, the largest of population, the longest of road infrastructure, and the highest number of vehicles, but road victims‟ fatality is low. Data reported by Indonesian state police are the victims at the scene, has not completed the data from hospital. A lot of estimated fatality data has not been recorded and reported. Such data is not fit for use in the study of road safety. In many countries, the researchers only used Andreassen model with two variable input data to predict the actual number of fatality, which are the population and number of vehicles. This model is not in accordance with the conditions in Indonesia, so that the new model with multivariable is required which suits Indonesia‟s conditions. The purpose of this study is to determine the predictions number of fatalities factual with multiple variables by considering the characteristics of the population density/km2 in West Java, using Artificial Neural Network (ANN) predictive models. After obtained the validation testing with error model test, the number of factual predictions fatality in 2010 was 1546 people, more than 94% from the police data which are 797 people, or 20.1% more than the police data with the fatality amount at the hospital.