Road traffic accidents models using threshold levels of fuzzy linear regression

Abstract

It has been hypothesized that number of road traffic accidents and road casualties are increased in line with the rapid recent increase in the variables of registered vehicles, population and road length. However the effects of these variables toward road traffic accidents are still inconclusive. Therefore, this paper develops models based on the variables which can be used to determine number of road traffic accidents in Malaysia. In order to explain the effects of these variables to road traffic accident, fuzzy linear regression models with three threshold levels, h=0.1, 0.5, 0.9 are examined. Historical data from the year 1974 to 2007 were collected to test performances of the models. The results show that by applying a multi-variable approach to fuzzy linear regression, the model provides not only crisp output but also output range for road traffic accident in Malaysia. The fuzzy linear regression model with threshold level h=0.1 was outperformed the other two models. The variables of registered vehicles and population were notable predictors to number of road traffic accidents in Malaysia.