Assessing the Predictive Abilities of Statistical Injury-Severity Prediction Modelling Considering non-behavioral Factors of Accident

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title




In the developing country, Road Traffic Accidents are among the leading cause of death and injury; Ethiopia in particular experiences the highest rate of such accidents. Road Traffic Accidents cannot be absolutely eradicated, however, it is possible to prevent them to some extent as long as the contributing factors are identified and tackled appropriately. The driver behavior plays a crucial role in the occurrence of a crash; but, it is usually complex and unpredictable, and also focusing too much on the driver as the cause of a crash often masked the ability to see other causes that could reduce crash rates and crash severity. So it is important to figure out the role of the non-behavioral factors in traffic accidents, based on which cost-effective countermeasures can be recommended to reduce the chance of accidents. Previously, significant studies were undertaken to predict the magnitude of road traffic accidents and black spot areas in Addis Ababa City considering the driver as the main causing factor. However significant studies were not undertaken to predict the magnitude of accident severity considering the non-behavioral factors as the main causing factor. As a result, this study developed accident severity prediction models using Multinomial logistic regression that link accident severity to non-behavioral contributing factors. For this study, a total of 5251 traffic accident data from June 30/2011 to June 30/2016GC were collected from Yeka sub-city. Multinomial logistic regression was used to estimate the model parameters. The data set obtained for this study were applied to examine the goodness-of-fit regression models. The dependent variable used in this study was crash severity. As part of the study, the models have been tested to see how well they predict the accidents observed during a one year accident period. From the model developed, road type, road surface condition, crash type, maneuvering condition and lighting conditions were found as significant explanatory variables that influence the prediction of crashes in the yeka sub-city. This indicates that non-behavioral factors have an effect on the occurrence of accident.



Road traffic accident, Non-behavioral factors, Accident severity Prediction Models, Multinoial Logistic regression, , Crash Severity level