Evaluation of the Impact of Different Variables on Motocyclists' Safety

Motorcyclists due to lack of protection at the scene of the crash and also being less exposed in the eyesight of other vehicles are more vulnerable for sever crashes. The main purpose of this study, is to identify the factors affecting the severity of motorcycle crashes.

Neural network, random forest and statistical analysis (comparison of the average severity of accidents in variables) have been used in this paper. To conduct this research, two series of data (accident data and behavioral data collected from motorcyclists in Tehran based on field interview) have been used.

The severity of crashes increases with the rise of the frequency of speeding, passing red lights, using mobile phones, performing dangerous movements, not usign helmet and crossing the pedestrain pathway. However, with the increase in the health status of motorcyclists and the technical appearance of the motorcycle the severity of accidents will decrease. Among the districts of Tehran, district 6 had the highest frequency of crahses. Neural network model, is able to predict fatalities, injuries and propoerty damage crashes with 24, 76 and 51 percent accuracy. Furthermore, in the random forest model, the accuracy of predicting fatal, injury and property damage crashes are 53, 69 and 32 percent.

It indicate that the traffic violations of the motorcyclists are the main reasons for crash severities. Thus, it is necessary to have an overlook about performing more restrict traffic fines for this group. In addition, becasuse of the importance of the individual health condition and technical appearance of the motorcycle it is necessary to determine rules for controling them with issuing cycling licence and other intervals after that. Finally, the high rate of motorcycle crashes in the district of 6 in Tehran can be related to high capacity rate of transit  in the zone.