RIASSUNTO
ABSTRACT
A three-classification model is established to analyze factors that have influence on severity of water traffic accidents with SVM, on this basis, this paper made an empirical combined with Cross Validation by use of the traffic accident data from Shanghai Maritime Safety Administration, which improve the accuracy of the forecast data was 70%. Then the SVM-RFE algorithm is used to get the weight value of the factors and confirms the most influence factor.
INTRODUCTION
At water transportation is more and more paid great attention in China today, the number of water traffic accidents also in growing larger, more and more people lost their life in water traffic accidents. So water traffic safety is becoming a major concern, and there are many people did researches on how to improve the water traffic safety degree. It's necessary for us to analysis water traffic accident severity, though that we can know the relationships between water traffic accident severity and various explanatory factors such as the type of the accidents, the time of accidents occurred, the size of the ship and so on which can help us understand the impacts of contributing factors on water traffic accident severity.
Thus far, many studies have been conducted in this area, and their study focused on a particular type of ships. WK Talley et al. (2008a) used a Tobit regression and the ordered Probit model to estimate the damage costs and injury of ferry vessel accident and concluded that allision, collision and fire ferry vessel accident incur more property damage and human error leads to heaveier casualties. D Jin (2014) utilized the ordered Probit model to investigate the determinants of fishing ship accident severity, result showed that fishing vessel damage severity is positively associated with several factors like wind speed during the day, vessel age, and distance to the shore. TL Yip et al. (2015) used Poisson regression to investigate determinants of casualties in passenger ship accident, result showed that the number of passenger injuries is positively related to the number of crew injuries in ferry, ocean cruise and river cruise passenger vessel accidents. Yishu Zheng et al. (2016) investigated determinants of the probability of non-fatal and fatal crew injuries in container ship accidents. WK Talley et al. (2008b) also investigated factors that have influence on cruise ship accident severity. Jinxian Weng et al. (2015) used a binary logistic regression model to describe water traffic accident severity and concluded that the probability of fatal accidents and mortalities are greater for accidents, which occurr in adverse weather and night.