In the Melbourne metropolitan area in Australia, an average of 34 pedestrians were killed in traffic accidents every year between 2004 and 2013, and vehicle-pedestrian crashes accounted for 24% of all fatal crashes. Mid-block crashes accounted for 46% of the total pedestrian crashes in the Melbourne metropolitan area and 49% of the pedestrian fatalities occurred at mid-blocks. Many studies have examined factors contributing to the frequency and severity of vehicle-pedestrian crashes. While many of the studies have chosen to focus on crashes at intersections, few studies have focussed on vehicle-pedestrian crashes at mid-blocks. Since the factors contributing to vehicle crashes at intersections and mid-blocks are significantly different, more research needs to be done to develop a model for vehicle-pedestrian crashes at mid-blocks. In order to identify factors contributing to the severity of vehicle-pedestrian crashes, three models using different decision trees (DTs) were developed. To improve the accuracy, stability, and robustness of the DTs, bagging and boosting techniques were used in this chapter. The results of this study show that the boosting technique improves the accuracy of individual DT models by 46%. Moreover, the results of boosting DTs (BDTs) show that neighbourhood social characteristics are as important as traffic and infrastructure variables in influencing the severity of pedestrian crashes.
Top4.1. Introduction
Walking is the most basic and active mode of travel in transportation systems. In order to reduce air pollution and obtain better public health outcomes, efforts to encourage non-motorized transport modes have increased in recent years (Wey & Chiu 2013). To increase the number of walking trips, concerns about pedestrian safety must be addressed. Pedestrians are more likely to be harmed or killed in traffic crashes. They are 23 times more likely to be killed than vehicle occupants (Miranda-Moreno et al, 2011) and more than 22% of traffic deaths in the world are of pedestrians (WHO 2013). Every year, 34 pedestrians are killed in traffic crashes in the Melbourne metropolitan area, representing 24% of the total traffic fatalities. Mid-block crashes account for 46% of total pedestrian crashes in the Melbourne metropolitan area and 49% of pedestrian fatalities occur at mid-blocks (Crash Statistics Data 2016).
Many studies have been conducted to examine the factors contributing to the frequency and severity of vehicle-pedestrian crashes (Anderson et al, 1997; Zajac & Ivan, 2003; Kim et al, 2008; Tulu et al, 2015). Whereas many of the studies have chosen to focus on crashes at intersections (Lee & Abdel-Aty, 2005), only a few studies have chosen to focus on vehicle-pedestrian crashes at mid-blocks. Since the factors contributing to vehicle crashes at intersections and mid-blocks are significantly different (Al-Ghamdi, 2003; Bennt & Yiannakoulias, 2015), more research needs to be done to develop a model for vehicle-pedestrian crashes at mid-blocks. In terms of the methodologies used to analyse vehicle-pedestrian crashes, our review of the literature shows that different regression techniques, such as logit and probit models, are widely used. However, these statistical models require specific assumptions on the distribution of the random term and the relationship between the dependent and independent variables (Chang & Wang 2006). To circumvent these restrictions, decision trees (DTs) have been increasingly used in road safety studies (Lord et al, 2007). One disadvantage of this approach is that the results obtained in standard DTs may be changed significantly with changes in training and testing the data (Lord et al, 2007). To increase stability and robustness, ensemble methods, such as bagging and boosting, have recently been used in some traffic safety studies (Abdelwahab & Abdel-Aty, 2001, Zajac & Ivan 2003, Lefler & Gabler 2004, Chong et al. 2005). However, the relative performance of these methods has yet to be investigated.