This study applied tree-based machine learning methods to investigate the contributing factors to both crash frequency and injury severity in vehicle-pedestrian crash events. Vehicle and roadway characteristics, driver and pedestrian attributes, traffic controls and land use conditions, transit provision and weather conditions are used as covariates to predict pedestrian crash frequencies (by roadway segment) and injury severity levels (for pedestrians struck by vehicles on public roadways). In both cases, tree-based models offered significantly more prediction accuracy than traditional statistical models (using negative binomial and ordered probit specifications, with the same covariates). The tree-based models also offer valuable interpretability through the regression tree graph itself (with clear branching based on variable cut-points), variable importance plots (for each covariate), and partial dependence plots to help analysts understand the relationship between contributing factors and the target variable (count or severity). Average daily vehicle-miles travelled (DVMT) on each road segment, population density, segment length, census tract-level job density, distance from nearest K-12 school, transit stop density, and segment speed limits were estimated to be the top contributing factors for increasing pedestrian crash counts. DVMT has been found as the single most responsible factor for vehicle-pedestrian crashes and in a way representing pedestrian exposure to such situations. In terms of predicting injury outcomes, intoxication of the pedestrian and/or driver, higher speed limits at the site, crash location not being in the traffic way, older pedestrian, interstate highway locations, and dark and unlit conditions were predictors for more severe outcomes. Importantly, if the surrounding urban area’s population is reasonably high (over 25,000 persons), the probability of the pedestrian dying falls significantly, which supports the ‘safety in numbers’ idea, for more people available to help save the crash victims, or drivers going more slowly due to crowded conditions, closer hospitals, and so on. While few crash studies have included land use variables and local demographics, including proximity to schools, hospitals, and transit stops, population and jobs density variables appeared to add to crash counts and severity for pedestrians, though this is moderated by the 25,000-population threshold and distance variables

From the Federal Highway Administration to local departments of transportation, traffic safety is a persistent concern for transportation planners and engineers. Pedestrians are among the most vulnerable road users and require consideration beyond typical analysis of vehicle safety. This study has two objectives: to identify environmental, demographic, and behavioral factors explaining crash severity, and to compare methods for determining the significance of these factors. Binary and ordered logistic regression models were developed and compared to assess factor significance. Environmental and local factors, such as lighting and speed limit, had the strongest correlation with crash severity in all cases. However, inclusion of driver and pedestrian behavior and demographic characteristics improved the fit of the model and, in some cases, predictive ability. The two model types identified the same significant variables in traffic safety, but the magnitudes of the effects differed by model. This finding demonstrates that while the simpler method may yield the same overall results, combining methods can differentiate factors which contribute to the most severe crashes.

Binary logistic regression was shown to be an effective technique for modeling fault in pedestrian crashes. Results were used to classify fault, and identify additional factors that influence fault. Logistic models correctly classified fault in anywhere from 84% to 97% of cases as compared to the existing Florida Department of Transportation (FDOT) algorithm, which only classified fault correctly in 56% to 58% of the same cases. Improvements in classification accuracy were shown to stem from two sources. The first was the use of abundant data, which took advantage of data augmentation; a process by which additional fields of data were made available for investigation. The second source of improvements was from the use of more accurate data; data which had undergone quality enhancements to correct errors, and to fill in incomplete or missing information. Both the improvements in quality and quantity of data came from using additional data sources and manual case reviews.

The increasingly serious pedestrian safety issue in the City of Austin aroused the concern. Other than conducting quantitative analysis at aggregate level via collecting and examining the secondary data extracted from the existing datasets, the authors shifted towards the disaggregate level analysis, focusing on twenty-six hotspots of pedestrian collisions via mixed method research. Qualitative data was collected in the field survey to precisely capture the contextual features of collision locations, and was interpreted and coded as explanatory variables for the quantitative analysis. Instead of the frequency of pedestrian collision, crash rate measured by incident count per million pedestrians was the dependent variable to identify the factors truly influencing the pedestrian safety issue, not just the total number of walkers. The stepwise bivariate analysis and negative binomial regression examined the association between pedestrian collision rate and independent variables. Finally, the average block length, speed limit posted, sidewalk condition, and the degree of proximity to major pedestrian attractors were statistically significant factors correlating with the pedestrian collision risk.