Geographically weighted random forests for macro-level crash frequency prediction

Machine learning models such as random forests (RF) have been widely applied in the field of road safety. RF is a prominent algorithm, overcoming the limitations of using a single decision tree such as overfitting and instability. However, the traditional RF is a global concept, and thus may fail to capture spatial variability. In macro-level analysis of road safety, the relationship between crash frequency and risk factors can vary spatially. To address this issue, we employ a modified RF algorithm, named geographically weighted random forest (GWRF). Based on the data from London at the level of Middle-super-output-area (MSOA), the predictive performances of RF and GWRF are compared using mean absolute error (MAE) and root mean square error (RMSE). Moreover, considering MSOAs are geographically connected with each other, several factors related to the discrepancies between adjacent zones are also included in the models. Our results indicate that GWRF outperforms the traditional RF and GWR when an appropriate bandwidth is selected. We further explore the effects of multicollinearity on model performance. The results show that prediction accuracy of GWRF models are not susceptible to the multicollinearity. However, the importance values of those variables with multicollinearity may reduce. Finally, and of equal importance, it is found that the importance of each explanatory variable varies across zones. The density of minor road makes the highest contribution to crash frequency in downtown area, while the crash frequency in peripheral area is more sensitive to the discrepancy of road environment between MSOAs. With such information, road safety interventions can be designed and implemented according to the locally important factors, avoiding thus general guidelines addressed for the entire city.

Predicting intersection crash frequency using connected vehicle data: A framework for geographical random forest

Accurate crash frequency prediction is critical for proactive safety management. The emerging connected vehicles technology provides us with a wealth of vehicular motion data, which enables a better connection between crash frequency and driving behaviors. However, appropriately dealing with the spatial dependence of crash frequency and multitudinous driving features has been a difficult but critical challenge in the prediction process. To this end, this study aims to investigate a new Artificial Intelligence technique called Geographical Random Forest (GRF) that can address spatial heterogeneity and retain all potential predictors. By harnessing more than 2.2 billion high-resolution connected vehicle Basic Safety Message (BSM) observations from the Safety Pilot Model Deployment in Ann Arbor, MI, 30 indicators of driving volatility are extracted, including speed, longitudinal and lateral acceleration, and yaw rate. The developed GRF was implemented to predict rear-end crash frequency at intersections. The results show that: 1) rear-end crashes are more likely to happen at intersections connecting minor roads compared to major roads; 2) a higher number of hard acceleration and deceleration events beyond two standard deviations in the longitudinal direction is a leading indicator of rear-end crashes; 3) the optimal GRF significantly outperforms Global Random Forest, with a 9% lower test error and a substantially better fit; and 4) geographical visualization of variable importance highlights the presence of spatial non-stationarity. The proposed framework can proactively identify at-risk intersections and alert drivers when leading indicators of driving volatility tend to worsen.

From the past to the future: Modeling the temporal instability of safety performance functions

In roadway safety management, safety performance functions (SPFs) are widely used by state and local agencies to predict crashes for base site conditions. SPFs are developed based on historical traffic safety data and are used to make predictions for anticipated site characteristics in the future. An underlying assumption in SPF development is that the relationships between crash frequency and site conditions are stationary from the past (when the model data were collected) to the future (for which SPFs are applied). The assumption using the past to represent the future could be fundamentally problematic. This study proposes a modeling framework that can relax this assumption. Specifically, this framework integrates temporal modeling with time-series analysis to strengthen the current SPF estimation methods. The temporal modeling approach is Temporally Weighted Negative Binomial Regression (TWNBW), and the time-series analysis is tried by employing the Seasonal Autoregressive Integrated Moving Average (SARIMA) and Artificial Neural Networks (ANN) methods. The temporal modeling is to uncover the temporal variations of SPFs and the time-series analysis explains and predicts the relationship between the SPF's temporal variation and time. The outcome of the framework is a set of Future SPFs that capture the temporal unobserved heterogeneity in safety data and describe the predicted relationships between safety performance and site characteristics in the future. A case study using six-year safety datasets from Georgia was conducted to illustrate the key components of the modeling framework. The temporal modeling results showed significant variations in SPFs across time. The parameters for traffic volume, i.e., Average Annual Daily Traffic (AADT), and segment length are associated with an increasing trend with time, and for access point density there is a descending trend. The SPF parameters are found to have a strong seasonality. Both time-series modeling methods appear to be appropriate to explain the temporal variations of SPF parameters, and the models are able to predict SPF parameters with acceptable errors smaller than 1% on average. Future SPFs can be used to support the roadway safety management that affects future traffic safety performance.

Safety analysis of urban signalized intersections under mixed traffic

INTRODUCTION

This study examined the crash causative factors of signalized intersections under mixed traffic using advanced statistical models.

METHOD

Hierarchical Poisson regression and logistic regression models were developed to predict the crash frequency and severity of signalized intersection approaches. The prediction models helped to develop general safety countermeasures for signalized intersections.

RESULTS

The study shows that exclusive left turn lanes and countdown timers are beneficial for improving the safety of signalized intersections. Safety is also influenced by the presence of a surveillance camera, green time, median width, traffic volume, and proportion of two wheelers in the traffic stream. The factors that influence the severity of crashes were also identified in this study.

PRACTICAL APPLICATION

As a practical application, the safe values of deviation of green time provided from design green time, with varying traffic volume, is presented in this study. This is a useful tool for setting the appropriate green time for a signalized intersection approach with variations in the traffic volume.