Applying a deep learning approach for transportation safety planning by using high-resolution transportation and land use data

Analytical Transportation Safety Planning (TSP) is an important concept for integrating and improving both planning and safety and achieving better policies and decision making. In the recent decade, considerable efforts have been devoted to providing better prediction results with the consideration of zonal systems, mathematical methods, input variables, etc. In previous studies, transportation and land use data have been widely used as input to predict crashes. Meanwhile, the previous studies required all input variables to be aggregated at the zonal level. With the aggregation process, the collected data fell into low resolution and lost details, which may introduce low accuracy and even biases. The primary objective of this study is to validate the viability of applying a deep learning approach to predict crashes for TSP with the high-resolution data. A framework of collecting high-resolution data is first introduced. Then, a deep learning architecture of a convolutional neural network (CNN) is adopted to predict traffic crashes. To validate the proposed method, an empirical study is conducted and the proposed method is compared with three counterparts: two statistical models (i.e., negative binomial model and spatial Poisson lognormal model) and a traditional machine learning model (i.e., artificial neural network) using low-resolution data (i.e., data that are aggregated based on zones). The results indicate that the proposed deep learning method with high-resolution data could provide significantly higher prediction accuracy than the three conventional models using low-resolution data, which validates the concept of using the deep learning approach with detailed data for traffic crash prediction. It is expected that the deep learning approach for traffic crash prediction in this study could provide new and valuable insights into the future directions of transportation safety planning.