Statistical and Geostatistical Analysis of Accident Data for Developing Injuries Forecasting Models

During the last years the road safety has become increasingly important because of the number of dead and injured people and therefore the economically and socially costs associated. The most recent international studies about the number of accidents show that most of them happen in urban areas. In these accidents more than 50% of victims are vulnerable users (i.e. cyclists, motorcyclists and/or pedestrians). Clearly the dangerousness of urban areas accidents is much higher than that in rural areas. This is mainly due to the major vulnerability of vulnerable users compared to other road users. Analyzing the accident data in urban and rural areas trouhgt the Statistics and Geostatistics, our research focuses on the definition of the accidents causes (in relation to road characteristics, traffic flow conditions, etc..) and defining algorithms for predicting the risk of accidents (for which vulnerable users could die and/or be injured). Internationally, the most popular models regarding the accident forecast were built upon statistical techniques. The innovativeness of the proposed work arises precisely from the use of geostatistical techniques for the analysis of the accident data. In particular Geostatistics, created historically in mining field in order to solve problems concerning the correct evaluation of the ore bodies, supplies a collection of techniques addressed to the study of the correlation between experimental values of a specific variable (which represents the phenomenon in study) and for the definition of unknown values through interpolation which takes account of that law (i.e. kriging). The study is structured as follows: • data collection about accidents where vurnerable users are affected; • identification of “danger” road sections characterized by a high number of accidents; • study and analysis of speed and traffic flow nearly these “danger” sections; • definition and analysis of all the geometric and the design features for the roads affected by the accident; • construction of a database for storing the speeds recorded for each section (of vehicles and vulnerable users distinguished by category), flows of traffic (vehicles and vulnerable users, distinguishing them by category), road-related defects related to planimetric or cross-section; • preliminary and exploratory analysis of collected data through statistical techniques; • study and modeling of the correlation law between the accident data and the road characteristics through geostatistical techniques; • definition of forecasting geostatistical models; • forecasting models validation.