Yapılı çevre-trafik güvenliği ilişkisinde başlıca faktörlerin belirlenmesi için tabu-arama tabanlı bir yaklaşım

Traffic crashes are a significant contributing factor to fatalities globally, with an increasing pace of crashes occurring in urban areas. Thus, this study aims to analyze the relationship between urban built environment and traffic safety (BE-TS) to reveal the factors that threaten TS so that they could be avoided in early planning phases. BE was operationalized through the D-classification (e.g., density and diversity) whereas, TS was operationalized through varying crash categories (e.g., severity) and representation (i.e., frequency and density). A GIS-based process was applied to assign crashes and BE variables at the census block group level, followed by a dimension reduction process to eliminate correlated variables in an extensive set. Then the relationship between BE and TS was examined through a series of generalized linear regression models with several statistical distributions. A tabu-search-based model generation was applied to each BE-TS model to identify the set of BE variables that minimizes the modeling error. Moreover, variables were traced during the algorithm to identify consistent and then critical variables. Those models were analyzed for varying case studies, two cases within the developed country context, from the US, single county (Leon), and tri-county (Broward, Miami-Dade, and Palm-Beach) cases; and a case within developing country context, from Turkey (Konya). According to the results, exposure variables (e.g., bus stops and traffic volume) have positive associations with all crash types. However, there are variables with mixed relations (e.g., primary road percentage has positive relations with crash types except for vulnerable road user involved crashes, and employment density has positive or negative impacts according to the local economic structure). Finally, non-motorized modes are found to have positive, and motorized modes are found to have negative associations with all crash types within the developed country context, which are not common in safety research.