Enhancing Support Vector Machine Performance in Forecasting the Number of Vehicles Involved in Traffic Crashes via Metaheuristic Optimization Algorithms

Traffic accidents cause huge financial, human, and emotional losses every year. With increased urbanization and population growth, safety problems are also increasing. This study develops a novel approach by integrating Support Vector Machine (SVM) with advanced meta-heuristic algorithms to build a classification model of the number of vehicles involved in a traffic accident and the ranking of factors contributing to the accident, which is essential for improving transportation safety management. For this purpose, nine factors contributing to impacting road safe transport in urban areas of Cosenza, southern Italy, are assessed and ranked through the application of a stochastic approach. These features were determined using machine learning techniques, including SVM, combined with some metaheuristic optimization methods involving the Cuckoo Search Algorithm (CSA), Multi-verse Optimizer (MVO), and Whale Optimization Algorithm (WOA). With the SVM-WOA algorithm offering a great accuracy value, the results reveal that these approaches are successful in detecting elements influencing traffic safety management in road transportation.