An Improved Particle Swarm Optimization for Prediction of Accident Severity

Abstract

In recent time, road traffic accidents are increased day by day due to incremental growth in number of vehicles. Traffic accidents are also the crucial and important issue for achieving the sustainable transportation development. One of the tasks of sustainable transportation is to reduce the number of accidents and also design the traffic assessment policy. Many researchers consider the traffic accident issue of sustainable transportation and developed prediction models for measuring severity of accidents. But the accuracy of accident severity is one of the major issues. In this work, an attempt is made to improve the accuracy of accident severity. To achieve the same, a particle swarm optimization-based algorithm is applied for evaluating the accident severity. Prior to implement the PSO, two modification are incorporated into PSO algorithm, called improved PSO. These modifications can be described as mutation operator and trail candidate generation. The performance of improved PSO is examined over accident traffic severity dataset and results are evaluated using accuracy, recall@5 and precision@5 metrics. Several existing techniques are considered for comparing the results of IPSO algorithm. It is revealed that IPSO achieves more accurate results among all techniques.