Traffic congestion identification based on parallel SVM
Zhanquan Sun, Jinqiao Feng, Wei Liu, Xiaomin Zhu · 2012
Traffic congestion auto identification is a complicated problem. Many identification methods have been developed. SVM is taken as one of the most efficient traffic congestion identification methods. But the training computation cost of SVM is expensive. General SVM is difficult to be used in practical applications because that traffic congestion identification is a real-time task. Parallel SVM can improve the training speed markedly. It is possible to apply PSVM to practical applications. In this paper, PSVM is adopted to identify traffic congestion. Through example analysis, the training speed is improved without decreasing the traffic congestion identification precision. It illustrates that PSVM is suitable to be applied in practice.