Study and Application of Traffic Flow Forecast Based on PSO-WLSSVM

Abstract:

With wavelet function introduced to improve least square support vector machine kernel function and the wavelet least square support vector machine (WLSSVM) improved using particle swarm optimization (PSO), PSO-WLSSVM for traffic flow prediction is proposed. PSO-WLSSVM inherits good time and frequency domain distinguishing ability from wavelet transform, and the nonlinear learning performance from LSSVM; PSO is used to conduct global optimum search of super parameters so that the blindness of human selection could be avoid. Thus the accuracy of model predictions is improved. The simulation result shows that the forecasted traffic flow of PSO-WLSSVM are in good agreement with the measured value, and the forecasting precision of PSO-WLSSVM than the traditional LSSVM, thus indicating that the PSO-WLSSVM is feasible and precise and can be well applied to the forecast of traffic flow.

• Periodical: Advanced Materials Research (Volumes 779-780)
• Pages: 453-456
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.779-780.453
• Online since: September 2013
• Authors: Fei Xu, Yao Zhong Shi, Ke Wang
• Keywords: Least Squares Support Vector Machines (Ls-SVM), Particle Swarm Optimization (PSO), Simulation, Traffic Flow, Wavelet Transform (WT)

References

[1] B.Y. Sun, L.Y. Wei: CHINA JOURNAL OF HIGHWAY AND TRANSPORT. Vol. 6 (1993), pp.84-89 (In Chinese).

[2] J.C. Burge: Data Mining and Knowledge Discovery. Vol. 2 (1998), pp.121-167.

[3] Q. Zhang, A. Benveniste: IEEE Trans on Neural Networks. Vol. 3 (1992), pp.889-898.

[4] R.C. Eberhart, J. Kennedy, in: Proceedings of Sixth International Symposium on Micro Machine and Human Science, edited by Nagoya, Japan(1995), in press.

[5] H.B. Zhao, X.T. Feng: Chinese Journal of Geotechnical Engineering, Vol. 25 (2003), pp.468-471.