Short Term Traffic Flow Prediction Based on Improved Support Vector Machine

Abstract

ABSTRACT The traditional forecasting methods are not suitable for short term traffic flow prediction, due to strong non-linear, time varying characteristics of urban transportation system. In order to improve forecasting accuracy of short term traffic flow, short term traffic flow prediction model based on support vector machine is presented. The most important parameter of support vector machine is parameter selection including the kernel function parameter and the penalty factor, which has significant influence on the properties of model prediction. Particle swarm optimization is used to optimize support vector machine, and particle swarm optimization is improved by means of adjusting inertia weight and choosing acceleration constant dynamically. Then improved particle swarm optimization is used to optimize support vector parameter. The experiment results show that predictive result based on improved particle swarm optimization LSSVM is closer to the real traffic flow data compared with support vector machine based on basic particle swarm optimization. Short term traffic prediction model based on improved particle swarm optimized support vector machine is feasible.