Improved Particle Swarm Optimized SVM for Short-term Traffic Flow Predication

Short-term traffic flow predication has played a key role in ITS. Since traffic flow has the property of periodicity and randomicity, a new short-tern traffic flow predication model based on support vector machine (SVM) is proposed, the parameter performance of SVM for regression estimation, the parameter of SVM is optimized by an improved particle swarm optimization (IPSO). The IPSO uses the dynamic best inertia weight and acceleration coefficient, which avoids the PSO plunging into local optima and make it converge faster. As the proposed model can reduce the dimensionality of data space and preserve features of traffic flow time series, it can predict traffic flow efficiently. The simulation results of traffic flow collected from Chinese national highway G107 show that the IPSO-SVM has greater efficiency and better performance than PSO-SVM. The average predication error is 3.3%, which proves the proposed model´s validity.