Abstract: Parameter selection for SVM exerts influence on the predicting function of SVM prediction model in traffic flow． The present paper introduces PSO optimization mechanism to solve parameter selection of SVM． The way is to select SVM training parameters by PSO; then better PSO-SVM prediction model is obtained． Finally the paper compares PSO-SVM prediction model with neural network by means of simulation experiments and shows that PSO-SVM prediction model has advantages over neural network．

Key words: support vector machine( SVM) , particle swarm optimization( PSO) , neural network, prediction, traffic flow

摘要： 参数选择问题影响了支持向量机预测模型在交通流量中的预测性能。为了解决支持向量机预测模型的参数 选择问题，引入了粒子群优化算法机制，通过粒子群优化选择支持向量机预测模型的学习训练参数，得到较优的 PSO-SVM预测模型。通过实例仿真实验，将PSO-SVM预测模型与神经网络预测模型进行了比较，显示了其优越 性。

关键词: 支持向量机, 粒子群优化算法, 神经网络, 预测, 交通流量