蚁群和粒子群算法优化支持向量机的公交行程时间预测

Prediction of Bus Travel Time Based on Support Vector Machines Using Ant Colony Optimization and Particle Swarm Optimization

为精确预测公交行程时间,提出了基于蚁群和粒子群(ACO+PSO)优化最小二乘支持向量机(LS-SVM)的预测方法。算法通过对LS-SVM的高斯核参数σ和正则化参数C进行搜索寻优,获得更为精确的LS-SVM,建立了ACO和PSO优化LS-SVM的公交行程时间预测模型,根据历史数据进行实例分析,并与LS-SSVM及BP神经网络进行比较。实验表明,ACO+PSO模型的预测精度高、鲁棒性及泛化能力强。

In order to accurately predict the travel time of the bus, a prediction method based on ant colony and particle swarm optimization(PSO) is proposed to optimize the least squares support vector machine(LS-SVM). Based on the Gaussian kernel parameters and regularization parameters of LS-SVM, the LS-SVM is obtained by searching the LS-SVM,and the bus travel time prediction model of ACO and PSO optimized LS-SVM is established. The model is analyzed according to the historical data compared with LS-SSVM and BP neural network. Experiments show that the ACO + PSO model has high prediction accuracy, robustness and generalization ability.