基于改进LSSVM的节假日高速公路行程时间预测

Travel Time Prediction of Freeway Based on Improved LSSVM Model

节假日高速公路交通量突增,导致路段行程时间不确定增加,严重扰乱人们的出行安排,因此有效的行程时间预测至为关键。首先对历史数据集按照节假日行程时间的分布规律进行分类,使得子数据集和特征向量之间的关系,与预测时段行程时间和特征向量之间的关系更加相似。然后对LSSVM (Least Squares Support Vector Machines)模型进行改进,通过构造混合核函数,降低了模型计算复杂度;对PSO优化算法进行改进,解决了标准PSO算法搜索精度低,容易陷入局部极值的缺点。最后使用改进LSSVM模型对不同数据集进行训练,完成行程时间的预测。研究表明:(1)对历史数据集的分类,提高了模型预测的准确性;(2)与传统模型相比,改进后的模型训练速度更快,预测精度更高。

The increase in traffic volume during the holidays will lead to an increase in the uncertainty of travel time,which will seriously disrupt people's travel arrangements.Therefore,effective travel time prediction is critical.At first,the historical data set is classified according to the distribution rule of travel time on holiday,so that the relationship between subsets and feature vector is more similar to the relationship between travel time of the predicted period and feature vector.Then the LSSVM(Least Squares Support Vector Machines)model is improved by constructing the mixed kernel function,which will reduce the computational complexity of the model.In view of the shortcomings of the traditional PSO algorithm,which is low in search precision and easy to fall into the local optimal value,the PSO algorithm is improved.Finally,the improved LSSVM model is used to train different data sets and predict the travel time.The research shows that the classification of historical data sets improves the accuracy of model prediction;being compared with the traditional models,the improved model has faster training speed and higher prediction accuracy.