论文部分内容阅读
城市道路交通中交通检测器获得的数据往往不完整,存在丢失现象,需要对其进行修补,以保证交通流预测模型的实际应用精度。以离散和连续缺失的线圈检测器交通流量数据为研究对象,提出一种基于最小二乘支持向量机(Least Squares Support Vector Machines,LS-SVM)的交通流时间序列数据修补模型,并将其结果与利用RBF神经网络模型和一元非线性回归模型的结果进行比较。研究结果表明,LS-SVM模型修补精度优于RBF(Radical Basis Function)神经网络模型和非线性回归模型。最后,针对历史数据缺失难以构建LS-SVM模型的问题,提出了两阶段故障数据修补组合模型,取得了好的效果。
The data obtained by the traffic detector in urban road traffic is often incomplete and there is a loss phenomenon, which needs to be repaired to ensure the practical application accuracy of the traffic flow prediction model. Taking the discrete and continually missing coil detector traffic flow data as the research object, this paper proposes a traffic flow time series data repair model based on Least Squares Support Vector Machines (LS-SVM) With the use of RBF neural network model and the results of a unified nonlinear regression model for comparison. The results show that the LS-SVM model is better than RBF (Radical Basis Function) neural network model and nonlinear regression model. Finally, aiming at the problem that it is difficult to construct LS-SVM model with missing history data, a two-phase fault data repair combined model is proposed and good results are achieved.