基于SVM与自适应时空数据融合的短时交通流量预测模型

针对短时交通流变化周期性与随机性特点,选取时间和空间序列流量观测值作为支持向量机训练样本进行训练,使用空间序列预测值对交通流时间序列预测结果进行修正,并通过对历史时间空间序列预测结果的分析,动态调整其对未来预测的影响,建立基于SVM与自适应时空数据融合的短时交通流量预测模型.最后,将提出的预测模型与支持向量机时间序列预测模型、指数平滑法、多元回归法预测结果进行对比,结果表明:自适应时空数据融合预测模型可将预测平均相对误差控制在4%,明显高于其他模型预测精度.

Short-term Traffic Flow Forecasting Model Based on SVM and Adaptive Spatio-Temporal Data Fusion

Centering around the periodicity and randomness properties of short-term traffic flow,a better model was derived by correcting the results predicted by time series with the prediction data of space series based on SVM in this paper. In additions, the future projections were dynamically adjusted through the analysis of spatial and temporal historical predictions. The proposed model is compared with three well-known prediction models including SVR, Holt exponential smoothing, and Multiple regression. The resultant performance comparisons suggest that the adaptive spatio-temporal data fusion model performs better than other models,and the average relative error is less than 4%.