融合粗糙集的GM(1,N)幂模型在短时交通流预测中的应用

Short-Term Traffic Flow Prediction Based on GM(1,N)Power Model Optimized by Rough Set Algorithm

实时准确的道路网短时交通流预测是智能交通系统的核心内容和难点问题.由于交通流的非线性、复杂性和不确定性,使得传统的交通流预测不能取得满意的效果.基于GM(1,N)幂模型,建立了一种道路网多段面短时交通流预测模型.考虑到上游路段与预测路段之间相关性,引入粗糙集理论确定模型的相关因素序列.针对传统参数求解方法最小二乘的不足,将问题转化为以拟合效果最优为目标的非线性优化模型求解,利用遗传算法(GA)搜索最优参数.通过实验例证,对比了GM(1,N)模型、GAGM(1,N)模型、BP神经网络和自适应神经模糊推理系统(ANFIS).结果表明,GAGM(1,N)幂模型对城市道路网多段面短时交通流的预测效果较好.

Real-time and accurate traffic flow forecasting of the road network is the core content and difficult problem of intelligent transportation system.Due to the nonlinearity,complexity and uncertainty of traffic flow,the traditional methods of traffic flow forecasting can not achieve satisfactory results.Based on the GM(1,N)power model,a short time traffic flow forecasting model for multi-sections in the road network is established in the paper.Concerning the correlation between the upstream and the section needed to be predicted,Rough set theory is introduced to determine the correlation factor sequence of the model.Against the deficiency of the traditional parameter solution method-the least square method,the problem is transformed into a nonlinear optimization model whose goal was to get the best fitting results,and the genetic algorithm is used to search the optimal parameters.The GAGM(1,N)power model was compared with the GM(1,N)model,GAGM(1,N)model,BP neural network and ANFIS through experimental analysis.The results show that the GAGM(1,N)power model has better results forecasting short time traffic flow for multi-sections in the road network.