基于NCAGA-投影寻踪混合优化城市客运量预测

为了提高参数投影寻踪回归(parameter projection pursuit regression,PPPR)模型对城市客运量的预测精度, 基于cat映射、高斯分布和精英局部搜索对加速遗传算法进行改进. 提出了新的混沌加速遗传算法(new chaosaccelerating genetic algorithm, NCAGA),用于对PPPR模型的最佳投影方向α的优选.建立了在外层优化岭函数个数M的同时,内层利用NCAGA优化最佳投影方向a的NCAGA-PPPR混合优化城市客运量预测模型,结合某市统 计资料进行了仿真预测.结果表明该方法的预测精度优于BP神经网络模型、传统PPR模型和基于加速遗传优选的PPPR模型, 平均绝对相对误差小于3.1%,提高了城市客运量的预测精度,可有效应用于城市客运量的预测.

Urban passenger prediction based on hybrid algorithm of new chaos accelerating genetic algorithm and PPPR model

To improve urban passenger prediction accuracy with parametric projection pursuit regression, acceleration genetic algorithm was improved with cat map,Gaussian distribution and local searching.A new chaos accelerating genetic algorithm(NCAGA) was presented,used to optimize the best projection direction a of PPPR model.A hybrid algorithm of NCAGA-PPPR urban passenger forecasting model was proposed,in which the best projection direction was hybrid optimized inner by the NCAGA at the time of optimizing outer the number of ridge functions M.The simulation prediction was made with observed data,compared with BP neural network model,traditional PPR model and PPPR model optimized by acceleration genetic algorithm.The urban passenger forecasting accuracy is higher than the others,which the mean absolute relative error is less than 3.1%.The new hybrid algorithm can improve prediction accuracy of urban passenger and can be used efficaciously to forecast the urban passenger.