为了解决网约车行业中供需不平衡问题,首先总结用户出行的分布特征,然后通过加权迭代DBSCAN算法得到用户出行热点区域,在此基础上,提出一种改进图卷积神经网络(G C N)的网约车需求预测模型.通过网约车实际出行数据,证明该模型的可行性和有效性,其预测精度明显优于图卷积神经网络(GCN)、遗传算法(GA)、遗传算法支持向量机(GA-SVM)、反向传播神经网络(BP)、径向基函数神经网络(RBF)预测模型,该研究有助于优化网约车供需结构,对车辆调度具有重要价值.
Prediction of ride hailing demand based on improved DBSCAN
In order to solve the imbalance between supply and demand in the car-hailing industry,the distribution characteristics of user trips were first summarized,and then the hot areas of user trips were obtained by weighted iterative DBSCAN algorithm.On this basis,an improved graph convolutional neural network(GCN)demand prediction model for car-hailing was proposed.The feasibility and effectiveness of this model are proved by the actual online car-hailing data.The prediction accuracy of this model is obviously better than that of graph convolutional neural network(GCN),genetic algorithm(GA),genetic algorithm support vector machine(GA-SVM),backpropagation neural network(BP)and radial basis function neural network(RBF).This study is helpful to optimize the supply and demand structure of ride hailing,which is of great value to vehicle scheduling,and provides a theoretical basis for managers to implement management strategies more accurately.
NETL
NSTL
万方数据
