Research On Demand Prediction Of Online Car Hailing Based On Deep Bidirectional Long Short Term Memory Network

The number of demand for online car hailing is affected by a variety of internal and external factors,so the prediction is complicated and difficult.To solve this problem,a demand forecasting method based on Deep Bidirectional Long Short Term Memory(DBLSTM)neural network was proposed by using online car booking order data to represent spatial characteristics with interest point information and combining time and meteorological dimension factors.First of all,the article illustrates the problems on the online car hailing travel demand,has been clear about the area and time granularity partition method,based on the online car hailing order data and other related data,the effect factors of online car hailing travel demand is studied,using the correlation coefficient analysis the influence of changes in Point of Interest(POI)quantity,weather conditions,historical demand and other factors on the demand for online car hailing,Determine the input variables of the prediction model;Secondly,according to the time series distribution characteristics and periodicity characteristics of online car hailing travel demand,the trend of online car hailing demand changes on working days,non-working days and different dates of the same week were analyzed to further provide a basis for the selection of input variables and establish a DBLSTM-based online car hailing demand prediction model.Finally,the DBLSTM demand prediction model constructed in this paper is trained to find the optimal value of multiple hyperparameters and the optimal network structure,and the comparative test is carried out.Experimental results show that the prediction accuracy of multi-feature DBLSTM demand prediction model is improved by 2.84% compared with single-feature DBLSTM demand prediction model,and the maximum prediction accuracy is improved by 55.71%compared with other traditional models,which verifies the validity and accuracy of the proposed model.There are 21 figures,27 tables and 51 references in this paper.