Online Car-hailing Users' Traveling Behavior Analysis And Demand Predicting

In recent years,with the acceleration of urbanization and the development of public transportation system,there are more and more ways for residents to travel: buses,cars,subways,car-hailing,bike-sharing and so on.Combining the information like satellite navigation signals obtained by the devices that equipped on these tools,and using the advanced technology of Internet of Things and Big Data,people can investigate the users' behavior,so as to solve the problem of traffic and urban planning,and promote the development of the city,improve the quality of people's life.Among the above various travelling tools,online car-hailing,as a new way of travelling,has accumulated a large number of users in recent years and been widely recognized by the society and supported by relevant national policies.It also solved the problem of information asymmetry between traditional taxi drivers and passengers.However,there is little analysis on the travelling behavior of ride-hailing users,and as more and more users and drivers join in,some passengers met the problems such as having difficulty in hailing a car and waiting for a quite long queue time at certain periods.How to make the use of the historical order data of car-hailing,how to analyze and study the travelling behavior of car-hailing users,and how to forecast the demand of car-hailing users in some regions have become pivotal research directions.Taking Haikou city as an example,this paper uses the historical data of DIDI service car-hailing orders and weather data to study and analyze the passengers' travelling behavior and forecast the demand of car-hailing orders.The main research contents are as follows:1.Data pre-processing – obtain online car-hailing order data;build a Web Crawler system to crawl the weather data within the orders' date range;carry out data cleaning,format conversion,data integration and other work on the above data,and obtain the pre-processed order data.2.Data analysis – conduct statistical and visual analysis of pre-processed ridehailing order data,and analyze the quantity and variation trend of ride-hailing users' demand in different time,regions and environments.Do the basic work for the prediction of the demand of the online car-hailing service.3.Model construction of demand prediction – prediction of car-hailing users' travel demand based on the visual results obtained from the completed visual analysis.Firstly,the feature selection is carried out to explore the influence of different variables on the proportion of demand.Then,the LSTM and CNN parallel combination model is proposed to realize the prediction of ride-hailing users' travel demand,and the preprocessed order data is used to determine the optimal parameters and the overall structure of the model.A comparative experiment is carried out,and the experimental results prove that the accuracy of the demand prediction model constructed in this paper is better than that of the traditional single-feature LSTM prediction model and the serial multi-feature LSTM prediction model.The research work of this paper has a strong practical significance for the operation of ride-hailing companies and the planning of urban public transport system.