Research On The Forecasting Of Taxi Demand Based On Machine Learning Methods

With the rapid development of the national economy and the continuous advancement of urbanization,urban road traffic has brought many conveniences to residents’ daily life.But at the same time,the continuous growth of population,the increasing number of vehicles and the consequent dramatic growth of residents’ daily traffic demand make traffic congestion more and more frequent.Intelligent transportation system(ITS)is a more effective way to solve urban road traffic problems,and traffic prediction is an important part of ITS research.In the urban public transportation system,taxis and Internet-based taxis are important components of them.However,due to the volatility and randomness of passenger travel demand,taxis are blindly searching for passengers by cruising.In some periods of time,the increasing number of Internet-based taxi users leads to the shortage of Internet-based taxis in some areas,while in other areas,there is excess transport capacity.Transport capacity scheduling through supply and demand predicting is the most effective way to solve the contradiction between passenger travel demand and vehicle service.Supply and demand forecasting belongs to the field of short-term traffic flow predicting.By predicting the passengers’ demand for taxis in a certain period of time and a certain area,the excess transport capacity is directionally dispatched to the surrounding specific area,thus effectively solving the contradiction between supply and demand.By consulting a large number of relevant literature,this thesis reviews the relevant literature at home and abroad from two aspects of taxi demand predicting and short-term traffic flow predicting,and studies the prediction of taxi demand.Firstly,the predicting model of taxi demand based on XGBoost is established,and the advantages of XGBoost algorithm compared with the traditional gradient boosting algorithm are analyzed theoretically.By preprocessing the sample data,the taxi demand is analyzed,and the model features are constructed.The experimental results show that the predicting model based on XGBoost is more effective in predicting the taxi demand and is superior to other predicting models.Secondly,a single variable linear regression model is established to extract the monthly trend of taxi demand,and a spline curve model is established to extract the daily trend of taxi demand.The features of the taxi demand predicting model based on XGBoost are optimized,the monthly trend features and the daily trend features are added.The experimental results show that the effectiveness of the new features to improve the prediction accuracy of the model.Finally,based on the research done in this thesis,the design and implementation of the taxi demand predicting system is carried out.The function of the system is composed of three modules: real-time demand,historical demand and system management.The implementation of the system verifies the effectiveness and practicability of the research done in this thesis.