Research On Online Car-hailing Supply-Demand Prediction Based On XGBoost Optimized Under The Cloud Computing

With the wave of "Internet+Travel",the intelligent travel platform has made great development,which attract more and more people to use the online car-hailing.However,using the service of platform still need to take a long time to wait for the car to arrive.To solve this problem,the research of the online-car supply-demand prediction come into being.By predicting,the online car-hailing can be dispatched in advance,so as to shorten the waiting time of users,maximize the revenue of the platform and promote the transport capacity of regions.This paper studies the relevant prediction algorithms and data mining technology,analyzes and constructs the characteristics that affect the online car-hailing supply-demand,designs the prediction model and predict the range of supply-demand gap in each area of the city in the next 10 minutes.The main research contents include:(1)Analyze the background and research status of the car-hailing supply-demand prediction.On this basis,study the relevant theory and the latest technology in the field of prediction algorithms and data mining,analyze the raw data and constructs the characteristics that affect the online car-hailing supply-demand.(2)Design and implement online car-hailing supply-demand prediction model based on XGBoost.Analyse of the principle of XGBoost and the implementation process of car-hailing supply-demand prediction model,set the input features,related parameters and prediction goal of the model,which lays a foundation for constructing a online-car supply-demand prediction model based on XGBoost.(3)Analyze and construct the point of interests(POIs)feature.Firstly,discuss the relationships between the POIs feature and the online car-hailing supply-demand prediction,then prove that different types and amount of POIs have different effects on predictions.On this basis,propose the method of extracting out the POIs feature.Finally,do experiments to prove the validity of the method of selecting POIs features and the positive effect of the selected POIs features on the prediction.(4)Use particle swarm optimization algorithm(PSO)to optimize the important parameters of the model.The value of important parameters in XGBoost has a great influence on the predictability of the model.Therefore,it is necessary to optimize the parameters.Firstly,analyze the principle and implementation process of the PSO.Then,analyze the role of parameters in XGBoost and select the important parameters to beoptimized.In the next,use PSO combined with prediction model to optimize the parameters.Finally,do experiments to verify the effect of optimization of parameters.(5)Use Spark platform to implement the parallelism of the model.Set up environment of a spark cluster,parallelize the model and compare the calculation time and operating efficiency of the parallelized model and the stand-alone model through experiments.The research shows that the online car-hailing supply-demand prediction model based on XGBoost proposed by this paper have good predictive effect and expansion ability.Use the feature engineering and optimization algorithm to improve the predictive ability of the model.Parallelize the model on the Spark platform also significantly improves the prediction efficiency of the model.The research in this paper provides a solution for the online car-hailing supply-demand prediction,which can be used as an effective prediction model in real life.