Logistics Distribution Time Prediction Based On SVR And Kalman Filter

Logistics is an important part of unblocking the national economic cycle,it has made tremendous contributions to promoting economic development,and it has gradually attracted people's attention.Logistics distribution is an important part of the logistics system and a customer-oriented service link.The quality of logistics distribution will directly affect the customer experience.Precisely predict the logistics contribution time,improve the punctuality of goods distribution,can save customers time,improve customer experience,help logistics companies improve distribution efficiency,reduce distribution costs,and enhance corporate competitiveness.Therefore,it is of great sighificance for the research of logistics distribution time prediction.This paper takes the precise prediction of urban logistics distribution time as the research goal.By analyzing the current research,status of logistics distribution problems,according to the influencing factors of its time prediction,the input variables of the prediction model are determined,and the logistics data collected by GPS are preprocessed to clear the dirty data in the data set.Taking into account the urban logi.stics distribution route is too long,the distribution route is divided and segment prediction is carried out;For the traditional logistics time prediction model that cannot correct the time error accumulation problem caused by accidents in the distribution process,a particle swarm optimization support vector regression and Kalman filter combined prediction model is proposed,use particle swarm optimization support vector regression model to predict the distribution time of logistics vehicles,and combine the real-time information of logistics vehicles and Kalman filter algorithm to correct the prediction time to avoid the accumulation of predict errors and achieve accurate prediction of logistics distribution time.After simulation verification,compared with the traditional support vector regression and BP neural network model,the prediction model proposed in this thesis reduces the average absolute percentage error by 8.92%.The results show that the joint prediction model has the ability to correct errors caused by accidents,and the prediction accuracy is higher,which provides reference value for logistics enterprises to grasp the real-time location of logistics vehicles.