Optimization Of Last-Mile Vehicle Route Problem By Consumer Logistics Profiling

In recent years,the scale of online shopping in China has continued to expand,and the demand for Last-Mile delivery has continued to increase.According to the "Statistical Bulletin on the Development of the Postal Industry in 2020" issued by the State Post Bureau,the national express delivery business volume in 2020 has reached 83.36 billion pieces,an increase of 31.2% over last year.The fast-growing express delivery business requires better Last-Mile delivery services to provide support.However,the current level of express delivery service cannot meet this demand.According to the statistics of the State Post Bureau in November 2021,the complaints of express service users increased by 62.9% year-on-year.Last-Mile delivery,as the link of direct contact and interaction between express service and customers,plays a vital role in improving the service level of express delivery service.Providing personalized terminal distribution service is an important means to improve customer satisfaction.According to a survey conducted by the China Institute of Smart Logistics,43% of consumers have demand for personalized services in Last-Mile delivery service.Then,(1)how to obtain and characterize the personalized logistics needs of consumers;on the basis of the characterization,(2)how to apply it to the Last-Mile delivery service strategy to ensure customer satisfaction while reducing enterprise distribution costs,Improve delivery efficiency.These are two very important issues faced by express delivery companies.In order to solve the above two problems,this thesis takes the e-commerce ordinary package Last-Mile delivery service of express companies as the study object.Based on the existing study,it first depicts the consumer logistics demand,and then introduces the vehicle routing optimization models based on the consumer logistics profile.The solution to this problem is: based on the precise description of the five logistics needs of consumers:delivery time,address,method,reassignment and complaint,quantify the preference probability of consumers on different characteristic values of different logistics needs,and on this basis Considering the trade-off between consumer satisfaction and company delivery cost,the company’s vehicle delivery route is planned.When characterizing consumer logistics needs,in order to solve the problem of cold start of new users,this thesis first describes the group profile of consumer logistics demand,and then studies the group-based precise profile of individual consumer logistics demand,and based on the characteristics of the two profiling results,respectively constructs the Last-mile delivery vehicle route optimization model based on consumer logistics demand group and individual profiles.While solving the Last-mile vehicle route optimization problem considering consumers’ individual logistics needs,it also provides inspirations for solutions of reducing the complexity of this type of VRP models.The innovations of this thesis mainly include the following four points:(1)It constructs a group profiling model of consumer logistics demand.This thesis proposes the consumer logistics demand profile for the first time.Based on the characteristics of the consumer logistics demand profile,it is proposed to comprehensively obtain consumers’ logistics characteristics data through the method of "questionnaire + registration information + delivery service database sampling".The method of logistics demand characteristic data solves the problems of difficulty in obtaining user dimension statistics and data privacy in the current user profile research.On this basis,the K-means++ clustering method,which is simple,effective,easy to understand and implement in the study of group user profiling,is used,and based on the ensemble classifier of the Bagging method,a group profile of consumer logistics demand is constructed.(2)It constructs a group-based profile of individual precise consumer logistics demand.This thesis proposes a group-based method for accurate individual profile of consumers’ logistics demand.In order to solve the problem of feature extraction of multidimensional,multi-scale and sparse data after the fusion of group profiling results and consumers’ historical delivery bill data,an attention mechanism is used to improve the current solution.A convolutional neural networks are used to extract the characteristics of individual profile of consumers’ logistics demand.On this basis,the group profile results are reconstructed by combining the characteristics of deep neural networks,and an accurate profile of consumers’ logistics needs based on group profile constructed.(3)A vehicle route optimization model for Last-mile delivery based on group profile is constructed.This thesis proposes a two-stage group profile-based vehicle routing optimization model for Last-mile delivery.Different from the previous set-division vehicle route optimization,in the first stage,this thesis takes the minimum expected total cost of each route as the objective function,and decides the service strategy of each consumer in the group while selecting the vehicle driving route,and while meeting constraints of capacity,overall preference satisfaction rate,and time windows of consumers and delivery sites,all feasible routes have been found;in the second stage,with the minimum expected total cost of the company,all customers in different groups are guaranteed to be allocated to different delivery vehicles.,and meeting the complaint rate of the entire route,the optimal distribution route plan was decided.(4)It constructs a vehicle route optimization model of Last-mile delivery based on individual profile,and designs a solution algorithm for it.The individual profile is a more accurate depiction of each consumer’s four logistics demand characteristic value preferences of time,location,method,reassignment and the characteristics of complaints of Last-mile delivery service.Based on the results of individual profiles,this thesis draws on the idea of constructing edge-flow model for vehicle routing optimization,and constructs a vehicle routing optimization model for Last-mile distribution based on individual profiles.This model is different from the previous vehicle routing optimization models,for the objective function,while deciding the driving route,it also decides the service strategy of each consumer’s different needs,and considering the complexity of the model,an algorithm is designed combing three heuristic algorithms of the Greedy algorithm,Adaptive Large Neighborhood Search and Tabu Search.There are 41 figures,44 tables,and 182 references.