Dynamic Takeout Order Dispatch Algorithm Driven By Data

With the rapid development of science and technology,the pace of human life is getting faster and faster,and the society has also spawned a number of new-type industries that adapt to the fast-paced,such as the take-away industry.As the take-away industry enters the normal track operation stage,under the premise of ensuring user satisfaction,improving the revenue of the take-away platform has become the primary goal of the platform.The problem of dynamic allocation of take-out orders is essentially a vehicle routing problem.A well-designed solution for vehicle routing problems can effectively improve the platform's profitability and optimize platform delivery efficiency.This paper analyzes and summarizes the dynamic allocation problem of orders,establishes the mathematical model of dynamic order vehicle routing problem with multi-vehicle time window and pickup and delievery constraints,and designs related solving algorithms.The research content mainly includes:Firstly,a descriptive analysis of the spatial and temporal distribution of take-out in Dalian is conducted.The investigation and analysis of the distribution of take-out space in Dalian has the characteristics of commercial circle;its time distribution has the characteristics of double peaks.Then we compare and analyze the existing order allocation model,summarize its characteristics,and conclude that in the current stage of the take-out industry,the dispatch mode is more suitable for the take-out platform to achieve higher returns.Secondly,the characteristics of the order allocation problem are compared and analyzed,which shows that the static order allocation problem has global transparency;while the dynamic order allocation problem has only partial transparency and strong change.Finally,the similarities and differences between the order allocation problem of the take-out industry and the order allocation problem of the e-commerce industry are compared,highlighting the characteristics of small batch size and high time efficiency;Through the comparison with the characteristics of the order allocation problem in the carpooling industry,it is characterized by its short path and large capacity.Based on the actual order allocation problem,the problem of the take-out order allocation in the[0,T]period is modeled.Firstly,the problem is mathematically described,the mathematical variables are defined,and then the goal of maximizing the platform's revenue,the time capacity of the knight's capacity,the time position of the order's starting position,and the end position are established on the premise of ensuring user satisfaction.Dynamic Vehicle Routing Model with Time Window and Picking Order Constraints for Multiple Vehicles with Constraints and Orders Delivered in Main ConstraintsThen,two heuristic algorithms,greedy algorithm and hybrid genetic algorithm,are considered for this problem.The idea of the greedy algorithm is to decompose the[0,T]period into several[k,k+l]segments,and then assign only the orders that make the platform get the highest return for the orders generated during the[k,k+1]period.iterative allocation until the allocation condition is not met.The process of greedy algorithm:first merge similar orders;then calculate the return matrix of the order in the order pool matching the knight,prioritize the high priority order;finally,assign the order with high yield until it can not be allocated.The idea of hybrid genetic algorithm is to decompose the[0,T]period into several[k,k+l]hour segments.For the orders generated in the[k,k+l]period,the genetic algorithm assigns each knight a specified number of orders each time.The neighborhood search algorithm then determines the order in which these orders are inserted into the Knight's existing dispatch list,looping iteratively until the end of the iteration.The process of hybrid genetic algorithm:firstly,the genetic algorithm is used to initialize the population,and the next generation of individuals with high fitness is selected by crossover,mutation and selection operators,and then the neighborhood search is performed in the next generation of individuals to determine the order to be assigned to insert the knight's existing order list.In the order,continue the iterative process of the genetic algorithm until the end of the iteration.Combining the delivery data of Dalian Roosevelt Business Circle in June 2016,we do two parts of the experiment.The first part of the experiment is based on the order data of one day in June 2016(the largest number of orders and the least number of orders each day)as the experimental object,and considering the benefit effect in the statistical sense from the perspective of one month,and considering the advantages and disadvantages of the two algorithms from the perspective of algorithm running time.The second part of the experiment is to use the greedy algorithm to analyze the impact of changes in the number of knights,time intervals,and estimated order service time on the platform's revenue,and find the best expected order service time and number of knights and time intervals in the Roosevelt business circle.