A DEA-based evaluation method for joint distribution network location

Abstract

The present invention relates to a method for evaluating the site selection of a common delivery network based on DEA, the method comprising the following steps: S1: using an unsupervised learning particle swarm algorithm to solve a clustering algorithm to obtain a reasonable express delivery station location placed in a city; S2: clustering the results to obtain the optimal express delivery station location, and then using an immune algorithm to obtain a reasonable common delivery network location, constructing a common delivery network efficiency index system based on DEA, and obtaining relevant express delivery network data; S3: using the obtained efficiency index to increase the site selection conditions of the common delivery network, so that the common delivery network site selection problem can achieve the goal of achieving the highest operating efficiency under the condition of the lowest economic cost. The present invention provides an effective method for solving the problems of express delivery route overlap and high costs in real scenarios, and is conducive to building an intelligent, collaborative, low-cost, and efficient overall urban express delivery network.

Description

Technical Field

[0001] The present invention relates to the field of joint distribution site selection evaluation, and in particular to a joint distribution network site selection evaluation method based on a DEA model. Background Art

[0002] Joint distribution aims to reduce the resource waste caused by individual express delivery companies through combined transportation. With the rapid development of my country's express delivery industry, the need for joint distribution is becoming increasingly urgent. The decentralized operations of individual express delivery companies lead to dispersed demand, which in turn leads to duplicated operations and wasted resources. Joint distribution is a key solution to this problem. Furthermore, the overall growth of the express delivery market has slowed in recent years, and the overall industry development is shifting from increasing incremental growth to competing for existing market share. Express delivery companies are also facing the challenges of market-driven integration and upgrading. The rational planning of a joint distribution network can effectively reduce the cost per express delivery and improve express delivery efficiency. Using a DEA-based joint distribution network location selection method, the selection of joint distribution network locations not only considers external space optimization but also considers the impact of internal enterprise management on location selection. Calculating the efficiency of different distribution centers before achieving joint distribution can address real-world problems such as overlapping express delivery routes and high costs. Summary of the Invention

[0003] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0004] The present invention plans the express network structure according to the whole chain, and completes the process planning between all the express nodes passed through. The goal is to achieve the triple goals of lowest total cost, shortest total time and highest total efficiency under the conditions of not exceeding the vehicle capacity and time window constraints. A three-stage algorithm solution method is adopted, and the solution is reversed layer by layer to finally obtain the express network structure with the highest superiority. The first stage solves the site selection problem of express stations, and uses the superiority of unsupervised learning algorithms to obtain the optimal number of station clusters. The second stage solves the site selection and distribution path optimization problems of co-distribution outlets, and uses the immune algorithm to determine the location of co-distribution outlets at known express station locations to achieve the lowest total route cost. The third stage uses the DEA evaluation method to calculate and sort the efficiency of express outlets, screen the locations of co-distribution outlets, and determine the optimal location. The present invention specifically adopts the following technical solutions:

[0005] A DEA-based method for evaluating the location of common distribution points includes the following steps:

[0006] (1) Model construction

[0007] It is known that vehicles can travel on any road between nodes i and j within the time range not subject to traffic control. The transportation distance at this time is set to Within the time range of traffic control, the vehicle needs to detour and reach the target distribution point via the shortest distance. The transportation distance at this time is set to Right now:

[0008]

[0009] Among them, p ij is the distance between the different q routes for transportation between the common distribution points ij, T ki is the start delivery time of the vehicle to the shared distribution point i, [μ1,μ2] is the time range of traffic control;

[0010] The objective function is expressed as:

[0011]

[0012] Among them, f2 is the fixed cost of mobilizing vehicles, K is the set of vehicles, and N is the set of common distribution outlets. is the average speed of the vehicle, λ is the average distance between adjacent traffic lights, φ is the expected travel time of each traffic light, c2 is the transportation cost per unit distance of the vehicle, X ijk2 is the decision variable for whether the vehicle arrives at point j from point i, R2 n is the actual load of the vehicle, W2 is the maximum load of the vehicle, Y k2 A decision variable indicating whether a vehicle will incur a call cost;

[0013] The constraints are expressed as:

[0014] Ensure that the total weight of the express delivered by the transport vehicle does not exceed the maximum loading capacity of the delivery vehicle

[0015]

[0016] Ensure that the transportation of vehicles is carried out within the working hours of the common distribution center

[0017]

[0018] in, The working hours of the joint distribution center;

[0019] Normal driving distance of a vehicle without traffic control

[0020] P ij =MinP ij

[0021] Minimum driving distance of vehicles under time constraints of traffic control

[0022] P ij=Min(P ij -minP ij )

[0023] The actual demand of the distribution point is the same as the actual transportation volume of the vehicle

[0024]

[0025] Among them, r n is the demand for outlets,

[0026] Each vehicle can only transport to one common distribution point

[0027]

[0028] Whether vehicle k2 is transported from distribution point i to distribution point j

[0029]

[0030] Does vehicle k2 incur a call cost?

[0031]

[0032] (2) Determine the location of the optimal post station and distribution network

[0033] Based on the model in step (1), the unsupervised learning particle swarm optimization algorithm is used to cluster the applicability of the clustering data vector, wherein K-means clustering is used to generate the initial population and obtain the reasonable express station location within the city; the optimal express station location is obtained according to the clustering results, and the reasonable co-distribution network point location is obtained through the immune algorithm;

[0034] (3) Using DEA to evaluate the efficiency of co-distribution points:

[0035] The efficiency of the shared distribution network is calculated according to the entropy weight method. The weights of the evaluation indicators are determined by data standardization, information entropy of each indicator, and information entropy calculation. After the weights are determined, the operational efficiency of the shared distribution network is calculated by the data envelopment analysis method based on the relative efficiency of similar decision-making units of multiple performance measurement indicators:

[0036] Assume that the weight of the output indicator is u n , the weight of the input indicator is v m , a total of y j Output indicators, x i Input indicators. For each part, the efficiency evaluation index is:

[0037]

[0038] Based on the calculated operating efficiency of each co-distribution point, screening is performed to obtain the optimal co-distribution point location.

[0039] Preferably, the output indicators include the efficiency of warehouse inbound and outbound work, express delivery processing efficiency, and service quality; the input indicators include the total area of the outlets, the number of outlet staff, the number of outlet equipment, and the total cost of the outlets;

[0040] The efficiency of warehouse entry and exit = (warehouse entry time - receipt time / average port entry time) + (shipping time - warehouse exit time / average port exit time);

[0041] Express delivery processing efficiency = delivery volume / number of couriers;

[0042] Service quality is the weighted average of the false receipt complaint rate, loss and damage rate, upgraded complaint acceptance rate, and secondary complaint rate.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] (1) There are too many existing express delivery stations and their coverage areas overlap, resulting in a waste of resources. The present invention uses an unsupervised learning particle swarm algorithm to solve a clustering algorithm to obtain the reasonable locations of express delivery stations within the city.

[0045] (2) The optimal location of the express delivery station is obtained through clustering results, and then the reasonable location of the shared distribution network is obtained through the immune algorithm. In addition to considering the impact of external space optimization on the location selection of the shared distribution network, the impact of internal management on the location selection is also added. The efficiency of each shared distribution network is calculated by combining the entropy weight method and the DEA algorithm, and the location conditions of the shared distribution network are increased, so that the shared distribution network location problem can achieve the goal of the highest operating efficiency under the condition of the lowest economic cost.

[0046] (3) Based on the existing traffic control, the traffic control of road sections that are common in daily life and have a greater impact on urban transportation is analyzed and studied at the express transportation level, and a path planning model for urban trunk transportation vehicles is constructed under this traffic control situation, which is conducive to building an intelligent, collaborative, low-cost and efficient overall urban express network. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a diagram of the joint distribution sharing optimization process.

[0048] Figure 2 It is a diagram of common delivery express transport mode.

[0049] Figure 3 It is a diagram of transportation route changes when there is traffic control.

[0050] Figure 4 This is a map of the distribution network in a certain city.

[0051] Figure 5 Cluster map of post stations in a certain city.

[0052] Figure 6 Co-distribution network site selection and optimal route map.

[0053] Figure 7 Transport route map between distribution points. DETAILED DESCRIPTION

[0054] Before sharing was realized, each express delivery company adopted a separate delivery method, with its own delivery distribution center, receiving distribution center, outlets and post stations in cities A and B, and trunk and branch transport vehicles with different load capacities. Figure 2 As shown in the figure, after sharing is implemented, a distribution center in delivery city A, a distribution center in receiving city B, a shared distribution network, and express delivery stations are set up separately. Vehicles depart from the distribution center in delivery city A, travel via trunk lines to the distribution center in receiving city B, and then arrive at the shared distribution network for delivery to express delivery stations in various regions, completing the entire express delivery process.

[0055] (1) Model construction

[0056] Figure 3 The simulation of transportation route changes under traffic control is conducted. Outside the traffic-constrained time period [μ1, μ2], trucks can deliver between distribution center A and common distribution points N = {1, 2, ..., n} via route A. Within the traffic-constrained time period [μ1, μ2], vehicles must detour via route B for transportation.

[0057] A: Distribution Center

[0058] N = {0, 1, 2, ..., n}: the set of common distribution points, where 0 represents the distribution center of the receiving city from which the goods are delivered.

[0059] H={0,1,2...,h}:the common distribution network point within the road section jurisdiction area

[0060] K2={1,2,...,k 2m}: A collection of vehicles

[0061] The distances [μ1,μ2] between the q different routes for transport between the common distribution points ij: the time range of traffic control

[0062] Working hours of the joint distribution center

[0063] W2: Maximum load capacity of the vehicle

[0064] Actual load of the vehicle

[0065] r n : Demand for outlets

[0066] f2: Fixed cost of vehicle operation

[0067] c2: the transportation cost per unit distance of the vehicle;

[0068] φ: represents the expected travel time of each traffic light

[0069] γ: represents the average distance between adjacent traffic lights

[0070] Average vehicle speed

[0071] T ki : The time when the vehicle starts delivering to the shared distribution point i

[0072] The decision variable represents whether the vehicle arrives at the node j from the node i.

[0073] The decision variable indicates whether the vehicle will incur a call cost

[0074] It is known that vehicles can travel on any road between nodes i and j within the time range not subject to traffic control. Then the transportation distance at this time is set to On the contrary, when the time range of traffic control is within the scope, the vehicle needs to detour and reach the target distribution point through the shortest distance. The transportation distance at this time is set to Right now:

[0075]

[0076] The objective function can be expressed as:

[0077]

[0078] The objective function consists of three parts: the fixed vehicle mobilization cost, the vehicle transportation cost, and the vehicle deployment cost. The deployment cost is determined by calculating the vehicle's full load factor.

[0079] The constraints can be expressed as:

[0080] Ensure that the total weight of the express delivered by the transport vehicle does not exceed the maximum loading capacity of the delivery vehicle

[0081]

[0082] Ensure that the transportation of vehicles is carried out within the working hours of the common distribution center

[0083]

[0084] Normal driving distance of a vehicle without traffic control

[0085] P ij =MinP j

[0086] Minimum driving distance of vehicles under time constraints of traffic control

[0087] P ij =Min(P ij -minP ij )

[0088] The actual demand of the distribution point is the same as the actual transportation volume of the vehicle

[0089]

[0090] Each vehicle can only transport to one common distribution point

[0091]

[0092] Whether vehicle k2 is transported from distribution point i to distribution point j

[0093] X ijk2 =0

[0094] X ijk2 =1

[0095] Does vehicle k2 incur a call cost?

[0096] Y k2 =1

[0097] Y k2 =0

[0098] (2) Algorithm implementation

[0099] (1) Using unsupervised clustering algorithm to select the optimal station location

[0100] Currently, most residential and office buildings in China use door-to-door delivery services. This service improves delivery safety and enhances the customer experience, but due to time constraints, it can lead to duplicate deliveries and waste resources when couriers from different courier companies repeatedly make deliveries to the same customer. Based on actual conditions, it has been found that express delivery stations, which can accommodate all express delivery needs and have broad and even coverage across all areas of a city, can meet customer needs most efficiently. Therefore, a cluster analysis was conducted on areas using door-to-door delivery services. Clustering nearby end-customer locations can effectively reduce vehicle delivery costs.

[0101] This paper exploits the applicability of particle swarm optimization (PSO) to cluster data vectors and develops a clustering algorithm that combines PSO with k-means, using K-means clustering to generate the initial population. K-means converges faster than PSO, but typically produces lower clustering accuracy. Using K-means to generate the initial population further improves the performance of the PSO clustering algorithm. Compared to using either traditional algorithm alone, it significantly reduces the dependency on the initial solution and achieves better convergence, thus reducing errors.

[0102]

[0103]

[0104] (2) Using efficiency evaluation methods and immune algorithms to determine the location of express delivery points

[0105] Aiming at the fact that outlets of different express delivery companies are repeatedly established in the same area, based on the known clustering results and service time of the previous level of stations, the time window constraint is added and the idea of TSPTW problem is used to obtain the construction quantity and location of common distribution outlets and the related service scope.

[0106] Given the coordinates of the common distribution center and outlets, and with the traffic path jurisdiction constraints, the vehicle path planning problem is solved by using the idea of VRPTW problem according to the rated load constraint and time window constraint of the first-level vehicle, with the goal of minimizing time and economic cost, and the optimal path from city B to each common distribution outlet is obtained.

[0107] After reasonably dividing the service area and determining the location of the post stations, the immune algorithm is used to determine the location of the distribution network points. The immune algorithm is an optimization form of the genetic algorithm. The immune algorithm can compensate for the immature convergence problem that traditional genetic algorithms are prone to when the initial population distribution is uneven.

[0108]

[0109]

[0110] In addition to the basic indicator of express delivery outlets' geographic location, the impact of efficiency on the location of shared distribution outlets also needs to be considered when conducting outlet cluster analysis. Terminal outlets possess express delivery resources such as vehicles, personnel, and equipment, and also undertake the functions of receiving and delivering parcels. Therefore, the impact of various express delivery resources on outlet efficiency evaluation cannot be ignored. The following table lists the indicators that affect express delivery outlet efficiency:

[0111] Table 1 Network efficiency evaluation index table

[0112]

[0113]

[0114] (3) Using DEA to evaluate the efficiency of outlets:

[0115] For a particular indicator, entropy can be used to determine its degree of dispersion. The smaller the information entropy value, the greater the indicator's degree of dispersion, and the greater its influence (i.e., weight) on the comprehensive evaluation. Therefore, information entropy can be used as a tool to calculate the weights of various indicators, providing a basis for comprehensive multi-indicator evaluation. The entropy weight method is used to calculate express delivery point efficiency. By standardizing data, calculating the information entropy of each indicator, and applying the information entropy calculation formula, the weights of each evaluation indicator are determined.

[0116]

[0117]

[0118]

[0119]

[0120] Assume that there are k indicators X1, X2, ..., Xk, where Xi = {x1, x2, ..., x n}, formula (1) uses data standardization to convert the data into Y1, Y2, ..., Y k, According to the definition of information entropy in information theory, formula (2) (3) can obtain the information entropy of each indicator E1, E2,..., E k ,Finally, formula (4) calculates the weight of the evaluation index through ,each information entropy.

[0121] After determining the weights of the evaluation indicators using the entropy weight method, the operational efficiency of the express delivery outlets was calculated by comparing the relative efficiency of similar decision-making units across multiple performance measurement indicators using the Data Envelopment Analysis (DEA) method. The following table lists the key input-output indicators:

[0122] Table 2 Key indicators of input and output

[0123]

[0124]

[0125] Assume that the weight of the output indicator is u n , the weight of the input indicator is v m , a total of y j Output indicators, x i Input indicators. For each part, the efficiency evaluation index is:

[0126]

[0127] The CCR model is constructed with n departments, called n decision-making units. Each decision-making unit has p inputs and q outputs, each represented by a different economic indicator. Thus, the multi-indicator input and multi-indicator output evaluation system composed of n decision-making units has a corresponding efficiency evaluation index for each decision-making unit DMUj:

[0128]

[0129] By taking appropriate weight coefficients v and u,

[0130] hj≤1,j=1,…,n

[0131] If the efficiency index of the j0th decision-making unit is taken as the target and the efficiency index of all decision-making units as the constraint, the following CCR (C2R) model is constructed:

[0132]

[0133]

[0134] u≥0,v≥0

[0135] X ij : The total amount of input of the jth decision-making unit to the i-th type of input. ij >0;

[0136] y rj : The total output of the jth decision-making unit for the rth type of output. rj >0;

[0137] v i : A measure of the i-th type of input, weight coefficient;

[0138] u r : A measure of the r-th type of output, weight coefficient;

[0139] i:1,2,…,m;

[0140] r:1,2,…,s;

[0141] j:1,2,…,n;

[0142] h j :efficiency evaluation index;

[0143] The above planning model is a fractional planning, using Charnes-Cooper transformation, let:

[0144]

[0145] Depend on

[0146] It can be transformed into the following linear programming model P:

[0147]

[0148] stw T x j -μ T y j ≥0,j=1,2,...n

[0149] w T x0=1

[0150] w≥0,μ≥0

[0151] The effectiveness of decision-making unit j0 is defined using the optimal solution of linear programming. From the model, it can be seen that the effectiveness of decision-making unit j0 is relative to all other decision-making units. The CCR model can be expressed using linear programming, and an important effective theory of linear programming is the duality theory. By establishing a dual model, it is easier to conduct in-depth analysis from a theoretical and economic perspective. Dual programming (D):

[0152] minθ

[0153]

[0154]

[0155] λ j ≥0,j=1,2,...n

[0156] θ unconstrained

[0157] Where λj is the linear coefficient of DMU, θ * is the efficiency value, 0<θ * <1.

[0158] For the convenience of discussion and calculation, we further introduce slack variables s+ and residual variables s-, and transform the above inequality constraints into equality constraints, which can be transformed into:

[0159] minθ

[0160]

[0161]

[0162] λ j ≥0,j=1,2,...n

[0163] θ is unconstrained, s + ≥0,s- ≤0

[0164] Among them, s + : slack variable, s - : Remaining variable.

[0165] The above plan (D) is directly defined as the dual plan of plan (P).

[0166] The DEA method requires no additional data processing, requiring only input and output data. It eliminates the need to consider the relationships between individual data expressions, eliminating the need for functional form assumptions and thus avoiding errors introduced by functional form in effective statistical analysis. The efficiency values of individual nodes can be determined by combining the entropy weight method with the DEA algorithm. The immune algorithm is used to determine the coordinates of potential nodes for co-location. These results are then filtered using the node efficiencies derived from the entropy weight method and the DEA algorithm to determine the optimal co-location node locations.

[0167] (4) Using GA algorithm to solve vehicle path planning problems with time windows

[0168] After determining the locations of express delivery stations using a PSO-based clustering algorithm and the immune algorithm and efficiency evaluation method, the routes for vehicles transporting between the shared distribution points and the shared distribution center are planned. When planning routes, it is necessary to consider the time window constraints of each shared distribution point and the traffic control constraints within the city. A genetic algorithm is used to solve the vehicle route planning problem with time windows, with the minimum path cost as the objective function:

[0169]

[0170]

[0171] (3) Case analysis

[0172] The total area of a certain city is 16,410.54 square kilometers, with 16 districts and a permanent population of about 22 million. According to the 2020 Statistical Bulletin on the Development of the Postal Industry in a Certain City, the cumulative business revenue of the city's postal enterprises and express service enterprises was about 40 billion yuan, and the total business volume was about 48 billion yuan. The development of the express industry in a certain city continued to move forward steadily. The present invention organizes the operation data of the express distribution centers and outlets in a certain city for analysis and puts forward reasonable development suggestions. Through simple organization, 150 post stations in a certain city were obtained, such as Figure 4 shown.

[0173] The method of combining particle swarm optimization algorithm and K-means clustering algorithm to solve the problem of express station clustering can improve the performance of particle swarm optimization algorithm and reduce errors. Using the real data of 150 express stations into the PSO clustering model, we can get the following Figure 5The cluster diagram shown.

[0174] After clustering the express delivery stations, we obtained the 30 best express delivery stations. We input this data into the second-stage site selection and route planning model, solved the model with the immune algorithm, and obtained the 15 best common distribution network locations with the lowest delivery cost as the goal. We restored the network locations to a city map, and we can get Figure 6 .

[0175] The efficiency evaluation indexes of 30 clustered outlets were calculated using the entropy weight method. The present invention evaluates the efficiency of express outlets from three different dimensions: express processing speed, labor productivity, and service quality. The following table can be obtained by collecting and processing basic data:

[0176]

[0177] The table shows that the combined score for these 30 outlets is 72.418, with a maximum of 122.003 and a minimum of 24.075. The significant difference between the maximum score and the outlet's combined score indicates that operating efficiency varies widely across outlets. Express delivery processing speed has the highest weight, followed by inbound and outbound efficiency and service quality, but the differences between the three are relatively small, indicating that the impact of the three evaluation dimensions on express delivery outlet efficiency decreases with each dimension, but remains relatively consistent.

[0178] After obtaining the weights of each output indicator, we bring them into the DEA model for solution, and we can get the efficiency values of the 30 outlets in the following table:

[0179] Table 3 Network evaluation indicators and efficiency values

[0180]

[0181] By analyzing the above evaluation indicators, it can be found that the maximum value of the comprehensive efficiency is 1, the minimum value is 0.613, and the average value is 0.88. The efficiency values vary greatly. By re-screening the 15 clustered points and removing the three points with a comprehensive efficiency lower than 0.7, 12 common distribution points with high efficiency and optimal location can be obtained. These 12 common distribution points are evenly distributed in various areas of a city in China. Then, the coordinates of the optimal common distribution points are input into the vehicle path planning model with a time window, and the following can be obtained: Figure 7 The running results.

[0182] Get the location of the distribution center and the corresponding transportation route:

[0183] Route1:0->3->7->10->11->6->4->2->0; Route2:0->5->9->1->0; Route3:0->8->0;

[0184] A basic cost calculation reveals that branch costs include delivery costs, site costs, staff costs, and facility costs. Delivery costs include the costs incurred at every stage of the transportation process. For example, for a shipment from Shanghai to Beijing, the shipping label fee is 2 yuan / shipping (including a 1.5 yuan delivery fee), transit fees are 2 yuan / kg for air freight and 0.5 yuan / kg for truck transport, a support fee of 0.5 yuan / shipping, and an operating fee of 0.5 yuan / shipping. Therefore, the delivery cost per shipment is 5.5 yuan. Site costs primarily cover branch rental costs. In one urban area, the daily rent per square meter for an express delivery branch is approximately 4 yuan. In addition to sorting operations, express delivery company sites also serve as offices and accommodation. Staff costs primarily cover the monthly salary of couriers. In Beijing, a courier's base salary is currently around 3,500 yuan, which, including commissions, totals approximately 5,800 yuan per month. Furthermore, most courier companies provide food and accommodation, averaging 200 yuan per person per month, for a total monthly salary of 6,000 yuan per person. The cost of facilities comes from the scanning equipment used by express delivery stations, with each PDA costing approximately 2,000 yuan. The fixed cost of the vehicles transporting goods from the distribution center to the various distribution points is 1,000 yuan per vehicle, with a variable cost of 10 yuan per kilometer. The fixed cost of the first-level vehicles transporting goods from the distribution points to the express delivery stations is 800 yuan per vehicle, with a variable cost of 8 yuan per kilometer.

[0185] Given the known locations of each shared distribution point, using first-level delivery vehicles to deliver express packages from these points to each shared express delivery station incurs another portion of costs. By discarding the three shared distribution points with lower operating efficiency, the remaining 11 shared distribution points are matched with the remaining 18 shared distribution stations and route optimization is performed. The shortest total delivery route using 10 first-level vehicles is 68.25 kilometers, with a total cost of 99.36241 million yuan. The following is a comparative analysis of the situation before and after implementing shared distribution:

[0186] Individual delivery Joint delivery Increase or decrease Increase / decrease ratio Number of first-level vehicles 150 10 -140 93.3％ Number of vehicles 30 3 -12 40％ Number of distribution centers / distribution centers (number) 13 1 -12 92.3％ Number of express delivery outlets / self-pickup outlets (units) 150 11 -139 92.7％ Transport mileage (km) 2654 537 -2117 80.0％ Delivery mileage (km) 862 68.25 -793.75 92.1％ Total cost (yuan) 23333436 9936241 -13397195 57.41％

[0187] After sharing, the transportation cost has been reduced by more than 50%. At the same time, by screening the efficiency of express delivery outlets, the operating efficiency of shared distribution outlets has been maintained at more than 70%, which can effectively improve the problems of low efficiency, ineffective and repetitive operations of express delivery outlets.

[0188] Joint delivery has been a recent trend in shared express delivery. This paper investigates the joint delivery network planning problem using a three-stage solution approach. The study finds that transitioning from individual deliveries to joint delivery can not only reduce the ineffective construction of express delivery facilities but also reduce delivery costs and improve transportation efficiency. Joint delivery holds broad prospects for development. This paper only examines a three-tiered express delivery network. Future research will continue to optimize and research a variety of complex express delivery network structures, continuously improving the accuracy of the solution algorithm, and providing effective solutions for even more complex joint delivery problems.

Claims

1. A DEA-based method for evaluating the location of common distribution points, characterized in that: The method comprises the following steps: (1) Model construction It is known that vehicles can travel on any road between nodes i and j within the time range not subject to traffic control. The transportation distance at this time is set to Within the time range of traffic control, the vehicle needs to detour and reach the target distribution point via the shortest distance. The transportation distance at this time is set to Right now: Among them, p ij is the distance between the different q routes for transportation between the common distribution points ij, T ki is the start delivery time of the vehicle to the shared distribution point i, [μ1,μ2] is the time range of traffic control; The objective function is expressed as: Among them, f2 is the fixed cost of mobilizing vehicles, K is the set of vehicles, N is the set of common distribution outlets, V k2 is the average speed of the vehicle, λ is the average distance between adjacent traffic lights, φ is the expected travel time of each traffic light, c2 is the transportation cost per unit distance of the vehicle, X ijk2 is the decision variable for whether the vehicle arrives at point j from point i, R2 n is the actual load of the vehicle, W2 is the maximum load of the vehicle, Y k2 A decision variable indicating whether a vehicle will incur a call cost; The constraints are expressed as: Ensure that the total weight of the express delivered by the transport vehicle does not exceed the maximum loading capacity of the delivery vehicle Ensure that the transportation of vehicles is carried out within the working hours of the common distribution center in, The working hours of the joint distribution center; Normal driving distance of a vehicle without traffic control P ij =MinP ij Minimum driving distance of vehicles under time constraints of traffic control Pij=Min(P ij -minP ij ) The actual demand of the distribution point is the same as the actual transportation volume of the vehicle Among them, r n is the demand for outlets, Each vehicle can only transport to one common distribution point Whether vehicle k2 is transported from distribution point i to distribution point j Does vehicle k2 incur a call cost? (2) Determine the location of the optimal post station and distribution network Based on the model in step (1), the unsupervised learning particle swarm optimization algorithm is used to cluster the applicability of the clustering data vector, wherein K-means clustering is used to generate the initial population and obtain the reasonable express station location within the city; the optimal express station location is obtained according to the clustering results, and the reasonable co-distribution network point location is obtained through the immune algorithm; (3) Using DEA to evaluate the efficiency of co-distribution points: The efficiency of the shared distribution network is calculated according to the entropy weight method. The weights of the evaluation indicators are determined by data standardization, information entropy of each indicator, and information entropy calculation. After the weights are determined, the operational efficiency of the shared distribution network is calculated by the data envelopment analysis method based on the relative efficiency of similar decision-making units of multiple performance measurement indicators: Assume that the weight of the output indicator is u n , the weight of the input indicator is v m , a total of y j Output indicators, x i Input indicators, for each part, the efficiency evaluation index is: Based on the calculated operating efficiency of each co-distribution point, screening is performed to obtain the optimal co-distribution point location.

2. A DEA-based common distribution network site selection evaluation method according to claim 1, characterized in that: The output indicators include the efficiency of warehouse inbound and outbound operations, express delivery processing efficiency, and service quality; the input indicators include the total area of outlets, the number of outlet staff, the number of outlet equipment, and the total cost of outlets; The efficiency of warehouse entry and exit = (warehouse entry time - receipt time / average port entry time) + (shipping time - warehouse exit time / average port exit time); Express delivery processing efficiency = delivery volume / number of couriers; Service quality is the weighted average of the false receipt complaint rate, loss and damage rate, upgraded complaint acceptance rate, and secondary complaint rate.

Images