Platform logistics transportation path planning method based on online learning of time-sensitive demand

Abstract

The present invention provides a platform logistics transportation route planning method based on online learning of time-sensitive demand, and relates to the field of logistics technology. The method takes into account the relationship between the fulfillment time and the average value of the demand in the time-sensitive demand, and the online data generated during the operation of the platform. It can realize the online learning of the customer's time-sensitive demand based on the online data, and further use the learned time-sensitive demand function to plan the platform logistics transportation route. Compared with the traditional transportation route planning method, it can effectively characterize the relationship between the customer demand and the fulfillment time, making the characterization of customer demand more accurate, and more conducive to designing a transportation route planning scheme that maximizes the platform's benefits, and avoiding the inaccuracy of the transportation route planning scheme due to changes in the fulfillment time; compared with the traditional transportation route planning method, this method can continuously adapt to the changing customer time-sensitive needs and has a wider range of applications.

Description

Technical Field

[0001] The present invention belongs to the field of logistics technology, and specifically relates to a platform logistics transportation path planning method based on time-sensitive demand online learning. Background Art

[0002] The platform logistics model refers to a logistics model that relies on an online platform, integrates the warehousing, transit and transportation resources of multiple logistics companies, and collaborates with multiple logistics companies to complete order fulfillment. Its most important feature is that the transportation tasks on different sections of the transportation routes planned by the platform for customers can be undertaken by different logistics companies.

[0003] The platform logistics model, by virtue of its integration and coordinated use of logistics resources, effectively improves the efficiency of logistics fulfillment and reduces the operating costs of enterprises. Currently, the platform logistics model has been successfully applied in industries such as retail and manufacturing. Transport route planning is an important part of the platform logistics operation process. Figure 1 The design of the transportation path planning scheme is completed by selecting the transfer nodes and the logistics enterprises that undertake the transportation tasks of different sections in the potential network composed of all logistics resources of the platform shown. Since different logistics enterprises usually have different transportation time and unit transportation cost, and different transfer centers have different processing time and unit processing cost, the different transportation path planning schemes designed by the platform usually have different fulfillment time and operation cost.

[0004] In the actual operation background represented by the retail industry, customers are sensitive to the order fulfillment time. This sensitivity is specifically manifested in that the average demand generated by the customer cluster in a certain area decreases as the fulfillment time increases. This type of demand is usually called time-sensitive demand. However, the existing transportation route planning technology usually ignores the impact of time-sensitive demand on the operation effect of the transportation route planning scheme. At the same time, the existing technology usually uses a static statistical regression method based on historical data to characterize customer demand. It lacks the mining and application of online data generated during the platform logistics operation process, and cannot effectively reflect the dynamic and changeable actual needs. Therefore, it is impossible to guarantee the accuracy of the customer demand characterization, resulting in poor adaptability of the designed transportation route plan to actual demand, waste or insufficient transportation capacity, and inflated platform operation costs, which makes it difficult to meet customer expectations. Summary of the invention

[0005] In view of the shortcomings of the above-mentioned prior art, the present invention provides a platform logistics transportation route planning method based on online learning of time-sensitive demand. The method takes into account the relationship between the fulfillment time and the average value of the demand in the time-sensitive demand, and the online data generated during the platform operation. It can realize online learning of customer time-sensitive demand based on online data, and further use the learned time-sensitive demand function to carry out platform logistics transportation route planning.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A platform logistics transportation path planning method based on time-sensitive demand online learning, the method includes the following steps:

[0008] Step 1: For the current commodity to be transported, its historical sales data is statistically analyzed to obtain the average demand and average fulfillment time of the commodity. Referring to the average demand and average fulfillment time, several possible time-sensitive demand functions are estimated to form the potential time-sensitive demand function set Φ required for online learning f ;

[0009] Step 2: Determine the information and parameters of the potential network composed of all logistics resources on the platform;

[0010] Step 3: From the set Φ f A function is randomly selected as the initial customer actual time-sensitive demand function identified by the platform;

[0011] Step 4: Based on the information and parameters of the potential network composed of all logistics resources on the platform, the actual time-sensitive demand function of customers identified by the platform, and with the goal of maximizing platform revenue, the selection of transit nodes and the selection of logistics companies that undertake transportation tasks are used as decision variables to establish a platform logistics transportation path planning model;

[0012] Step 5: Solve the platform logistics transportation path planning model and obtain the transportation path planning solution S p ;

[0013] Step 6: Platform Application S p , obtain the average value of online customers’ real time-sensitive demand data;

[0014] Step 7: Calculate the set Φ f All functions in S p The corresponding customer time-sensitive demand mean and their respective deviations from the mean of the actual time-sensitive demand data are found in the set Φ f The function with the smallest deviation value;

[0015] Step 8: If the minimum deviation value is within the allowable error range, output the current transportation path planning scheme; otherwise, determine whether the minimum deviation value is greater than the preset maximum allowable deviation range: if so, use the transportation path planning scheme S p The corresponding transportation time and the mean of the real time-sensitive demand data are used as references to estimate the new time-sensitive demand function and add it to the set Φ f Then go to step 9; if not, execute step 9;

[0016] Step 9: Φ fThe function with the smallest deviation value is updated to the actual time-sensitive function of the customer recognized by the platform, and steps 4 to 8 are repeated according to the method of steps 4 to 8 until the minimum deviation value is within the allowable error range and the current transportation path planning plan is output.

[0017] Furthermore, according to the platform logistics transportation route planning method based on time-sensitive demand online learning, the time-sensitive demand function expresses the average customer demand at the fulfillment time corresponding to a certain transportation route planning scheme for the goods to be transported.

[0018] Further, according to the platform logistics transportation path planning method based on time-sensitive demand online learning, the time-sensitive demand function obtained by historical learning of commodities of similar categories to the commodities to be transported is introduced in step 1, and together with all estimated time-sensitive demand functions of the current commodities to be transported, the set Φ f .

[0019] Furthermore, according to the platform logistics transportation path planning method based on time-sensitive demand online learning, the information and parameters of the potential network composed of all logistics resources of the platform described in step 2 include the connection relationship between different nodes in the potential network, the processing time and unit processing cost of the transit nodes, the number of logistics companies between connected nodes and the transportation time and unit transportation cost of each logistics company.

[0020] Further, according to the platform logistics transportation path planning method based on time-sensitive demand online learning, the platform logistics transportation path planning model includes:

[0021] The objective function shown in formula (4) is:

[0022] max R=bd p (T p )-C(S p ) (4)

[0023] In the above formula, R is the platform revenue, RMB; b is the unit remuneration obtained by the platform for completing the order, RMB / item; p is the actual time-sensitive demand function of the customer recognized by the platform, d p (t) In the potential time-sensitive demand function set Φ f Number in; S p Indicates that in the recognition p (t) The transportation route planning scheme adopted in the case of actual time-sensitive needs of customers; C(S p ) is the transportation path planning solution S p Corresponding operating cost, yuan; T p Plan the transport route S p Transportation time, h; d p (Tp ) is to identify p (t) Adopt the transportation route planning solution S for the actual time-sensitive needs of customers p The demand generated by the customer cluster at that time, pieces;

[0024] The constraints shown in equations (5)-(9) are:

[0025] Formula (5) is the flow balance constraint, which is used to ensure that the transportation path planning solution obtained by solving the model is a reasonable path:

[0026]

[0027] In the above formula, It is the set of all nodes in the potential network composed of all logistics resources of the platform; It is the set of all supply points in the potential network composed of all logistics resources of the platform; It is the set of all transfer nodes in the potential network composed of all logistics resources of the platform; It is the collection of all customers in the potential network composed of all logistics resources of the platform; is the set of all logistics enterprises on the transport section with starting point number i and end point number j; ijk is a decision variable, representing the choice of the logistics enterprise that undertakes the transportation task, indicating whether the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j is selected in the transportation route planning scheme, x ijk =1 means selected, x ijk =0 means not selected;

[0028] Formula (6) and Formula (7) are unique path constraints, which are used to ensure that each customer order is delivered through the only feasible path:

[0029]

[0030]

[0031] Formula (8) and Formula (9) are decision variable value range constraints, which are used to stipulate that the decision variables take values ​​in a reasonable {0,1} set;

[0032]

[0033]

[0034] In the above formula, y l is a decision variable, representing the choice of the transfer node, indicating whether the transfer node numbered l is selected in the transportation path planning scheme, y l =1 means selected, yl =0 means not selected; x ijk is a decision variable, representing the choice of the logistics enterprise that undertakes the transportation task, indicating whether the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j is selected in the transportation route planning scheme, x ijk =1 means selected, x ijk =0 means not selected.

[0035] Furthermore, according to the platform logistics transportation path planning method based on time-sensitive demand online learning, the transportation path planning solution S is calculated according to formula (2): p The transportation time T p ;

[0036]

[0037] Among them, TP ijk is the transportation time of logistics enterprise numbered k on the transportation section with starting point numbered i and end point numbered j, h; TC l is the processing time of the transfer node numbered l, h.

[0038] Furthermore, according to the platform logistics transportation path planning method based on time-sensitive demand online learning, the transportation path planning solution S is calculated according to formula (3): p The corresponding operating cost C(S p ),Yuan;

[0039]

[0040] Among them, CP ijk is the unit transportation cost of the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j, RMB; CC l is the unit processing fee of the transit node numbered l, RMB.

[0041] In general, the above technical solutions conceived by the present invention have the following beneficial effects compared with the prior art:

[0042] (1) Compared with the traditional transportation route planning method, the method of the present invention introduces time-sensitive demand into the transportation route planning model, which can effectively characterize the relationship between customer demand and fulfillment time, and thus can more effectively perform transportation route planning and avoid inaccurate transportation route planning schemes caused by changes in fulfillment time;

[0043] (2) Compared with the customer demand characterization method based on statistical regression of historical data, the method of the present invention uses online learning to explore and obtain the customer time-sensitive demand function based on online customer demand data, making the characterization of customer demand more accurate and more conducive to designing a transportation path planning solution that maximizes the platform's revenue;

[0044] (3) Compared with traditional transportation route planning methods, the method of the present invention can continuously adapt to changing customer time-sensitive needs and has a wider scope of application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific methods in the embodiments of the present invention, the relevant drawings involved in the embodiments will be briefly described below. The following drawings are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative changes.

[0046] Figure 1 It is a potential network diagram composed of all logistics resources of the platform under the platform logistics model;

[0047] Figure 2 This is a flow chart of a platform logistics transportation path planning method based on online learning of time-sensitive demand in this embodiment;

[0048] Figure 3 This is a flow chart of a platform logistics transportation path planning method based on online learning of time-sensitive demand in this embodiment;

[0049] Figure 4 This is a distribution diagram of real customer demand data collected by the platform of this implementation mode;

[0050] Figure 5 This is a schematic diagram of the transportation route planning scheme with the greatest platform benefit in this implementation mode;

[0051] Figure 6 This is a curve chart of daily revenue changes on the platform in this implementation mode;

[0052] Figure 7 This is a curve chart showing the cumulative revenue changes of the platform in this implementation mode;

[0053] Figure 8 A Gantt chart of the customer time-sensitive demand function set for the platform in this implementation mode. DETAILED DESCRIPTION

[0054] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present application are given in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thoroughly and comprehensively understood.

[0055] As those skilled in the art know, transportation route planning in platform logistics mode is a process in which Figure 1 The platform shown is a planning process for selecting available resources in a potential network consisting of all logistics resources. Figure 1 The source point 1 represents the starting point of the transportation task, the confluence point 10 represents the destination of the transportation task, and the intermediate nodes 2-9 represent the available transfer nodes such as warehouses, transfer stations, distribution centers, and docking and sorting centers owned by the platform. The directed edge between two nodes in the figure represents the logistics enterprise that can undertake the transportation task of the road section between the two nodes. Among them, there may be multiple available logistics enterprises between the two nodes, and their transportation capacity, transportation time, unit cost, etc. are not exactly the same, that is, the potential network of platform logistics transportation is represented as a multi-graph structure. The process of the platform completing the transportation task is to start from the starting point according to the designed transportation path planning scheme, and finally transport it to the destination through different logistics enterprises through different transfer nodes.

[0056] For platform logistics transportation route planning, the core idea of ​​the present invention is: first, a set of potential time-sensitive demand functions is constructed by estimating based on the statistical characteristics of historical commodity sales data; then, any function is selected from the set of potential time-sensitive demand functions as the initial platform-recognized customer time-sensitive demand function, a transportation route planning model is established and the corresponding transportation route planning scheme is solved; then, the platform collects real customer demand data by applying the obtained transportation route planning scheme, and calculates the average customer time-sensitive demand corresponding to all functions in the set of potential time-sensitive demand functions under the above transportation route planning scheme and their respective deviations from the average of the online real time-sensitive demand data. If the minimum deviation value is not within the preset allowable error range and is greater than the preset maximum allowable deviation range, a new time-sensitive demand function is estimated with the current data as the reference point and added to the set of potential time-sensitive demand functions, and the function with the smallest deviation value is updated and set as the actual customer time-sensitive function recognized by the platform. The above calculation process is repeated until the minimum deviation value is within the preset allowable range, and the transportation route planning scheme is output.

[0057] The method of the present invention is now further described in detail by taking a platform logistics operator's transportation route planning for a certain electronic product category as an example. Figure 2 and Figure 3is a flow chart of the platform logistics transportation path planning method based on time-sensitive demand online learning in this embodiment, such as Figure 2 and Figure 3 As shown, the platform logistics transportation path planning method based on time-sensitive demand online learning includes the following steps:

[0058] Step 1: Count the historical sales data of the goods to be transported to obtain the average demand and average fulfillment time of the goods, and estimate several possible time-sensitive demand functions based on the average demand and average fulfillment time to form a potential time-sensitive demand function set Φ required for online learning f ;

[0059] This embodiment obtains the statistical characteristics of the electronic product as shown in Table 1 by collecting statistics on the historical sales data of the electronic product to be transported.

[0060] Table 1 Statistical characteristics of the historical sales data of electronic products

[0061]

[0062] The features obtained from the limited historical sales data of the electronic products are As a reference point, several curves can be estimated around this reference point. A time-sensitive demand function, wherein there is no limit on the number of time-sensitive demand function estimates, such as the time-sensitive demand function estimated around the features shown in Table 1 as shown in Table 2.

[0063] Table 2 Estimation results of time-sensitive demand function

[0064]

[0065] Among them, t represents the fulfillment time, d p (t) represents the average customer demand when the fulfillment time is t under the time-sensitive demand function numbered p. From the time-sensitive demand function shown in Table 2, it can be understood that the time-sensitive demand function expresses the average customer demand d when the fulfillment time is t corresponding to the transported goods under a certain transport route planning scheme. p (t).

[0066] Construct a set of potential time-sensitive demand functions Φ required for online learning based on the estimated time-sensitive demand functions f :The set Φ is composed of all estimated time-sensitive demand functions f It is better to introduce the time-sensitive demand function obtained from the historical learning of commodities of similar categories to the commodities to be transported and combine it with all estimated time-sensitive demand functions to form a set Φ f ;

[0067] In order to improve the utilization of the potential time-sensitive demand function set Φ f To learn the effectiveness of customer demand, we can introduce the time-sensitive demand function obtained from the historical learning of similar categories of goods into the set Φ f Table 3 shows the time-sensitive demand function obtained by historical learning of commodities in similar categories to the electronic commodities. It should be noted that the introduction of the time-sensitive demand function obtained by historical learning of commodities in similar categories is only to improve the effect of online learning and is not necessary.

[0068] Table 3 Time-sensitive demand function obtained from historical learning of similar category products

[0069]

[0070] Step 2: Determine the information and parameters of the potential network composed of all logistics resources on the platform;

[0071] The information and parameters of the potential network composed of all logistics resources on the platform should include the connection relationship between different nodes in the potential network, the processing time and unit processing cost of the transit nodes, the number of logistics companies between the connected nodes, and the transportation time and unit transportation cost of each logistics company. Figure 1 The information and parameters of the transit nodes and logistics enterprises on the potential network shown in Table 4. The information and parameters of the potential network nodes composed of all logistics resources of a certain platform shown in Table 4, and the information and parameters of the potential network edges composed of all logistics resources of a certain platform shown in Table 5.

[0072] Table 4 Information and parameters of potential network nodes composed of all logistics resources on a certain platform

[0073]

[0074]

[0075] Table 5 Information and parameters of potential network edges composed of all logistics resources on a certain platform

[0076]

[0077] Step 3: From the set of potential time-sensitive demand functions Φ f Randomly select a function d from p (t) is the actual time-sensitive demand function of the customer identified by the initial platform.

[0078] Where p is the set of potential time-sensitive demand functions Φ f The number of the time-sensitive demand function in the range of {1,…,K}, where K is the set of potential time-sensitive demand functions Φ f The total number of functions present in .

[0079] In this embodiment, d is randomly selected 1 (t)=-10t+150 is used as the actual time-sensitive demand function of the customer recognized by the initial platform.

[0080] Step 4: Based on the information and parameters of the potential network composed of all logistics resources on the platform, the actual time-sensitive demand function of customers identified by the platform, and with the goal of maximizing platform revenue, the selection of transit nodes and the selection of logistics companies that undertake transportation tasks are used as decision variables to establish a platform logistics transportation path planning model;

[0081] In this embodiment, the decision variable y l Represents the selection of the transfer node, indicating whether the transfer node numbered l is selected in the transportation path planning scheme, y l =1 means selected, y l =0 represents not selected; the decision variable is x ijk represents the selection of the logistics enterprise that undertakes the transportation task, indicating whether the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j is selected in the transportation route planning scheme, x ijk =1 means selected, x ijk =0 means not selected.

[0082] The platform revenue refers to the difference between the remuneration obtained by completing the order and the operating cost paid. In this embodiment, the platform revenue R is calculated according to formula (1), RMB;

[0083] R = b·d p (T p )-C(S p ) (1)

[0084] In the above formula, b is the unit remuneration obtained by the platform for completing the order. In this embodiment, the unit remuneration obtained by the platform for completing the order is b=15, yuan / item; p is the actual time-sensitive demand function of the customer identified by the platform in the potential time-sensitive demand function set Φ f The number in d p (t) is the actual time-sensitive needs of the customer; S p Indicates that in the recognition p (t) The transportation route planning scheme adopted in the case of actual time-sensitive needs of customers; C(S p ) is the transportation path planning solution S p Corresponding operating cost, yuan; T p Plan the transport route S p Transportation time, h; d p (T p ) is to identify p(t) Adopt the transportation route planning solution S for the actual time-sensitive needs of customers p The demand generated by the customer cluster at that time, pieces.

[0085] In this embodiment, the transportation path planning solution S is calculated according to formula (2): p The transportation time T p , h;

[0086]

[0087] Among them, TP ijk is the transportation time of logistics enterprise numbered k on the transportation section with starting point numbered i and end point numbered j, h; TC l is the processing time of the transfer node numbered l, h; It is the set of all nodes in the potential network composed of all logistics resources of the platform; It is the set of all transfer nodes in the potential network composed of all logistics resources of the platform; It is the set of all logistics companies on the transport route with the starting point numbered i and the end point numbered j.

[0088] In this embodiment, the transportation path planning solution S is calculated according to formula (3): p The corresponding operating cost C(S p ),Yuan;

[0089]

[0090] Among them, CP ijk is the unit transportation cost of the logistics enterprise numbered k on the transportation section with starting point numbered j and end point numbered j, RMB; CC l is the unit processing fee of the transit node numbered l, RMB.

[0091] Then, the following platform logistics transportation path planning model is established, where the objective function is shown in formula (4):

[0092] max R=bd p (T p )-C(S p ) (4)

[0093] The constraints are shown in equations (5)-(9):

[0094] The flow balance constraint is shown in formula (5), which is used to ensure that the transportation path planning solution obtained by solving the model is a reasonable path:

[0095]

[0096] in, It is the set of all supply points in the potential network composed of all logistics resources of the platform; It is the set of all customers in the potential network composed of all logistics resources of the platform.

[0097] The unique path constraint is shown in equations (6) and (7), which is used to ensure that each customer order is delivered via the only feasible path:

[0098]

[0099]

[0100] The decision variable value range constraints are shown in equations (8) and (9), which are used to specify that the decision variables take values ​​in a reasonable {0,1} set:

[0101]

[0102]

[0103] Step 5: Solve the platform logistics transportation path planning model and obtain the transportation path planning solution S p ;

[0104] The platform logistics transportation path planning model shown in equations (4)-(9) is a 0-1 integer quadratic programming model. The optimal solution can be obtained through the existing integer programming technology, that is, the output decision variable x ijk and l The optimal value of is obtained to obtain the transportation path planning solution.

[0105] Step 6: The platform applies the transportation path planning solution S p , obtain the average value of online customers’ real time-sensitive demand data;

[0106] The specific method is: during the application period w, the transportation path planning solution S obtained in step 5 is used. p To fulfill the contract, the platform obtains a real sales data every day, that is, the platform obtains {X 1 ,…,X w There are w real sales data in total. The mean of online real customer time-sensitive demand data is calculated by formula (10);

[0107]

[0108] Where w is the application cycle of the transportation route planning scheme set by the platform, days; X q is the actual customer demand volume collected by the platform on the qth day of the application cycle, pieces; To apply the transport path planning solution S within a complete application cycle pThe average of the actual customer demand collected by the platform under the condition of .

[0109] In this embodiment, the platform transportation path planning scheme application cycle w is set to 60, and the platform will collect Figure 4 Actual customer demand data shown.

[0110] Step 7: Calculate the set of potential time-sensitive demand functions Φ f All functions in the transportation path planning solution S p The corresponding customer time-sensitive demand mean and their deviations from the mean of the online real time-sensitive demand data are used to find the potential time-sensitive demand function set Φ f The function with the smallest deviation value;

[0111] The specific method is as follows: Calculate the potential time-sensitive demand function set Φ f All functions in the transportation path planning solution S p The corresponding customer time-sensitive demand mean under the following conditions is used to calculate the potential time-sensitive demand function set Φ by formula (11): f In the transportation path planning scheme S p The function number m with the smallest deviation from the mean of the online real time-sensitive demand data is obtained, and the corresponding time-sensitive demand function d is obtained. m (t);

[0112]

[0113] Where, Δd m is the set of potential time-sensitive demand functions Φ f The function numbered m in the transportation path planning scheme S p The corresponding average value of customer time-sensitive demand Compared with the average of online real time-sensitive demand data The deviation value of .

[0114] D m (t) is set to the actual time-sensitive demand function of the customer identified by the updated platform;

[0115] Step 8: If the minimum deviation value If the error is within the allowable range, the current transportation route planning solution is output; otherwise, the minimum deviation value is determined. Is it greater than the preset maximum allowable deviation range Δd? max :If yes, then use the transportation route planning solution S p The corresponding transportation time T p , the average value of online real time-sensitive demand data As the reference point, estimate the new time-sensitive demand function and add it to the potential time-sensitive demand function set Φ f and go to step 10; if not, execute step 10;

[0116] Step 9: Take Φ f The function with the smallest deviation value is updated and set as the actual time-sensitive function of the customer recognized by the platform. According to the method of steps 4 to 8, steps 4 to 8 are repeated until the minimum deviation is reached. Output the current transportation route planning solution when it is within the allowable error range.

[0117] According to the method of the present invention, based on the information and parameters of the potential network composed of all logistics resources of the above platform and the application cycle of the platform transportation path planning scheme, the above platform logistics transportation path planning model is established through computer programming and online learning of customer time-sensitive demand based on online data is realized, and the optimal customer time-sensitive demand function is calculated as d 7 (t) = -30t + 350, the calculated potential time-sensitive demand function set Φ f The deviation values ​​of all functions in are shown in Table 6. The output of the transportation path planning scheme that maximizes the platform revenue is as follows: Figure 5 In this embodiment, the platform daily revenue change curve is as shown in the 600-day total accounting period of the platform. Figure 6 As shown; the platform cumulative revenue curve is as follows Figure 7 As shown; the Gantt chart of the customer time-sensitive demand function set by the platform is as follows Figure 8 As shown. Figure 8 As can be seen in the figure, in this embodiment, the platform obtains the transportation path planning scheme that maximizes the platform's revenue by adjusting the transportation path planning scheme only twice. The results show that after the platform uses online data to learn the customer's time-sensitive demand function online, it can quickly adjust the time-sensitive demand function set to maximize the platform's revenue, and calculate the transportation path planning scheme to achieve platform profitability.

[0118] Table 6 Optimal transportation time and demand deviation values ​​of all functions in the potential time-sensitive demand function set

[0119]

[0120]

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. Therefore, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A platform logistics transportation path planning method based on online learning of time-sensitive requirements, characterized by: The method comprises the following steps: Step 1: For the current commodity to be transported, its historical sales data is statistically analyzed to obtain the average demand and average fulfillment time of the commodity. Referring to the average demand and average fulfillment time, several possible time-sensitive demand functions are estimated to form the potential time-sensitive demand function set Φ required for online learning f ; Step 2: Determine the information and parameters of the potential network composed of all logistics resources on the platform; Step 3: From the set Φ f A function is randomly selected as the initial customer actual time-sensitive demand function identified by the platform; Step 4: Based on the information and parameters of the potential network composed of all logistics resources on the platform, the actual time-sensitive demand function of customers identified by the platform, and with the goal of maximizing platform revenue, the selection of transit nodes and the selection of logistics companies that undertake transportation tasks are used as decision variables to establish a platform logistics transportation path planning model; Step 5: Solve the platform logistics transportation path planning model and obtain the transportation path planning solution S p ; Step 6: Platform Application S p , obtain the average value of online customers’ real time-sensitive demand data; Step 7: Calculate the set Φ f All functions in S p The corresponding customer time-sensitive demand mean and their respective deviations from the mean of the actual time-sensitive demand data are found in the set Φ f The function with the smallest deviation value; Step 8: If the minimum deviation value is within the allowable error range, output the current transportation path planning scheme; otherwise, determine whether the minimum deviation value is greater than the preset maximum allowable deviation range: if so, use the transportation path planning scheme S p The corresponding transportation time and the mean of the real time-sensitive demand data are used as references to estimate the new time-sensitive demand function and add it to the set Φ f Then go to step 9; if not, execute step 9; Step 9: Φ f The function with the smallest deviation value is updated to the actual time-sensitive function of the customer recognized by the platform, and steps 4 to 8 are repeated according to the method of steps 4 to 8 until the minimum deviation value is within the allowable error range and the current transportation path planning plan is output.

2. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 1 is characterized in that: The time-sensitive demand function expresses the average customer demand for the goods to be transported at the corresponding fulfillment time using a certain transportation route planning scheme.

3. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 1 is characterized in that: The time-sensitive demand function obtained by historical learning of commodities of similar categories to the commodities to be transported in step 1 and all estimated time-sensitive demand functions of the current commodities to be transported together constitute the set Φ f .

4. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 1 or 2 is characterized in that: The information and parameters of the potential network composed of all logistics resources of the platform described in step 2 include the connection relationship between different nodes in the potential network, the processing time and unit processing cost of the transit nodes, the number of logistics companies between connected nodes, and the transportation time and unit transportation cost of each logistics company.

5. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 1 is characterized in that: The platform logistics transportation path planning model includes: The objective function shown in formula (4) is: maxR=b·d p (T p )-C(S p ) (4) In the above formula, R is the platform revenue, RMB; b is the unit remuneration obtained by the platform for completing the order, RMB / item; p is the actual time-sensitive demand function of the customer recognized by the platform, d p (t) In the potential time-sensitive demand function set Φ f Number in; S p Indicates that in the recognition p (t) The transportation route planning scheme adopted in the case of actual time-sensitive needs of customers; C(S p ) is the transportation path planning solution S p Corresponding operating cost, yuan; T p Plan the transport route S p Transportation time, h; d p (T p ) is to identify p (t) Adopt the transportation route planning solution S for the actual time-sensitive needs of customers p The demand generated by the customer cluster at that time, pieces; The constraints shown in equations (5)-(9) are: Formula (5) is the flow balance constraint, which is used to ensure that the transportation path planning solution obtained by solving the model is a reasonable path: In the above formula, It is the set of all nodes in the potential network composed of all logistics resources of the platform; It is the set of all supply points in the potential network composed of all logistics resources of the platform; It is the set of all transfer nodes in the potential network composed of all logistics resources of the platform; It is the collection of all customers in the potential network composed of all logistics resources of the platform; is the set of all logistics enterprises on the transport section with starting point number i and end point number j; ijk is a decision variable, representing the choice of the logistics enterprise that undertakes the transportation task, indicating whether the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j is selected in the transportation route planning scheme, x ijk =1 represents being selected, x ijk =0 means not selected; Formula (6) and Formula (7) are unique path constraints, which are used to ensure that each customer order is delivered through the only feasible path: Formula (8) and Formula (9) are decision variable value range constraints, which are used to stipulate that the decision variables take values ​​in a reasonable set of {0, 1}; In the above formula, y l is a decision variable, representing the choice of the transfer node, indicating whether the transfer node numbered l is selected in the transportation path planning scheme, y l =1 means selected, y l =0 means not selected; x ijk is a decision variable, representing the choice of the logistics enterprise that undertakes the transportation task, indicating whether the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j is selected in the transportation route planning scheme, x ijk =1 means selected, x ijk =0 means not selected.

6. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 5 is characterized in that: According to formula (2), the transportation path planning solution S is calculated. p The transportation time T p ; Among them, TP ijk is the transportation time of logistics enterprise numbered k on the transportation section with starting point numbered i and end point numbered j, h; TC l is the processing time of the transfer node numbered l, h.

7. The platform logistics transportation path planning method based on time-sensitive demand online learning according to claim 5 is characterized in that: According to formula (3), the transportation path planning solution S is calculated. p The corresponding operating cost C(S p ),Yuan; Among them, CP ijk is the unit transportation cost of the logistics enterprise numbered k on the transportation section with the starting point numbered i and the end point numbered j, RMB; CC l is the unit processing fee of the transit node numbered l, RMB.

Images