Cargo matching method, system and terminal based on digital multi-parameter dynamic optimization

Abstract

The present invention discloses a cargo quantity matching method, system and terminal based on digital multi-parameter dynamic optimization. The method includes: obtaining cargo quantity data of multiple freight objects in multiple rounds of single cycles, constructing and training a long short-term memory network model based on the cargo quantity data to predict the current cargo quantity data of all freight objects in the current cycle; inputting the cargo quantity data into the constructed cargo quantity matching theoretical model, and outputting the optimal service level weight of the exporter; constructing a decision process model, inputting the optimal service level weight into the decision process model, and outputting the optimal cargo quantity matching decision. The present invention focuses on the instant single-cycle transportation service procurement scenario. For the enterprise's multi-cycle instant transportation orders and uncertain transportation needs, it adopts multi-attribute reverse selection of transportation service providers and determines the cargo quantity of freight, effectively solving the problem of inflexibility of long-term procurement agreements, reducing procurement costs, and improving the exporter's freight efficiency throughout the entire cycle.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a cargo quantity matching method, system, terminal and computer-readable storage medium based on digital multi-parameter dynamic optimization. Background Art

[0002] As the exporter with demand for transportation capacity, it is necessary not only to consider the transportation rates offered by international logistics and transportation providers, but more importantly, to consider the service level that the freight object can provide, such as transportation frequency, transportation time, risk management, response speed, and its adaptability to the exporter's own logistics needs.

[0003] However, the supply of international logistics capacity is susceptible to the complex influence of many uncertain factors such as sudden disasters, international relations, and trade decisions. In the short term, there will be a cliff-like decline in cross-border capacity supply or even a direct interruption of supply. For major exporters, the uncertainty of cross-border logistics is reflected in the logistics procurement link, and in the prominent business problems of logistics capacity supply such as container shortages, unstable capacity supply, and large fluctuations in transportation costs. As a result, multiple shipments are required, resulting in cost waste for exporters and significantly reducing the efficiency of freight transportation for exporters.

[0004] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0005] The main purpose of the present invention is to provide a cargo matching method, system, terminal and computer-readable storage medium based on digital multi-parameter dynamic optimization, aiming to solve the problem that the cargo matching decision-making mechanism adopted by the exporter in the prior art is difficult to effectively and quickly respond to the supply and demand mismatch problem within a large cycle, thereby resulting in an overall reduction in the exporter's freight efficiency.

[0006] To achieve the above objectives, the present invention provides a cargo quantity matching method based on digital multi-parameter dynamic optimization, the cargo quantity matching method based on digital multi-parameter dynamic optimization comprising the following steps:

[0007] Obtaining freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, using the freight information to train the constructed long short-term memory network model, inputting the current cargo volume into the trained long short-term memory network model, and outputting the current cargo volume data of all freight objects in the current cycle;

[0008] Inputting the current cargo volume data into the constructed cargo volume matching theoretical model, analyzing the current cargo volume data, and outputting the exporter's optimal service level weight in the current period;

[0009] A decision process model is constructed based on the current cargo volume and the current cargo volume data, the optimal service level weight is input into the decision process model, the decision process model is analyzed using a deep network algorithm, and the optimal cargo volume matching decision of the exporter is output.

[0010] Optionally, in the cargo volume matching method based on digital multi-parameter dynamic optimization, the cargo information includes cargo volume data and predicted cargo volume data;

[0011] The method includes obtaining freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, using the freight information to train the constructed long short-term memory network model, inputting the current cargo volume into the trained long short-term memory network model, and outputting the current cargo volume data of all freight objects in the current cycle. Specifically, the method includes:

[0012] Obtain cargo volume data for multiple freight objects in multiple single cycles and predicted cargo volume data for all freight objects in each cycle, and build an initial long-short-term memory network model;

[0013] Inputting the cargo volume data and the predicted cargo volume data into the initial long short-term memory network model for training to obtain a trained long short-term memory network model;

[0014] Input the current cargo volume of the exporter in the current period into the trained long short-term memory network model, and output the current cargo volume data of all freight objects in the current period;

[0015] The sum of the current cargo volume data of all freight objects is equal to the current cargo volume.

[0016] Optionally, the cargo volume matching method based on digital multi-parameter dynamic optimization, wherein the cargo volume data and the predicted cargo volume data are input into the initial long short-term memory network model for training to obtain a trained long short-term memory network model, specifically includes:

[0017] Obtain model parameters of the initial long short-term memory network model, input the cargo volume data and the predicted cargo volume data into the initial long short-term memory network model for training, and output the difference between the cargo volume data and the predicted cargo volume data:

[0018] ;

[0019] in, Indicates the freight object The difference between the forecast volume data and the volume data for the period, Indicates the freight object Forecast cargo volume data for the period, Indicates that the freight object is Periodic cargo volume data, Indicates the freight object In the Forecast cargo volume data for the period, Indicates the freight object In the Periodic cargo volume data, Indicates the quantity of the freight object, Indicates that the exporter The supply of goods in the cycle, Indicates the Freight objects;

[0020] The gradient descent algorithm is used to update the model parameters according to the difference to obtain a trained long short-term memory network model:

[0021] ;

[0022] in, represents the updated model parameters, represents the model parameters, It represents the delay cost that the exporter needs to pay for the delayed goods order. represents the gradient of the model parameters.

[0023] Optionally, the cargo volume matching method based on digital multi-parameter dynamic optimization, wherein the current cargo volume data is input into a constructed cargo volume matching theoretical model, the current cargo volume data is analyzed, and the optimal service level weight of the exporter in the current cycle is output, specifically includes:

[0024] The current cargo volume data is extracted and input into the constructed cargo volume matching theoretical model. The optimal transportation service level of each freight object in the current cargo volume data is analyzed in the cargo volume matching theoretical model to obtain the optimal service level weight of the exporter in the current period and output it:

[0025] ;

[0026] ;

[0027] ;

[0028] ;

[0029] in, represents the optimal single item benefit, The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object In the The optimal transport service level for the cycle, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, It represents the basic income brought by the freight object completing one unit of transportation. Express obedience The probability distribution function of the uniform distribution on , Indicates that the freight object is Periodic cargo volume data, express The probability density function of represents the optimal service level weight of the exporter in a single period, represents the cost conversion factor, The utility conversion factor that represents the exporter's transportation service level for the freight object.

[0030] Optionally, the cargo volume matching method based on digital multi-parameter dynamic optimization, wherein the current cargo volume data is input into a constructed cargo volume matching theoretical model, the current cargo volume data is analyzed, and the optimal service level weight of the exporter in the current period is output, further comprises:

[0031] Determine the bidding standards for the freight objects in the freight volume data within a single period, and construct a cost function based on the bidding standards:

[0032] ;

[0033] in, Indicates the freight object Bidding standards, Indicates the freight object In the Bidding standards for the cycle, Indicates the freight object In the The unit transport rate for the cycle, Indicates the freight object In the The transport service level of the cycle;

[0034] ;

[0035] in, Indicates the freight object The cost function, Indicates the freight object Cost conversion coefficient that converts cost into service capability, Indicates the basic cost of each freight object;

[0036] Construct a scoring function based on the bidding criteria and the cost function:

[0037] ;

[0038] in, Indicates the freight object Scoring function for a single bid, Indicates that the exporter Service level weight within the cycle;

[0039] Construct the utility reward function based on the cost function:

[0040] ;

[0041] ;

[0042] in, represents the utility return function of the freight object, Represents the value function of logistics service level for freight objects, Indicates the logistics service level for freight objects The value function of represents the cost function of the freight object, represents the service level weight of the exporter, Indicates the transportation service level of the freight object;

[0043] According to the utility reward function, the optimal equilibrium service bidding level of each freight object is obtained, and the optimal decision space of each freight object is determined according to the optimal equilibrium service bidding level.

[0044] Optionally, the cargo volume matching method based on digital multi-parameter dynamic optimization, wherein a decision process model is constructed based on the current cargo volume and the current cargo volume data, the optimal service level weight is input into the decision process model, and a deep network algorithm is used for analysis to output the optimal cargo volume matching decision of the exporter, further comprising:

[0045] According to the number of cycles and the remaining inventory of the exporter, a state transition equation is constructed and the decision space of the exporter is determined:

[0046] ;

[0047] ;

[0048] in, Indicates that the exporter is in a single cycle The state transfer equation within, Indicates the cycle, represents the total number of auction cycles, Indicates the The inventory quantity of the cycle, Indicates the total inventory quantity. represents the exporter’s decision space, Indicates that the exporter is in a single cycle Transportation needs within Indicates the basic release quantity. Indicates that the freight object is Cargo volume data for the period.

[0049] Optionally, the cargo volume matching method based on digital multi-parameter dynamic optimization, wherein the decision process model is constructed based on the current cargo volume and the current cargo volume data, the optimal service level weight is input into the decision process model, and a deep network algorithm is used for analysis to output the optimal cargo volume matching decision of the exporter, specifically includes:

[0050] Construct a reward function based on the current cargo volume and the current cargo volume data:

[0051] ;

[0052] in, represents the reward function, Indicates that the freight object is Periodic cargo volume data, It represents the basic income brought by the freight object completing one unit of transportation. It represents the delay cost that the exporter needs to pay for the delayed goods order. The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, express The probability density function of Express obedience The probability distribution function of the uniform distribution on ;

[0053] A decision process model is constructed, the reward function, the state transition equation, and the optimal service level weight are input into the decision process model, and a deep network algorithm is used for calculation to determine and output the optimal cargo volume matching decision that maximizes the freight efficiency of the exporter within a large period in the decision space:

[0054] ;

[0055] ;

[0056] in, represents the optimal quantity matching decision in the decision space, represents a function that seeks the maximum value of an independent variable. Indicates that the exporter is from the current state Start Value, take action And follow a certain cargo matching decision The expected cumulative reward after Indicates the current state of the exporter. represents a cargo matching decision in the decision space, Indicates that the freight object follows the cargo matching decision The expected return, Indicates future accumulated rewards, Indicates the The status of the cycle export side, Indicates the The optimal cargo matching decision taken by the exporter during the period;

[0057] The large cycle represents all single cycles before the current cycle. According to the optimal cargo matching decision, the current reward of the exporter in the current cycle and the future state in the next cycle are obtained. According to the current reward and the future state, the deep network algorithm is updated to obtain the target deep network algorithm.

[0058] ;

[0059] ;

[0060] in, Represents the goal of the target deep network algorithm value, represents the current reward of the current cycle, A discount factor that indicates the relative importance of the reward, represents an updated copy of the deep network algorithm, representing the state Start Value, take action And follow a certain cargo matching decision The expected cumulative reward after represents the future state of the exporter, represents the updated deep network algorithm, Indicates the target Value and export side from current state Start The mean squared error between the values.

[0061] In addition, to achieve the above-mentioned purpose, the present invention further provides a cargo quantity matching system based on digital multi-parameter dynamic optimization, wherein the cargo quantity matching system based on digital multi-parameter dynamic optimization includes:

[0062] A data acquisition module is configured to acquire freight information of multiple freight objects within multiple single cycles and the current cargo volume of the exporter in the current cycle, use the freight information to train the constructed long short-term memory network model, input the current cargo volume into the long short-term memory network model, and output the current cargo volume data of all freight objects in the current cycle;

[0063] A weight acquisition module is used to input the current cargo volume data into the constructed cargo volume matching theoretical model, analyze the current cargo volume data, and output the optimal service level weight of the exporter in the current period;

[0064] A decision determination module is used to construct a decision process model based on the current cargo volume and the current cargo volume data, input the optimal service level weight into the decision process model, use a deep network algorithm to perform analysis, and output the optimal cargo volume matching decision of the exporter.

[0065] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a cargo quantity matching program based on digital multi-parameter dynamic optimization stored in the memory and runnable on the processor, and when the cargo quantity matching program based on digital multi-parameter dynamic optimization is executed by the processor, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described above are implemented.

[0066] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a cargo quantity matching program based on digital multi-parameter dynamic optimization, and when the cargo quantity matching program based on digital multi-parameter dynamic optimization is executed by the processor, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described above are implemented.

[0067] In the present invention, freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle are obtained, and a constructed long short-term memory network model is trained using the freight information. The current cargo volume is input into the trained long short-term memory network model, and the current cargo volume data of all freight objects in the current cycle is output; the current cargo volume data is input into a constructed cargo volume matching theoretical model, the current cargo volume data is analyzed, and the optimal service level weight of the exporter in the current cycle is output; a decision process model is constructed based on the current cargo volume and the current cargo volume data, the optimal service level weight is input into the decision process model, the decision process model is analyzed using a deep network algorithm, and the optimal cargo volume matching decision of the exporter is output. The present invention focuses on the instant single-cycle transportation service procurement scenario. For the enterprise's multi-cycle instant transportation orders and uncertain transportation needs, it adopts multi-attribute reverse selection of transportation service providers at each stage, clarifies various influencing factors and adaptively determines the cargo volume deployment and service level weights, and determines the cargo volume within a single cycle. It effectively solves the problem of inflexibility of long-term procurement agreements, avoids the situation of multiple incomplete cargo shipments, reduces procurement costs, and improves the exporter's freight efficiency throughout the entire cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of a preferred embodiment of the cargo quantity matching method based on digital multi-parameter dynamic optimization of the present invention;

[0069] Figure 2 This is a single-stage decision flow chart of a preferred embodiment of the cargo quantity matching method based on digital multi-parameter dynamic optimization of the present invention;

[0070] Figure 3 This is a multi-stage decision flow chart of a preferred embodiment of the cargo quantity matching method based on digital multi-parameter dynamic optimization of the present invention;

[0071] Figure 4 This is a schematic diagram of the principle of a preferred embodiment of the cargo quantity matching system based on digital multi-parameter dynamic optimization of the present invention;

[0072] Figure 5 Schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of the present invention more clear and distinct, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0074] The cargo quantity matching method based on digital multi-parameter dynamic optimization described in the preferred embodiment of the present invention is as follows: Figure 1As shown, the cargo volume matching method based on digital multi-parameter dynamic optimization includes the following steps:

[0075] Step S10: Obtain freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, use the freight information to train the constructed long short-term memory network model, input the current cargo volume into the trained long short-term memory network model, and output the current cargo volume data of all freight objects in the current cycle.

[0076] Among them, the freight information of multiple freight objects in multiple single cycles includes freight information such as the quantity of goods shipped by these freight objects in previous single cycles, transportation efficiency, etc. The current cargo volume of the exporter in the current cycle represents the total amount of goods that need to be transported in the current cycle. The long short-term memory network model is trained and constructed based on the freight information of all freight objects obtained in previous cycles, and the current cargo volume is input into it to output the predicted current cargo volume data that can be transported by all freight objects in the current cycle.

[0077] Specifically, the freight information includes cargo volume data and predicted cargo volume data; the cargo volume data of multiple freight objects in multiple single cycles and the predicted cargo volume data of all freight objects in each cycle are obtained, and an initial long short-term memory network model is constructed; the cargo volume data and the predicted cargo volume data are input into the initial long short-term memory network model for training to obtain a trained long short-term memory network model; the current cargo volume of the exporter in the current cycle is input into the trained long short-term memory network model, and the current cargo volume data of all freight objects in the current cycle is output; wherein the sum of the current cargo volume data of all freight objects is equal to the current cargo volume.

[0078] Among them, in addition to the quantity of goods transported by the freight object and the transportation efficiency in the previous single cycle, the freight information also includes the predicted cargo volume data of the freight object in a certain single cycle (i.e., predicted cargo volume data). Based on the cargo volume data and predicted cargo volume data of all freight objects in each round of single cycle, an initial long-short-term memory network model can be constructed, and the cargo volume data and predicted cargo volume data are used to train the initial long-short-term memory network model to obtain an optimized long-short-term memory network model. The current cargo volume of the exporter in the current cycle is then used to predict the possible transportation cargo volume data of all freight objects in the current cycle, and the current cargo volume data of each freight object in the current cycle is obtained.

[0079] Furthermore, the model parameters of the initial long short-term memory network model are obtained, the cargo volume data and the predicted cargo volume data are input into the initial long short-term memory network model for training, and the difference between the cargo volume data and the predicted cargo volume data is output:

[0080] ;

[0081] in, Indicates the freight object The difference between the forecast volume data and the volume data for the period, Indicates the freight object Forecast cargo volume data for the period, Indicates that the freight object is Periodic cargo volume data, Indicates the freight object In the Forecast cargo volume data for the period, Indicates the freight object In the Periodic cargo volume data, Indicates the quantity of the freight object, Indicates that the exporter The supply of goods in the cycle, Indicates the freight objects; using the gradient descent algorithm to update the model parameters according to the difference to obtain a trained long short-term memory network model:

[0082] ;

[0083] in, represents the updated model parameters, represents the model parameters, It represents the delay cost that the exporter needs to pay for the delayed goods order. represents the gradient of the model parameters.

[0084] During the training of the initial LSTM model, the model parameters are initialized. The predicted and cargo volume data are then input into the model. The model then predicts the predicted cargo volume based on the input cargo volume data. The difference between the technical and predicted cargo volume data is then calculated using a loss function based on the mean squared difference. This loss function is then used to train the initial LSTM model, minimizing the difference between the cargo volume data and the predicted cargo volume data. A backpropagation algorithm is then used to calculate the gradient of the loss with respect to the model parameters, and a gradient descent algorithm is used to update the model parameters, resulting in an optimized LSTM model. Based on historical freight information for freight objects, the LSTM model is used to predict the available capacity (indicates the volume of cargo that a freight object can transport) for each freight object in the current cycle. This allows for simultaneous capacity matching of multiple freight objects within a single cycle based on the exporter's current cargo volume, thereby maximizing cargo transportation efficiency for the exporter.

[0085] Step S20: Input the current cargo volume data into the constructed cargo volume matching theoretical model, analyze the current cargo volume data, and output the optimal service level weight of the exporter in the current period.

[0086] Among them, such as Figure 2 As shown, the cargo volume data of the freight objects obtained in the above steps are sorted out, and a cargo volume matching theoretical model is constructed. The cargo volume matching theoretical model is used to analyze the current cargo volume data of the freight objects, and a multi-attribute winner decision rule with exporter preferences is designed (that is, the exporter selects multiple freight objects according to its own needs to make equilibrium decisions, where the equilibrium decision refers to the freight strategy designed for the freight object). The cargo volume data of the freight objects are used to mine the benefits and efficiency achieved by the freight strategy of each freight object, thereby outputting the optimal service level weight that the exporter can use to achieve the highest freight efficiency in the current cycle.

[0087] Among them, according to the reconstructed exporter's transport capacity process, the entire transport capacity process can be divided into multiple cycles including all cargo transportation sessions. (i.e. the entire cycle ) and a single cycle of each freight transport, Among them, the exporter is in multiple cycles Has total transport capacity demand , in a single cycle The transportation demand within . Each single cycle In the process, the exporter also has a certain amount of basic cargo release , representing its minimum transportation demand, there are , .

[0088] Specifically, the current cargo volume data is extracted and input into the constructed cargo volume matching theoretical model. The optimal transportation service level of each freight object in the current cargo volume data is analyzed in the cargo volume matching theoretical model to obtain the optimal service level weight of the exporter in the current period and output it:

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] in, represents the optimal single item benefit, The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object In the The optimal transport service level for the cycle, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, It represents the basic income brought by the freight object completing one unit of transportation. Express obedience The probability distribution function of the uniform distribution on , Indicates that the freight object is Periodic cargo volume data, express The probability density function of represents the optimal service level weight of the exporter in a single period, represents the cost conversion factor, while the marginal cost of providing additional service level to the freight object increases, , The utility conversion factor that represents the exporter's transportation service level for the freight object.

[0094] Among them, when the transport capacity mechanism is in equilibrium and all freight objects are in equilibrium, the exporter can maximize the freight efficiency of a single item and achieve the maximum freight benefit. Among them, the value range of the optimal service level weight is , and only when the transport capacity supply is infinite, the optimal service level weight value is The single-stage optimal service level weight is equal to the actual weight under the actual utility of the exporter, and its setting is consistent with the current exporter's demand release The optimal value depends mainly on the exporter's utility factor. , freight object cost factor and the current cycle capacity supply For exporters, the current supply of transport capacity Since it is a random variable, it is difficult for the exporter to obtain the specific value of the transport capacity supply. Therefore, when making decisions, the exporter needs to use Estimated value of Therefore, when the exporter uses the weight level parameter to adjust the auction results, regardless of the number of freight objects, the importance it attaches to the utility brought by the service level is always higher than the importance it attaches to the benefits brought by the low price, and the increase It can bring positive expected benefits to the freight objects, thereby attracting more potential freight objects to participate in cargo transportation. For the exporter, it can improve It can also improve the overall revenue and efficiency of its own freight transportation, but If the threshold is exceeded, the exporter's freight efficiency will be reduced. Therefore, continuously increasing the weight of service level is not necessarily more beneficial to the exporter. The exporter needs to adjust the service level according to its current freight demand and multi-cycle In addition, The optimal value of approaches the real utility weight of the exporter as the freight supply increases.

[0095] Furthermore, a bidding standard for the freight object in the cargo volume data within a single period is determined, and a cost function is constructed based on the bidding standard:

[0096] ;

[0097] in, Indicates the freight object Bidding standards, Indicates the freight object In the Bidding standards for the cycle, Indicates the freight object In the The unit transport rate for the cycle, Indicates the freight object In the The transport service level of the cycle;

[0098] ;

[0099] in, Indicates the freight object The cost function, Indicates the freight object The cost conversion coefficient that converts cost into service capacity is: The smaller the freight object The lower the cost, Indicates the basic cost of each freight object;

[0100] Construct a scoring function based on the bidding criteria and the cost function:

[0101] ;

[0102] in, Indicates the freight object Scoring function for a single bid, Indicates that the exporter Service level weight within the cycle;

[0103] In each single cycle The exporter purchases transport capacity through auctions. Before each phase begins, the exporter releases a certain amount of cargo orders and discloses the demand to the freight recipient. and service level weights , the exporter uses the weight parameter To evaluate the service level provided by freight objects within a single cycle (using the scoring function ) for scoring; if the total supply of international shipping capacity in this cycle , select the freight object in descending order according to the scoring function until the current transportation demand is met, and the exporter pays according to the decision of the freight object; if the total supply of international transportation capacity in this cycle is , then select all freight objects, If some of the transport orders cannot be transported within this period, the exporter must pay the cost of the delay for each unit of the order that has not been transported. After that, it will be postponed to the next cycle for transportation, such as Figure 3 As shown, after the single-stage transportation task assignment is completed, it enters the next cycle until the multi-cycle Total transport tasks within Finish.

[0104] Furthermore, if the exporter accepts the cargo A single decision provided , then the profit brought to the exporter by this decision is ,in Additional benefits from the service level provided to freight forwarders, The basic income for completing one unit of transportation, Less than .

[0105] In the period of fluctuating capacity supply, the problem faced by freight forwarders is how to determine the additional service level to provide in order to maximize their own interests. Given the scoring rules and utility function, in the multi-attribute agreement, the additional service level that maximizes the freight forwarder's utility is The choice is independent of the information of other participants, the freight object The utility scoring function will be based only on the utility score disclosed by the exporter Make a decision, that is, for any utility of the exporter, the freight object has an optimal service level corresponding to it , so that under the fixed utility of the exporter, this optimal attribute combination can minimize its own cost, so the utility return function is constructed according to the cost function:

[0106] ;

[0107] ;

[0108] in, represents the utility return function of the freight object, Represents the value function of logistics service level for freight objects, Indicates the logistics service level for freight objects The value function of represents the cost function of the freight object, represents the service level weight of the exporter, Representing the transportation service level of the freight object; obtaining the optimal equilibrium service bidding level of each freight object according to the utility reward function, and determining the optimal decision space of each freight object according to the optimal equilibrium service bidding level.

[0109] Among them, under the given scoring rules and the exporter's scoring function, in the agreement considering the calculation service level, the goal of the exporter to provide additional service level is to complete the cargo transportation with the least possible cargo transportation times, to maximize the efficiency of the exporter's cargo transportation, and thus obtain the largest decision space , thus avoiding the situation of multiple incomplete shipments, reducing procurement costs and improving the exporter's freight efficiency throughout the entire cycle.

[0110] Furthermore, we can draw the following conclusions: (I) Freight objects The optimal service level bid is (II) Freight objects The optimal supply cost is , and the cost conversion coefficient The larger the freight object, the lower the cost. Indicates the freight object The optimal supply cost is:

[0111] Among them, the freight object Transportation service level revenue function With freight objects Cost function Substitute the utility return function and beg The partial derivative of This can prove the above conclusion.

[0112] Among them, freight objects The profit function It can also be written as:

[0113] ;

[0114] in, Indicates the freight object The supply cost, Indicates the freight object The utility rating of Indicates that except for freight objects Utility ratings of freight objects other than The meaning of this is that if the freight object If you want to transport the exporter's goods, a more robust decision is to ensure that the utility scores of all other freight objects are less than ,because It follows is monotonically decreasing, so we have ,in The probability that a value falls within a specified interval.

[0115] Furthermore, if all freight objects adopt the same incrementally differentiable equilibrium bidding strategy , then any freight object The unit shipping rate is:

[0116] ;

[0117] in, .

[0118] Furthermore, based on the number of cycles and the remaining inventory of the exporter, a state transition equation is constructed, and the decision space of the exporter is determined:

[0119] ;

[0120] ;

[0121] in, Indicates that the exporter is in a single cycle The state transfer equation within, Indicates the cycle, represents the total number of auction cycles, Indicates the The inventory quantity of the cycle, Indicates the total inventory quantity. represents the exporter’s decision space, Indicates that the exporter is in a single cycle Transportation needs within Indicates the basic release quantity. Indicates that the freight object is Cargo volume data for the period.

[0122] In each decision cycle, the state of the exporter is determined by the current cycle number. and remaining inventory Composition, forming a two-dimensional vector, according to which the state of the exporter is determined: , reflecting that the exporter needs to make different decisions at different times and at different inventory levels; in each cycle, the exporter's decision and the realization of shipping capacity will affect the status of the next cycle, where the number of cycles will increase as the decision is made, and the remaining inventory will decrease according to the amount of goods that can be shipped in the current cycle. When the shipping volume is greater than the shipping capacity, some goods will be stranded due to capacity constraints. The stranded goods will continue to match the shipping capacity for transportation in subsequent stages. When the shipping volume is less than the shipping capacity, the shipping volume planned for the current cycle can be fully transported.

[0123] Step S30: construct a decision process model based on the current cargo volume and the current cargo volume data, input the optimal service level weight into the decision process model, and use a deep network algorithm to analyze the decision process model to output the exporter's optimal cargo volume matching decision.

[0124] Among them, the current cargo volume (the amount of cargo that the exporter needs to transport) and the current cargo volume data (the amount of cargo to be transported in the decision made by each freight object) in the above data are used to construct a decision process model based on the Markov decision process, and the optimal service level weight obtained in the above steps is input into the decision process model. The decision process model uses a deep network algorithm to analyze the optimal service level weight to output the exporter's optimal cargo volume matching decision.

[0125] Specifically, in the process of determining the transportation decision that maximizes efficiency, a reward function may be constructed based on the current cargo volume and the current cargo volume data:

[0126] ;

[0127] in, represents the reward function, Indicates that the freight object is Periodic cargo volume data, It represents the basic income brought by the freight object completing one unit of transportation. It represents the delay cost that the exporter needs to pay for the delayed goods order. The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, express The probability density function of Express obedience A probability distribution function of uniform distribution on the upper boundary; constructing a decision process model, inputting the reward function, the state transition equation and the optimal service level weight into the decision process model, performing calculations through a deep network algorithm, and determining and outputting the optimal cargo matching decision that maximizes the freight efficiency of the exporter in a large cycle in the decision space:

[0128] ;

[0129] ;

[0130] in, represents the optimal quantity matching decision in the decision space, represents a function that seeks the maximum value of its independent variable. Indicates that the exporter is from the current state Start Value, take action And follow a certain cargo matching decision The expected cumulative reward after Indicates the current state of the exporter. represents a cargo matching decision in the decision space, Indicates that the freight object follows the cargo matching decision The expected return, Indicates future accumulated rewards, Indicates the The status of the cycle export side, Indicates the The optimal cargo matching decision taken by the exporter during the period;

[0131] in, Can be used as update state action and The value of the reward signal is different from using a separate reward signal. (Accumulated rewards) can provide information about the optimal actions, thereby accelerating the learning of deep network algorithms and improving the efficiency of the exporter's optimal cargo matching decisions.

[0132] The large cycle represents all single cycles before the current cycle. According to the optimal cargo matching decision, the current reward of the exporter in the current cycle and the future state in the next cycle are obtained. According to the current reward and the future state, the deep network algorithm is updated to obtain the target deep network algorithm.

[0133] ;

[0134] ;

[0135] in, Represents the goal of the target deep network algorithm value, represents the current reward of the current cycle, A discount factor that indicates the relative importance of the reward, represents an updated copy of the deep network algorithm, representing the state Start Value, take action And follow a certain cargo matching decision The expected cumulative reward after represents the future state of the exporter, represents the updated deep network algorithm, Indicates the target Value and export side from current state Start The mean squared error between the values.

[0136] Among them, after action selection and execution at each time step, the environment returns a reward and the new state This experience The experience is stored in the experience replay pool for subsequent learning. Subsequently, a batch of experiences are randomly selected from the experience replay pool for learning.

[0137] Among them, the deep network algorithm can adjust the function to adapt to the ever-changing transportation situation by constantly observing the transportation status (such as the behavior of freight objects, transportation service levels, etc.). In addition, the deep network algorithm also introduces two technologies, target network and experience replay, to update the optimal bidding strategy. At each time step, the current state is observed. , and according to The greedy strategy chooses an action The core idea of ​​this strategy is to use probability Randomly select an action (that is, the probability of selecting an action is ),by The probability of choosing The action with the largest value.

[0138] Furthermore, the parameters of the deep network algorithm are updated through back propagation and gradient descent. This process is repeated at each time step until a certain termination condition is met, such as reaching the maximum number of training steps or the performance of the deep network algorithm reaches a certain threshold. Through this process, the deep network algorithm can gradually learn and optimize the optimal cargo matching decision, and constantly adapt to changes in the freight environment. The present invention focuses on the immediate single-cycle transportation service procurement scenario. For the enterprise's multi-cycle instant transportation orders and uncertain transportation needs, it adopts multi-attribute reverse selection of transportation service providers at each stage, clarifies the various influencing factors and adaptively determines the cargo delivery and service level weights, and determines the cargo volume within a single cycle. It effectively solves the problem of inflexibility of long-term procurement agreements, avoids the situation of multiple insufficiency of freight, reduces procurement costs, and improves the exporter's freight efficiency throughout the cycle.

[0139] Further, if Figure 4 As shown, based on the above-mentioned cargo quantity matching method based on digital multi-parameter dynamic optimization, the present invention also provides a cargo quantity matching system based on digital multi-parameter dynamic optimization, wherein the cargo quantity matching system based on digital multi-parameter dynamic optimization includes:

[0140] Data acquisition module 51 is used to obtain freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, use the freight information to train the constructed long short-term memory network model, input the current cargo volume into the long short-term memory network model, and output the current cargo volume data of all freight objects in the current cycle;

[0141] The weight acquisition module 52 is used to input the current cargo volume data into the constructed cargo volume matching theoretical model, analyze the current cargo volume data, and output the optimal service level weight of the exporter in the current period;

[0142] The decision determination module 53 is used to construct a decision process model based on the current cargo volume and the current cargo volume data, input the optimal service level weight into the decision process model, use a deep network algorithm to perform analysis, and output the optimal cargo volume matching decision of the exporter.

[0143] Further, if Figure 5 As shown, based on the above-mentioned cargo quantity matching method and system based on digital multi-parameter dynamic optimization, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 5 Only some of the components of the terminal are shown, but it should be understood that implementation of all of the shown components is not required, and more or fewer components may be implemented instead.

[0144] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as the terminal's hard drive or memory. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the terminal. Furthermore, the memory 20 may include both the terminal's internal storage unit and an external storage device. The memory 20 is used to store application software installed on the terminal and various types of data, such as program code of the installed terminal. The memory 20 may also be used to temporarily store data that has been output or is about to be output. In one embodiment, the memory 20 stores a cargo quantity matching program 40 based on digital multi-parameter dynamic optimization. The cargo quantity matching program 40 based on digital multi-parameter dynamic optimization can be executed by the processor 10, thereby implementing the cargo quantity matching method based on digital multi-parameter dynamic optimization in this application.

[0145] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, configured to execute program codes or process data stored in the memory 20, such as executing the cargo quantity matching method based on digital multi-parameter dynamic optimization.

[0146] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. Components 10-30 of the terminal communicate with each other via a system bus.

[0147] In one embodiment, when the processor 10 executes the cargo quantity matching program 40 based on digital multi-parameter dynamic optimization in the memory 20, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described above are implemented.

[0148] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a cargo quantity matching program based on digital multi-parameter dynamic optimization, and when the cargo quantity matching program based on digital multi-parameter dynamic optimization is executed by a processor, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described above are implemented.

[0149] In summary, the present invention provides a cargo volume matching method and related equipment based on digital multi-parameter dynamic optimization, the method comprising: obtaining freight information of multiple freight objects in multiple rounds of single cycles and the current cargo volume of the exporter in the current cycle, using the freight information to train a constructed long short-term memory network model, inputting the current cargo volume into the trained long short-term memory network model, and outputting the current cargo volume data of all freight objects in the current cycle; inputting the current cargo volume data into a constructed cargo volume matching theoretical model, analyzing the current cargo volume data, and outputting the optimal service level weight of the exporter in the current cycle; constructing a decision process model based on the current cargo volume and the current cargo volume data, inputting the optimal service level weight into the decision process model, analyzing the decision process model using a deep network algorithm, and outputting the optimal cargo volume matching decision of the exporter. The present invention focuses on the instant single-cycle transportation service procurement scenario. For the enterprise's multi-cycle instant transportation orders and uncertain transportation needs, it adopts multi-attribute reverse selection of transportation service providers at each stage, clarifies various influencing factors and adaptively determines the cargo volume deployment and service level weights, and determines the cargo volume within a single cycle. It effectively solves the problem of inflexibility of long-term procurement agreements, avoids the situation of multiple incomplete cargo shipments, reduces procurement costs, and improves the exporter's freight efficiency throughout the entire cycle.

[0150] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or terminal comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or terminal comprising the element.

[0151] Of course, those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware (such as a processor, controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium that can be read by a computer. When executed, the program can include the processes in the above-described method embodiments. The computer-readable storage medium can be a memory, a magnetic disk, an optical disk, etc.

[0152] It should be understood that the application of the present invention is not limited to the above examples. For those skilled in the art, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A cargo volume matching method based on digital multi-parameter dynamic optimization, characterized in that: The cargo volume matching method based on digital multi-parameter dynamic optimization includes: Obtaining freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, using the freight information to train the constructed long short-term memory network model, inputting the current cargo volume into the trained long short-term memory network model, and outputting the current cargo volume data of all freight objects in the current cycle; The freight information includes cargo quantity, transportation efficiency, cargo volume data and forecasted cargo volume data; The freight information of multiple freight objects in multiple single cycles includes the freight information of all freight objects shipped in the previous single cycles; Extracting the current cargo volume data and inputting it into the established cargo volume matching theoretical model, analyzing the optimal transportation service level of each freight object in the current cargo volume data in the cargo volume matching theoretical model, obtaining the optimal service level weight of the exporter in the current period and outputting it; Obtaining an optimal equilibrium service bid level for each freight object according to the utility return function, and determining an optimal decision space for each freight object according to the optimal equilibrium service bid level; According to the number of cycles and the remaining inventory of the exporter, a state transition equation is constructed, and the decision space of the exporter is determined; constructing a decision process model based on the current cargo volume and the current cargo volume data, inputting the optimal service level weight into the decision process model, and performing analysis using a deep network algorithm on the decision process model to output an optimal cargo volume matching decision for the exporter; Reversely select transportation service providers at each stage, identify multiple influencing factors, and adaptively determine cargo volume deployment, service level weights, and cargo volume within a single cycle.

2. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 1 is characterized in that: The method includes obtaining freight information of multiple freight objects in multiple single cycles and the current cargo volume of the exporter in the current cycle, using the freight information to train the constructed long short-term memory network model, inputting the current cargo volume into the trained long short-term memory network model, and outputting the current cargo volume data of all freight objects in the current cycle. Specifically, the method includes: Obtain cargo volume data for multiple freight objects in multiple single cycles and predicted cargo volume data for all freight objects in each cycle, and build an initial long-short-term memory network model; Inputting the cargo volume data and the predicted cargo volume data into the initial long short-term memory network model for training to obtain a trained long short-term memory network model; Input the current cargo volume of the exporter in the current period into the trained long short-term memory network model, and output the current cargo volume data of all freight objects in the current period; The sum of the current cargo volume data of all freight objects is equal to the current cargo volume.

3. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 2 is characterized in that: Inputting the cargo volume data and the predicted cargo volume data into the initial long short-term memory network model for training to obtain a trained long short-term memory network model specifically includes: Obtain model parameters of the initial long short-term memory network model, input the cargo volume data and the predicted cargo volume data into the initial long short-term memory network model for training, and output the difference between the cargo volume data and the predicted cargo volume data: ； in, Indicates the freight object The difference between the forecast volume data and the volume data for the period, Indicates the freight object Forecast cargo volume data for the period, Indicates that the freight object is Periodic cargo volume data, Indicates the freight object In the Forecast cargo volume data for the period, Indicates the freight object In the Periodic cargo volume data, Indicates the quantity of the freight object, Indicates the Freight objects; The gradient descent algorithm is used to update the model parameters according to the difference to obtain a trained long short-term memory network model: ； in, represents the updated model parameters, represents the model parameters, It represents the delay cost that the exporter needs to pay for the delayed goods order. represents the gradient of the model parameters.

4. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 1 is characterized in that: Inputting the current cargo volume data into the constructed cargo volume matching theoretical model, analyzing the current cargo volume data, and outputting the exporter's optimal service level weight in the current cycle specifically includes: ； ； ； ； in, represents the optimal single item benefit, The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object In the The optimal transport service level for the cycle, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, It represents the basic income brought by the freight object completing one unit of transportation. Express obedience The probability distribution function of the uniform distribution on , Indicates that the freight object is Periodic cargo volume data, express The probability density function of represents the optimal service level weight of the exporter in a single period, represents the cost conversion factor, The utility conversion factor that represents the exporter's transportation service level for the freight object.

5. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 4 is characterized in that: The step of inputting the current cargo volume data into the constructed cargo volume matching theoretical model, analyzing the current cargo volume data, and outputting the optimal service level weight of the exporter in the current cycle, further includes: Determine the bidding standards of the freight object in the freight volume data within a single period, and construct a cost function based on the transportation service level in the bidding standards: ； in, Indicates the freight object Bidding standards, Indicates the freight object In the Bidding standards for the cycle, Indicates the freight object In the The unit transport rate for the cycle, Indicates the freight object In the The transport service level of the cycle; ； in, Indicates the freight object The cost function, represents the cost conversion factor, Indicates the freight object Cost conversion coefficient that converts cost into service capability, Indicates the basic cost of each freight object; Construct a scoring function based on the transportation service level and the unit transportation rate: ； in, Indicates the freight object Scoring function for a single bid, Indicates that the exporter The service level weight within the cycle, The utility conversion factor that represents the exporter's transport service level for the freight object; Construct the utility reward function based on the cost function: ； ； in, represents the utility return function of the freight object, Represents the value function of logistics service level for freight objects, Indicates the logistics service level for freight objects The value function of represents the cost function of the freight object, represents the service level weight of the exporter, Indicates the transportation service level of the freight object, The cost conversion coefficient that represents the freight object's conversion of cost into service capacity.

6. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 1 is characterized in that: The step of constructing a decision-making process model based on the current cargo volume and the current cargo volume data, inputting the optimal service level weight into the decision-making process model, analyzing the model using a deep network algorithm, and outputting the optimal cargo volume matching decision of the exporter, further includes: ； ； in, Indicates that the exporter is in a single cycle The state transfer equation within, Indicates the cycle, represents the total number of auction cycles, Indicates the The inventory quantity of the cycle, Indicates the total inventory quantity. represents the exporter’s decision space, Indicates that the exporter is in a single cycle Transportation needs within Indicates the basic release quantity. Indicates that the freight object is Cargo volume data for the period.

7. The cargo quantity matching method based on digital multi-parameter dynamic optimization according to claim 6 is characterized in that: The step of constructing a decision-making process model based on the current cargo volume and the current cargo volume data, inputting the optimal service level weight into the decision-making process model, analyzing the model using a deep network algorithm, and outputting the optimal cargo volume matching decision of the exporter specifically includes: Construct a reward function based on the current cargo volume and the current cargo volume data: ； ； in, represents the reward function, Indicates that the freight object is Periodic cargo volume data, It represents the basic income brought by the freight object completing one unit of transportation. It represents the delay cost that the exporter needs to pay for the delayed goods order. The cost conversion coefficient that represents the freight object's conversion of cost into service capacity. express Cost conversion coefficient that converts cost into service capability, represents the maximum cost conversion coefficient, represents the minimum cost conversion coefficient, Indicates the freight object Provide the highest level of service possible, Indicates the freight object In the The highest shipping rate for the period, express The probability density function of Express obedience The probability distribution function of the uniform distribution on , Indicates in The service level weight of the cycle, The utility conversion factor that represents the exporter's transport service level for the freight object, represents the cost conversion factor; A decision-making process model is constructed, the reward function, the state transition equation, and the optimal service level weight are input into the decision-making process model, and a deep network algorithm is used for calculation to determine and output the optimal cargo volume matching decision that maximizes the exporter's freight efficiency over a large period in the exporter's decision space: ； ； in, represents the optimal quantity matching decision in the exporter’s decision space, represents a function that seeks the maximum value of its independent variable. Indicates that the exporter is from the current state Start value, take the optimal cargo matching decision The expected cumulative reward after Indicates the current state of the exporter. represents a cargo matching decision in the exporter’s decision space, Indicates that the freight object follows the cargo matching decision The expected return, Indicates future accumulated rewards, Indicates the The status of the cycle export side, Indicates the The optimal cargo matching decision taken by the exporter during the period; The large cycle represents all single cycles before the current cycle. According to the optimal cargo matching decision, the current reward of the exporter in the current cycle and the future state in the next cycle are obtained. According to the current reward and the future state, the deep network algorithm is updated to obtain the target deep network algorithm. ； ； in, Represents the goal of the target deep network algorithm value, represents the current reward of the current cycle, A discount factor that indicates the relative importance of the reward, represents an updated copy of the deep network algorithm, representing the state Start Value, take a certain quantity matching decision After the expected accumulation of rewards, represents the future state of the exporter, represents the updated deep network algorithm, Indicates the target Value and export side from current state Start The mean squared error between the values.

8. A cargo matching system based on digital multi-parameter dynamic optimization, characterized in that: The cargo quantity matching system based on digital multi-parameter dynamic optimization is applied to the cargo quantity matching method based on digital multi-parameter dynamic optimization according to any one of claims 1 to 7, and the cargo quantity matching system based on digital multi-parameter dynamic optimization includes: A data acquisition module is configured to acquire freight information of multiple freight objects within multiple single cycles and the current cargo volume of the exporter in the current cycle, use the freight information to train the constructed long short-term memory network model, input the current cargo volume into the long short-term memory network model, and output the current cargo volume data of all freight objects in the current cycle; A weight acquisition module is used to input the current cargo volume data into the constructed cargo volume matching theoretical model, analyze the current cargo volume data, and output the optimal service level weight of the exporter in the current period; A decision determination module is used to construct a decision process model based on the current cargo volume and the current cargo volume data, input the optimal service level weight into the decision process model, use a deep network algorithm to perform analysis, and output the optimal cargo volume matching decision of the exporter.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a cargo quantity matching program based on digital multi-parameter dynamic optimization stored in the memory and runnable on the processor. When the cargo quantity matching program based on digital multi-parameter dynamic optimization is executed by the processor, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a cargo quantity matching program based on digital multi-parameter dynamic optimization. When the cargo quantity matching program based on digital multi-parameter dynamic optimization is executed by a processor, the steps of the cargo quantity matching method based on digital multi-parameter dynamic optimization as described in any one of claims 1 to 7 are implemented.

Images