Logistics density prediction method, device, equipment and medium

By constructing a logistics distribution map and predicting the logistics density value, the problems of delivery delay and resource waste on food delivery platforms during peak periods are solved, and the accuracy and user experience of logistics density prediction are improved.

CN118982114APending Publication Date: 2024-11-19SHANGHAI FENGNIAO JIPEI INFORMATION TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411062995.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The existing technology is difficult to accurately predict logistics density, resulting in delays in delivery and waste of resources on takeaway platforms during peak hours.

Method used

By obtaining the logistics of the target area, performing a diagram and a global logistics diagram, extracting node information, building a logistics distribution map, predicting the logistics density value based on the distribution map, adjusting the prediction model to adapt to the actual situation, and scientifically formulating logistics planning and scheduling plans.

Benefits of technology

Improve the accuracy of logistics density prediction, ensure smooth delivery of orders during peak periods, improve user experience and optimize operational costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982114A_ABST
    Figure CN118982114A_ABST
Patent Text Reader

Abstract

The invention provides a logistics density prediction method and device, equipment and a medium, the method is applied to the technical field of computers, and the method comprises the steps: obtaining a logistics proceeding graph and a logistics global graph of a target area; performing node information extraction on the logistics proceeding graph to obtain proceeding graph node information corresponding to the logistics proceeding graph; performing node information extraction on the logistics global graph to obtain global graph node information corresponding to the logistics global graph; based on the proceeding graph node information, the global graph node information and the logistics global graph, constructing a logistics distribution graph of the target area in a target time period, the target time period being a second time period after the current time point; and based on the logistics distribution map, obtaining a logistics density value of the target area in the target time period, and performing logistics density prediction based on the logistics density value. The method can improve the accuracy of logistics density prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more specifically, to a logistics density prediction method, device, equipment and medium in the field of computer technology. Background Art

[0002] In the rapidly developing food delivery industry, logistics density prediction has become an indispensable part of improving platform efficiency and service quality. With the growing demand for food delivery services, food delivery platforms are facing logistics challenges. Logistics density not only affects the speed and safety of food from restaurants to consumers, but also has a profound impact on the overall user experience and the operating costs of the platform.

[0003] The logistics supply and demand relationship, as the core of this complex system, directly reflects the dynamic balance between food delivery rider resources and online food order demand. When the order volume surges and the rider resources are relatively insufficient, the logistics density increases sharply, which may lead to delivery delays and food quality damage, thereby damaging the user experience and increasing the risk of user churn. On the contrary, if the rider resources are excessive and the orders are insufficient, it will cause a waste of resources and push up the platform's operating costs.

[0004] Therefore, it is crucial for food delivery platforms to accurately predict logistics density in order to rationally allocate rider resources and ensure efficient delivery services even during peak hours. Summary of the invention

[0005] The present application provides a logistics density prediction method, device, equipment and medium, which can improve the accuracy of logistics density prediction.

[0006] In a first aspect, a method for predicting logistics density is provided, the method comprising:

[0007] Obtain the logistics progress map and logistics global map of the target area. The logistics progress map is used to represent the logistics information of the items at the current time point, and the logistics global map is used to represent the logistics information of the items in the historical time period. The historical time period is the first time period before the current time point;

[0008] Extract node information from the logistics graph to obtain the graph node information corresponding to the logistics graph, and the graph node information represents the order information corresponding to each graph node in the logistics graph; extract node information from the logistics global graph to obtain the global graph node information corresponding to the logistics global graph, and the global graph node information represents the order information corresponding to each global graph node in the logistics global graph;

[0009] Based on the ongoing graph node information, the global graph node information and the logistics global graph, a logistics distribution graph of the target area within the target time period is constructed, where the target time period is the second time period after the current time point;

[0010] Based on the logistics distribution map, the logistics density value of the target area within the target time period is obtained to make a logistics density prediction based on the logistics density value.

[0011] In a second aspect, a logistics density prediction device is provided, the device comprising:

[0012] A graph acquisition unit, used to acquire a logistics progress graph and a logistics global graph of a target area, wherein the logistics progress graph is used to represent the logistics information of items at a current time point, and the logistics global graph is used to represent the logistics information of items within a historical time period, wherein the historical time period is the first time period before the current time point;

[0013] The first information extraction unit is used to extract the node information of the logistics graph to obtain the graph node information corresponding to the logistics graph, and the graph node information represents the order information corresponding to each graph node in the logistics graph;

[0014] The second information extraction unit is used to extract node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, where the global graph node information represents order information corresponding to each global graph node in the logistics global graph;

[0015] A graph construction unit, for constructing a logistics distribution graph of a target area within a target time period based on the graph node information, the global graph node information and the logistics global graph, where the target time period is a second time period after the current time point;

[0016] The density acquisition unit is used to acquire the logistics density value of the target area within the target time period based on the logistics distribution map, so as to make logistics density prediction based on the logistics density value.

[0017] According to a third aspect, an electronic device is provided, including:

[0018] A memory for storing executable program codes;

[0019] The processor is used to call and run the executable program code from the memory, so that the electronic device executes the method in the above-mentioned first aspect or any possible implementation manner of the first aspect.

[0020] In a fourth aspect, a computer program product is provided, the computer program product comprising: a computer program code, when the computer program code is run on a computer, the computer executes the method in the implementation manner of the above-mentioned first aspect.

[0021] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the implementation manner of the above-mentioned first aspect.

[0022] In an embodiment of the present application, by combining the real-time information of the logistics process map and the historical data of the logistics global map, the operation rules of the logistics system can be understood more comprehensively and deeply, thereby improving the accuracy of future logistics density predictions. This method can capture subtle changes in the logistics system and adjust the prediction model accordingly to better adapt to actual conditions. Based on the predicted logistics density information of the logistics distribution map, logistics planning and scheduling plans can be formulated more scientifically. Adjustments can be made in advance based on the prediction results to ensure smooth delivery of orders during the peak logistics period and improve the user experience of delivery personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a scene schematic diagram of a logistics density prediction method provided in an embodiment of the present application;

[0024] Figure 2 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0025] Figure 3 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0026] Figure 4 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0027] Figure 5 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0028] Figure 6 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0029] Figure 7 It is a flow chart of a method for predicting logistics density provided in an embodiment of the present application;

[0030] Figure 8 It is a structural schematic diagram of a logistics density prediction device provided in an embodiment of the present application;

[0031] Fig. 9 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] The technical solution in the present application will be described clearly and in detail below in conjunction with the accompanying drawings. In the description of the embodiments of the present application, unless otherwise specified, " / " means or, for example, A / B can mean A or B: "and / or" in the text is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, "multiple" means two or more than two.

[0033] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features.

[0034] Figure 1 It is a scenario diagram of a logistics density prediction method provided in an embodiment of the present application.

[0035] For example, currently density prediction is based on business districts, but business districts may contain many different types of businesses and consumer groups. These internal differences may lead to significant differences in logistics density in different areas of the business district. Therefore, treating the business district as a whole for prediction may mask these internal differences, resulting in inaccurate prediction results.

[0036] Based on the above situation, the present application embodiment proposes a logistics density prediction method, which can be Figure 1The electronic device 100 in the target area obtains the logistics progress map 200 and the logistics global map 300, the logistics progress map 200 is used to represent the item logistics information at the current time point, and the logistics global map 300 is used to represent the item logistics information in the historical time period, and the historical time period is the first time period before the current time point. The node information of the logistics progress map 200 is extracted to obtain the progress map node information corresponding to the logistics progress map 200, and the progress map node information represents the order information corresponding to each progress map node in the logistics progress map 200. The node information of the logistics global map 300 is extracted to obtain the global map node information corresponding to the logistics global map 300, and the global map node information represents the order information corresponding to each global map node in the logistics global map 300. Based on the progress map node information, the global map node information and the logistics global map 300, a logistics distribution map of the target area in the target time period is constructed, and the target time period is the second time period after the current time point. Based on the logistics distribution map, the logistics density value of the target area in the target time period is obtained, so as to predict the logistics density based on the logistics density value. This method can combine the real-time information of the logistics map with the historical data of the global logistics map to more comprehensively and deeply understand the operation rules of the logistics system, thereby improving the accuracy of future logistics density prediction. This method can capture subtle changes in the logistics system and adjust the prediction model accordingly to better adapt to the actual situation. Based on the predicted logistics density information of the logistics distribution map, logistics planning and scheduling plans can be formulated more scientifically. Adjustments can be made in advance based on the prediction results to ensure smooth delivery of orders during the logistics peak period and improve the user experience of delivery personnel.

[0037] based on Figure 1 The scene shown is shown below. Figure 2-Figure 7 , the logistics density prediction method provided in the embodiments of the present application is introduced in detail.

[0038] See also Figure 2 , which is a flow chart of a method for predicting logistics density according to an embodiment of the present application. Figure 2 As shown, the method of the embodiment of the present application may include the following steps S101-S105.

[0039] S101, obtaining a logistics process map and a logistics global map of a target area, wherein the logistics process map is used to represent the logistics information of items at a current time point, and the logistics global map is used to represent the logistics information of items in a historical time period, where the historical time period is the first time period before the current time point;

[0040] The target area refers to a specific geographical area directly related to logistics activities, which can be determined based on the location where logistics density prediction is required.

[0041] Among them, the logistics progress diagram refers to a visual chart built based on the order information and order quantity of the target area at the current time point.

[0042] In one embodiment, a set of orders of the target area at the current time point may be obtained, such as order information, order quantity, order distance, etc. Then, based on the order information, such as order starting point location information and order terminal location information, multiple process graph nodes and multiple process graph edges of the logistics process graph to be constructed are determined. The order information may be determined as the process graph node information of the logistics process graph node, and the order quantity and order distance may be determined as the process graph edge information of the logistics process graph. After the process graph nodes, process graph node information, process graph edges, and process graph edge information of the logistics process graph are determined, the logistics process graph of the target area at the current time point may be constructed.

[0043] Among them, the progress graph node represents a specific geographical area (Area of ​​Interest, AOI). A progress graph node represents an AOI. Each progress graph node has a progress graph node ID. For example, suppose People's Square represents an AOI and Community A represents an AOI. The progress graph node information of the progress graph node may include AOI meal pickup time, AOI order quantity, AOI meal pickup time, AOI order delivery time, progress graph node ID and other information.

[0044] Among them, the progress graph edge refers to the line segment or link connecting two progress graph nodes in the progress graph. These progress graph edges represent the association between progress graph nodes. For example, the progress graph edge information of the progress graph edge can include the number of orders between AOIs, the delivery distance between AOIs, the average delivery time between AOIs, the delivery speed of riders between AOIs, and other information.

[0045] The current time point refers to the specific moment when data is collected. It is a relative concept used to distinguish the past (i.e., historical time period) from the future.

[0046] Among them, the global logistics map refers to a visual chart built based on the historical order information and historical order quantity of the target area within a historical time period.

[0047] In one embodiment, a historical order set of a target area within a historical time period may be obtained, such as historical order information, the number of historical orders, and the like. Then, based on the multiple progress graph nodes obtained in the above embodiment and the historical order information in the historical order set, such as the historical order starting point location information and the historical order end point location information, multiple global graph nodes and multiple global graph edges are determined. The progress graph node information and the historical order information may be determined as the global graph node information of the logistics global graph. The progress graph edge information, the number of historical orders, and the historical order distance may be determined as the global graph edge information of the logistics global graph. After determining the global graph nodes, the global graph node information, the global graph edges, and the global graph edge information, a logistics global graph of the target area within the historical time period may be constructed.

[0048] Among them, the global map node represents a specific AOI. A global map node represents an AOI. Each global map node has a global map node ID.

[0049] The historical time period refers to a specific and known time range. It is located before the current time point. This time range can be of any length and can be determined based on the needs of the analysis and the availability of data. For example, the historical time period can be the past week, month, quarter, etc.

[0050] S102, extracting node information from the logistics graph to obtain graph node information corresponding to the logistics graph, where the graph node information represents order information corresponding to each graph node in the logistics graph;

[0051] Among them, node information extraction refers to identifying and collecting relevant data and information of each process node from the logistics process graph.

[0052] The process graph node information refers to the detailed data and information associated with each process graph node in the logistics process graph. For example, the process graph node information may include the process graph node ID, the order starting point location information at the current time point, the order end point location information at the current time point, the order quantity, and other information.

[0053] In one embodiment, the image information of the logistics process graph can be first extracted to obtain the first node information, the first edge information and the first graph global information of the logistics process graph (for example, the overall structure information of the logistics process graph). Then, based on the first node information, the first edge information and the first graph global information, the number of process graph nodes of the logistics process graph is aggregated into a preset first node number to obtain the aggregate process graph corresponding to the logistics process graph. After that, the node information of the aggregate process graph is extracted to obtain the process graph node information corresponding to the logistics process graph.

[0054] S103, extracting node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, where the global graph node information represents order information corresponding to each global graph node in the logistics global graph;

[0055] Among them, node information extraction refers to identifying and collecting relevant data and information of each global graph node from the logistics global graph.

[0056] Among them, the global graph node refers to the detailed data and information associated with each global graph node in the logistics global graph. For example, the global graph node information of the global graph node may include the AOI pickup time, the average AOI delivery time, the average AOI daily order volume, the AOI delivery time, the average AOI collection time, and the global graph node ID in the historical time period.

[0057] In one embodiment, the global logistics graph may be first extracted for image information to obtain the second node information, the second edge information, and the second graph global information of the global logistics graph (e.g., the overall structural information of the global logistics graph). Based on the second node information, the second edge information, and the second graph global information, the number of global graph nodes of the global logistics graph is aggregated into a preset second node number to obtain an aggregated global graph corresponding to the global logistics graph. Afterwards, the node information of the aggregated global graph is extracted to obtain the global graph node information corresponding to the global logistics graph.

[0058] S104, constructing a logistics distribution map of the target area within a target time period based on the ongoing graph node information, the global graph node information, and the logistics global graph, where the target time period is a second time period after the current time point;

[0059] The target time period refers to a specific time period after the current time point, which is used to predict and plan logistics activities. This time period is set based on the current time point and factors such as logistics demand and trends, and is intended to provide data support for logistics management and decision-making in the future. For example, the target time period can be within 5 minutes, 10 minutes, 1 hour, etc. after the current time point.

[0060] The logistics distribution map refers to a chart used to characterize and predict the distribution of logistics information of items in a target area within a target time period that has not yet arrived.

[0061] In one embodiment, the logistics distribution information of each progress graph node of the logistics progress graph within the historical time period can be generated based on the progress graph node information and the global graph node information. Then the logistics environment information of the target area is obtained. Based on the logistics distribution information and the logistics environment information, the environmental logistics distribution information of each progress graph node of the logistics progress graph is generated. After that, the environmental logistics distribution information is converted into an adjacency matrix for representing the logistics distribution structure of the target area within the target time period. Then, based on the adjacency matrix, multiple target global graph nodes that can be used as distribution graph nodes of the logistics distribution graph are determined from the logistics global graph, and the node information of each target global graph node is obtained. Finally, based on the adjacency matrix, multiple target global graph nodes and the node information of each target global graph node, a logistics distribution graph of the target area within the target time period is generated.

[0062] S105. Based on the logistics distribution map, obtain the logistics density value of the target area within the target time period, and perform logistics density prediction based on the logistics density value.

[0063] The logistics density value is a quantitative indicator used to measure the burden or tension of the logistics system in the target area within the target time period after the current time point. This logistics density value can be calculated based on the distribution map node information in the logistics distribution map.

[0064] In one embodiment, the node information of the logistics distribution map can be extracted to obtain the distribution map node information corresponding to the logistics distribution map. Then, the distribution map node information is pooled to obtain the logistics characteristic information of the logistics distribution map. After the logistics characteristic information is converted into a numerical value, the logistics density value of the target area in the target time period can be obtained.

[0065] In the embodiment of the present application, by combining the real-time information of the logistics process map and the historical data of the logistics global map, the operation rules of the logistics system can be understood more comprehensively and deeply, thereby improving the accuracy of future logistics density predictions. This method can capture subtle changes in the logistics system and adjust the prediction model accordingly to better adapt to the actual situation. Based on the predicted logistics density information of the logistics distribution map, logistics planning and scheduling plans can be formulated more scientifically. Adjustments can be made in advance based on the prediction results to ensure smooth delivery of orders during the peak logistics period and improve the user experience of delivery personnel.

[0066] See also Figure 3-Figure 4 , which is a flow chart of a method for predicting logistics density according to an embodiment of the present application. Figure 3-Figure 4 As shown, the method of the embodiment of the present application may include the following steps S201-S219.

[0067] S201, obtaining an order set of a target area at a current time point, and determining a plurality of progress graph nodes and a plurality of progress graph edges based on the starting point location information and the ending point location information of each order in the order set;

[0068] Among them, the process graph nodes and process graph edges of the logistics process graph to be generated can be created based on the starting point location information and the end point location information of each order in the order set. The process graph node can represent a specific AOI. One process graph node represents one AOI. Each process graph node has a process graph node ID. The process graph edge refers to the line segment or link connecting two process graph nodes in the process graph. These process graph edges represent the association between the process graph nodes.

[0069] S202, determining the order information in the order set as the progress graph node information of the progress graph node, and determining the order quantity and order distance in the order set as the progress graph edge information of the progress graph edge;

[0070] The order information refers to all the data related to the order in the order set, such as order number, user information, item information, order starting point location information, order ending point location information, order time, etc.

[0071] S203, constructing a logistics progress graph of the target area at the current time point based on the progress graph nodes, the progress graph node information, the plurality of progress graph edges, and the progress graph edge information;

[0072] Specifically, after determining the nodes and edges of the process graph, appropriate graph visualization tools, such as graph visualization software (Graphviz), data-driven documents (D3.js), etc., can be used to generate a logistics process graph.

[0073] S204, obtaining a historical order set of the target area within a historical time period, and determining a plurality of global graph nodes and a plurality of global graph edges based on a plurality of ongoing graph nodes and historical starting point position information and historical end point position information of each historical order in the historical order set;

[0074] The process graph nodes and process graph edges of the logistics process graph to be generated can be created based on the starting point location information and the end point location information of each order in the order set. The global graph nodes and global graph edges of the logistics global graph to be generated can be created based on the historical order starting point location information, the historical order end point location information, and multiple process graph nodes in the historical order information.

[0075] S205, determining the graph node information and the historical order information in the historical order set as the global graph node information of the global graph node, and determining the graph edge information, the number of historical orders in the historical order set, and the historical order distance as the global graph edge information of the global graph edge;

[0076] Among them, historical order information refers to all data related to orders in the historical order collection, such as historical order number, user information, item information, historical order starting point location information, historical order end point location information, historical order time, average delivery time of merchants corresponding to historical order starting point location information, average daily order volume of users corresponding to historical order end point location information, average meal collection time of users corresponding to historical order end point location information, etc.

[0077] S206, constructing a logistics global graph of the target area within a historical time period based on multiple global graph nodes, global graph node information, multiple global graph edges, and global graph edge information;

[0078] Specifically, after determining the global graph nodes and global graph edges, appropriate graph visualization tools, such as Graphviz, D3.js, etc., can be used to generate a logistics global graph.

[0079] S207, extracting image information from the logistics graph to obtain first node information, first edge information, and first graph global information of the logistics graph;

[0080] Specifically, a graph neural network can be used to extract the image feature information of the logistics process graph to obtain the first node information, the first edge information and the first graph global information. The first node information refers to the node information of the process graph node of the logistics process graph. For example, the node information here may include the starting position information of the order, the end position information, the production order quantity including the start and end intervals of the order, the pickup and delivery time, etc. The first edge information refers to the edge information of the process graph edge of the logistics process graph. For example, the edge information here may include the number of orders, the delivery distance, the delivery time, etc. The first graph global information refers to the overall information of the logistics process graph, which can characterize the structural relationship of the entire graph of the logistics process graph.

[0081] S208, based on the first node information, the first edge information and the first graph global information, the number of process graph nodes of the logistics process graph is aggregated into a preset first node number, to obtain an aggregated process graph corresponding to the logistics process graph;

[0082] Specifically, a fully connected network can be used to aggregate the nodes of the logistics process graph to aggregate the number of process graph nodes to a preset first number of nodes. Through aggregation, all graphs can be unified to a uniform size, which is convenient for subsequent processing. And the input graphs (for example, logistics process graphs, logistics global graphs) are all aggregated to obtain a uniform size, which can improve versatility and make it applicable to any input.

[0083] It can be understood that in the process of aggregating the logistics graph to obtain the aggregated graph, only a part of the graph nodes are aggregated together according to the preset rules, so that the number of graph nodes is the first number of nodes. However, the graph node information and graph edge information are not lost. This information can be extracted from the graph nodes obtained after aggregation.

[0084] S209, extracting node information of the aggregated graph to obtain graph node information corresponding to the logistics graph;

[0085] Specifically, a graph neural network can be used to extract the node information of the aggregated process graph. Because the process graph node information and edge information are not lost in the process graph node obtained after aggregation, this information can be extracted from the process graph node obtained after aggregation. Therefore, when a graph neural network is used to extract the node information of the aggregated process graph, the process graph node information corresponding to the logistics process graph can be obtained.

[0086] S210, extracting image information from the logistics global graph to obtain second node information, second edge information, and second graph global information of the logistics global graph;

[0087] Specifically, a graph neural network can be used to extract the image feature information of the global graph of logistics to obtain the second node information, the second edge information and the second graph global information. The second node information refers to the node information of the global graph node of the global logistics graph. For example, the node information here may include the starting position information of the order, the end position information, the production order quantity including the starting point and end point interval of the order, the pickup and delivery time, etc. The second edge information refers to the edge information of the global graph of the global logistics graph. For example, the edge information here may include the number of orders, the delivery distance, the delivery time, etc. The second graph global information refers to the overall information of the global graph, which can characterize the structural relationship of the entire graph of the global graph.

[0088] S211, based on the second node information, the second edge information and the second graph global information, aggregating the number of global graph nodes of the logistics global graph into a preset second node number, to obtain an aggregated global graph corresponding to the logistics global graph;

[0089] Specifically, a fully connected network can be used to aggregate the global graph nodes of the logistics global graph to aggregate the number of global graph nodes to a preset first node number. Through aggregation, all graphs can be unified to a uniform size, which is convenient for subsequent processing. And the input graphs (for example, logistics graphs, logistics global graphs) are all aggregated to obtain a uniform size, which can improve versatility and make it applicable to any input.

[0090] It is understandable that in the process of aggregating the logistics global graph to obtain the aggregated global graph, only a part of the global graph nodes are aggregated together according to the preset rules, so that the number of global graph nodes is the second number of nodes. However, there is no information loss for the global graph node information and the global graph edge information. This information can be extracted from the global graph nodes obtained after aggregation.

[0091] The preset number of first nodes and the preset number of second nodes may be the same or different, and the number may be set according to actual needs.

[0092] S212, extracting node information from the aggregated global graph to obtain global graph node information corresponding to the logistics global graph;

[0093] Specifically, a graph neural network can be used to extract node information of the aggregated global graph. Because the global graph node information and edge information are not lost in the process of aggregating the number of global graph nodes, this information can be extracted from the global graph nodes obtained after aggregation. Therefore, when a graph neural network is used to extract node information of the aggregated global graph, the global graph node information corresponding to the logistics global graph can be obtained.

[0094] S213, performing attention processing on the progress graph node information and the global graph node information to obtain the logistics distribution information of each progress graph node of the logistics progress graph in the historical time period;

[0095] Specifically, attention processing refers to a method for processing the logistics distribution information of each progress graph node in the logistics progress graph in the historical time period by inputting the progress graph node information and the global graph node information into the cross attention mechanism layer. Specifically, the encoder of the cross attention mechanism layer can output the first query vector converted from the progress graph node information, and output the first key vector and the first value vector converted from the global graph node information. Then, the first key vector and the first query vector are subjected to a dot product operation to obtain a node weight value. This node weight value can characterize the similarity between the node information of each progress graph node in the logistics progress graph and the node information of each global graph node in the logistics global graph. Then, the node weight value and the first value vector are weighted and summed to obtain node fusion information. The node fusion information is the node comprehensive information obtained by selecting and emphasizing the global graph node information based on similarity. Finally, the node fusion information and the global graph node information are fused to obtain the logistics distribution information of each progress graph node in the logistics progress graph in the historical time period.

[0096] S214, obtaining logistics environment information of the target area, performing attention processing on the logistics distribution information and the logistics environment information, and obtaining environmental logistics distribution information of each process graph node of the logistics process graph;

[0097] Among them, the current weather information, delivery personnel information and traffic information of the target area can be obtained. After the weather information, delivery personnel information and traffic information are integrated, the logistics environment information of the target area can be obtained.

[0098] Among them, attention processing refers to a processing method of inputting the logistics distribution information and logistics environment information into the cross-attention mechanism layer to output the environmental logistics distribution information of each graph node of the logistics graph.

[0099] Specifically, the encoder of the real-time learning layer can output a second query vector converted from the logistics environment information, and output a second key vector and a second value vector converted from the logistics distribution information. Then, a dot product operation is performed on the second key vector and the second query vector to obtain an environmental weight value. This environmental weight value can characterize the degree of influence of the logistics environment information on the logistics distribution information. Then, a weighted sum is performed on the environmental weight value and the second value vector to obtain environmental weighted logistics information. The environmental weighted logistics information is an information representation obtained by adjusting the logistics distribution information based on the logistics environment information. Finally, the environmental weighted logistics information and the logistics distribution information are fused to obtain the environmental logistics distribution information of each process graph node in the logistics process graph.

[0100] The environmental logistics distribution information represents the logistics distribution of each process node in the logistics process diagram under the environment indicated by the logistics environmental information.

[0101] In an embodiment of the present application, through the attention processing mechanism, it is possible to automatically identify and weight important node information and environmental information, thereby improving the accuracy and reliability of logistics distribution information. This mechanism can distinguish the factors and nodes that have a greater impact on logistics distribution, so that the generated logistics distribution map is closer to the actual situation. By introducing logistics environment information (such as traffic conditions, weather conditions, geographical location, etc.) and performing attention processing, the impact of external factors on logistics distribution can be fully considered. This makes the generated logistics distribution map not only reflect the relationship between orders and nodes, but also incorporates the complexity and dynamics of the actual environment, and improves the adaptability to the environment. By generating an adjacency matrix to represent the logistics distribution structure, the connection relationship between nodes and the path of logistics flow can be clearly displayed.

[0102] S215, performing matrix conversion on the environmental logistics distribution information to obtain an adjacency matrix, where the adjacency matrix is ​​used to represent the logistics distribution structure of the target area within the target time period;

[0103] Specifically, in the process of converting the environmental logistics distribution information into an adjacency matrix, the main purpose is to represent the logistics distribution characteristics of each node and its relationship with adjacent nodes in a matrix form, so as to be used for subsequent analysis and visualization of the logistics distribution structure. The relationship between nodes and edges can be defined first. Clarify the specific entity represented by each node, that is, each node of the logistics process graph, such as the AOI representing the starting point of the order and the AOI representing the end point of the order. And determine which nodes may have direct logistics interactions, that is, there are edges between them. This interaction may be based on relationships such as delivery distance and order quantity. Before converting the environmental logistics distribution information into an adjacency matrix, this information needs to be quantified. Next, based on the relationship between nodes and the quantified logistics information, an adjacency matrix is ​​constructed. The adjacency matrix is ​​a two-dimensional array whose rows and columns represent nodes in the graph.

[0104] In one embodiment, an adaptive learning layer (Adaptive Learning) can be used to generate an adjacency matrix of a graph from environmental logistics distribution information.

[0105] S216, determining a plurality of target global graph nodes from the logistics global graph based on the adjacency matrix, and obtaining node information of each target global graph node;

[0106] Specifically, the target global graph node refers to a global graph node that can be used as a distribution graph node of the logistics distribution graph. Specifically, the global graph nodes in the logistics global graph can be determined as the distribution graph nodes of the logistics distribution graph according to the adjacency matrix. For example, assuming that there are 1000 global graph nodes in the logistics global graph, each global graph node is numbered from 1 to 999. Then the adjacency matrix obtained, assuming that there is an order between No. 1 and No. 9, then the position between No. 1 and No. 9 will be marked as 1 in this matrix. Assuming that there is an edge between No. 2 and No. 100, then the position between No. 2 and No. 100 will be marked as 1. If there are two orders, it will be marked as 2. That is, the numbering of the elements in the adjacency matrix corresponds to the numbering of the global graph nodes in the logistics global graph. Therefore, after determining the adjacency matrix, it can be combined with the global graph node with the defined numbering from 1 to 1000 to determine multiple target global graph nodes. Then the node information corresponding to the target global graph node can be obtained.

[0107] The numbers of the global graph nodes and the numbers of the elements in the adjacency matrix may be agreed upon in advance. The adjacency matrix may include the numbers of the global graph nodes of the logistics global graph, and then according to the one-to-one correspondence between the numbers, multiple target global graph nodes may be determined from multiple global graph nodes of the logistics global graph.

[0108] S217, generating a logistics distribution map of the target area within the target time period based on the adjacency matrix, multiple target global graph nodes, and node information of each target global graph node;

[0109] Specifically, multiple target global graph nodes are used as distribution graph nodes of the logistics distribution graph to be generated, and the node information of each target global graph node is used as the node information of the distribution graph node. The adjacency matrix represents the logistics distribution structure of the target area within the target time period, and can represent the connection relationship between the distribution graph nodes. Therefore, it can be used

[0110] S218, extracting node information from the logistics distribution map to obtain distribution map node information corresponding to the logistics distribution map, and performing pooling processing on the distribution map node information to obtain logistics feature information of the logistics distribution map;

[0111] Specifically, the graph neural network can be used to extract the node information of the logistics distribution map. Then the extracted node information of the logistics distribution map is input into the pooling layer for pooling processing to obtain the logistics feature information of the logistics distribution map.

[0112] In an embodiment of the present application, by performing pooling processing on the distribution map node information, more representative logistics feature information can be obtained, while reducing the computational complexity of subsequent processing and improving processing efficiency.

[0113] S219. Perform numerical conversion on the logistics characteristic information to obtain the logistics density value of the target area within the target time period. The logistics density value is used to predict the logistics density of the target area within the target time period.

[0114] Specifically, the fully connected layer (Head layer) can be used to map the logistics feature information to one dimension to obtain the logistics density value of the target area within the target time period.

[0115] See also Figure 5 , is a flow chart of a logistics density prediction method provided in an embodiment of the present application, such as Figure 5 As shown, in the embodiment of the present application, step S213, attention processing is performed on the progress graph node information and the global graph node information to obtain the logistics distribution information of each progress graph node in the logistics progress graph within the historical time period, which may specifically include the following steps S301-S304:

[0116] S301, converting the global graph node information into a first key vector and a first value vector, and converting the graph node information into a first query vector;

[0117] In one embodiment, the global graph node information and the progress graph node information are respectively input into the cross attention mechanism layer (Cross Attention), and the encoder of the cross attention mechanism layer can output the first key vector (assuming K) and the first value vector (assuming V) converted from the global graph node information, and output the first query vector (assuming Q) converted from the progress graph node information. Then, the following formula (1) is used to perform a dot product operation on the first key vector (K) and the first query vector (Q) to obtain a node weight value. This node weight value can characterize the similarity between the node information of each progress graph node in the logistics progress graph and the node information of each global graph node in the logistics global graph. Then, the node weight value and the first value vector (V) are weighted and summed to obtain node fusion information. The node fusion information is the node comprehensive information obtained by selecting and emphasizing the global graph node information based on similarity. Finally, the node fusion information and the global graph node information are fused to obtain the logistics distribution information of each progress graph node in the logistics progress graph in the historical time period.

[0118]

[0119] Among them, softmax is a normalized exponential function. The role of the softmax function is to convert the attention score (i.e., the similarity or importance measure between the query and the key) into a probability distribution, so that each element ranges between (0,1) and the sum of all elements is 1.

[0120] Among them, Q: query vector (Query), represents the current target or query condition. In natural language processing tasks, the query vector is usually related to the state in the decoder.

[0121] Among them, K: key vector (Key), represents the key information in the input data. It is used together with the query vector to calculate the relevance score for each value (Value).

[0122] Among them, V: Value vector (Value), represents the actual content of the input data. Values ​​with high relevance scores will get more attention in the output.

[0123] in, The dimension of the key vector (Key), Scaling is done to prevent the dot product from being too large, so that the gradient remains stable and does not saturate prematurely.

[0124] Where T stands for transpose, which means swapping the rows and columns of a matrix. If "Q" and "K" are matrices, then "QK T " may refer to the result of multiplying the transposed matrices of "Q" and "K".

[0125] S302, performing a dot product operation on the first key vector and the first query vector to obtain a node weight value, where the node weight value is used to characterize the similarity between the node information of each process graph node in the logistics process graph and the node information of each global graph node in the logistics global graph;

[0126] Specifically, the formula (1) shown in the above embodiment can be used to perform a dot product operation on the first key vector and the first query vector. For example, assuming that the first key vector is K and the first query vector is Q, then KQ T Calculation can get the node weight value.

[0127] S303, performing weighted summation on the node weight value and the first value vector to obtain node fusion information, where the node fusion information is node comprehensive information obtained by selecting and emphasizing global graph node information based on similarity;

[0128] Specifically, the formula (1) shown in the above embodiment can be used to perform weighted summation on the node weight value and the first value vector. For example, assuming that the first value vector is V and the node weight value is KQ T , the node fusion information can be obtained through formula (1).

[0129] S304: Fuse the node fusion information and the global graph node information to obtain the logistics distribution information of each process graph node of the logistics process graph within a historical time period.

[0130] Specifically, the node fusion information can be added to the global node information, and the logistics distribution information of each node graph of the logistics process graph in the historical time period can be obtained.

[0131] In an embodiment of the present application, by converting the node information into a vector form (key vector, value vector and query vector), the similarity and relevance between nodes can be efficiently represented and compared in a high-dimensional space. This vectorized representation method not only reduces the redundancy of information, but also speeds up the speed of information processing. The similarity between the first key vector and the first query vector (i.e., the node weight value) is calculated using a dot product operation, which can accurately quantify the similarity between each node in the logistics process graph and each node in the logistics global graph. This similarity calculation provides a reliable basis for subsequent node information fusion. The first value vector is weighted and summed according to the node weight value to obtain node fusion information. This process realizes the intelligent selection and emphasis of the global graph node information, so that the fused information is closer to the actual situation of the logistics process graph. The node fusion information is merged with the global graph node information to obtain the logistics distribution information of each node in the logistics process graph in the historical time period. This process not only retains the richness and comprehensiveness of the global graph node information, but also incorporates the real-time and dynamic nature of the process graph node information, thereby enhancing the accuracy and practicality of the logistics distribution information.

[0132] See also Figure 6 , is a flow chart of a logistics density prediction method provided in an embodiment of the present application, such as Figure 6 As shown, in the embodiment of the present application, step S214, performing attention processing on the logistics distribution information and the logistics environment information to obtain the environmental logistics distribution information of each process node of the logistics process graph, may specifically include the following steps S401-S404:

[0133] S401, converting logistics distribution information into a second key vector and a second value vector, and converting logistics environment information into a second query vector;

[0134] In one embodiment, the logistics distribution information and the logistics environment information may be respectively input into the cross attention mechanism layer (Cross Attention), and the encoder of the cross attention mechanism layer may output a second key vector (assumed to be represented by K) and a second value vector (assumed to be represented by V) converted from the logistics distribution information, as well as output a second query vector (assumed to be represented by Q) converted from the logistics environment information.

[0135] S402, performing a dot product operation on the second key vector and the second query vector to obtain an environment weight value, where the environment weight value is used to characterize the influence of logistics environment information on logistics distribution information;

[0136] Specifically, in the above embodiment, the formula (1) in step S301 of the above embodiment can be used to perform a dot product operation on the second key vector (K) and the second query vector (Q) to obtain an environmental weight value. This environmental weight value is used to characterize the influence of logistics environment information on logistics distribution information.

[0137] S403, performing weighted summation on the environment weight value and the second value vector to obtain environment weighted logistics information, where the environment weighted logistics information is information representation obtained by adjusting the logistics distribution information based on the logistics environment information;

[0138] Specifically, in the above embodiment, the formula (1) of step S301 of the above embodiment can be used to perform weighted summation on the environment weight value and the second value vector (V) to obtain environment weighted logistics information. The environment weighted logistics information is an information representation obtained by adjusting the logistics distribution information based on the logistics environment information.

[0139] S404. The environmental weighted logistics information and the logistics distribution information are integrated to obtain the environmental logistics distribution information of each process node in the logistics process diagram. The environmental logistics distribution information represents the logistics distribution of each process node in the logistics process diagram under the environment indicated by the logistics environmental information.

[0140] Specifically, the environmental weighted logistics information can be added to the logistics distribution information, and the environmental logistics distribution information of each process graph node in the logistics process graph can be obtained.

[0141] In an embodiment of the present application, by converting the logistics environment information into a second query vector and performing a dot product operation with the second key vector of the logistics distribution information, the obtained environmental weight value can accurately quantify the impact of the logistics environment information on the logistics distribution information. The generation of environmental weighted logistics information is based on the adjustment of the logistics distribution information by the logistics environment information. This adjustment enables the logistics distribution information to reflect changes in environmental factors in real time. This process makes full use of advanced technologies such as vectorization and attention mechanisms to achieve deep integration and intelligent processing of logistics information and environmental information. More accurate environmental logistics distribution information can be obtained.

[0142] The following will be combined Figure 7 The logistics density prediction method provided in the embodiments of the present application is described in detail with reference to a specific example.

[0143] Collect the order set of the target area at the current time point (for example, the time point at which the order set is collected) and the historical order set of the target area in the historical time period (for example, within 3 months before the current time point). Draw the logistics process based on the order set at the current time point Figure 1 . Draw the overall logistics picture based on the historical order collection within the historical time period Figure 2. Then the graph neural network extracts the flow Figure 1 The first node information, the first edge information and the first graph global information are obtained by using the linear layer to transform the logistics into Figure 1 The number of nodes in the progress graph is aggregated into the number of first nodes, and the aggregate progress is obtained. Figure 3 Extracting global flows through graph neural networks Figure 2 The second node information, the second edge information and the second graph global information are obtained by using the linear layer based on the second node information, the second edge information and the second graph global information. Figure 2 The number of global graph nodes is aggregated into the second number of nodes, and the aggregated global Figure 4 . Extract the polymer separately Figure 3 and aggregate global Figure 4 Node information of the aggregation progress graph, obtain the progress graph node information corresponding to the aggregation progress graph, and obtain the aggregation global Figure 4 The corresponding global graph node information.

[0144] The graph node information and the global graph node information are input into the cross attention mechanism layer 5, and the encoder of the cross attention mechanism layer can output the first key vector (assuming it is represented by K) and the first value vector (assuming it is represented by V) converted from the global graph node information, and output the first query vector (assuming it is represented by Q) converted from the graph node information. Then, the formula (1) shown in the above embodiment is used to perform a dot product operation on the first key vector (K) and the first query vector (Q) to obtain the node weight value. Then, the node weight value and the first value vector (V) are weighted and summed to obtain the node fusion information. When the node fusion information is added to the global graph node information, the logistics distribution information of each graph node of the logistics graph in the historical time period is obtained.

[0145] Collect the current weather information, delivery personnel information and traffic information of the target area. By fusing the weather information, delivery personnel information and traffic information, the logistics environment information 6 of the target area can be obtained. Then the logistics environment information and logistics distribution information are respectively input into the real-time learning layer 7 to obtain the second query vector converted from the logistics environment information, and the second key vector and the second value vector converted from the logistics distribution information are output. Then the second key vector and the second query vector are dot-producted to obtain the environmental weight value. Then the environmental weight value and the second value vector are weighted summed to obtain the environmental weighted logistics information. Finally, the environmental weighted logistics information is added to the logistics distribution information to obtain the environmental logistics distribution information of each process node in the logistics process graph.

[0146] The environmental logistics distribution information is input into the adaptive learning layer 8, and the adjacency matrix used to generate the logistics distribution map can be output. Figure 2 Determine multiple target global graph nodes from multiple global graph nodes, and obtain the global graph node information corresponding to each target global graph node. Generate the logistics distribution of the target area within the target time period based on the adjacency matrix, multiple target global graph nodes, and each global graph node information Fig. 9 .

[0147] The logistics distribution map is input into the graph neural network layer 10 to extract the node information of the logistics distribution map through the graph neural network layer 10 to obtain the distribution map node information. Then the distribution map node information is input into the pooling layer 11 (pool) for pooling processing to obtain the logistics feature information of the logistics distribution map. After the logistics feature information is input into the fully connected layer 12, the dimension of the logistics feature information is mapped to one dimension through the fully connected layer 12, and the logistics density value 13 of the target area in the target time period can be obtained.

[0148] based on Figure 1 The following is a schematic diagram of the scene. Figure 8 , the logistics density prediction device provided in the embodiment of the present application is introduced in detail. It should be noted that, Figure 8 The logistics density prediction device in the present application is used to execute Figure 2-Figure 7 For the convenience of explanation, only the part related to the embodiment of the present application is shown. For the specific technical details not disclosed, please refer to the present application. Figure 2-Figure 7 The embodiment shown.

[0149] See also Figure 8 , is a schematic diagram of the structure of a logistics density prediction device provided in an embodiment of the present application. Figure 8 As shown, the logistics density prediction device 1 of the embodiment of the present application may include: a graph acquisition unit 11, a first information extraction unit 12, a second information extraction unit 13, a graph construction unit 14 and a density acquisition unit 15.

[0150] A graph acquisition unit 11 is used to acquire a logistics progress graph and a logistics global graph of a target area, wherein the logistics progress graph is used to represent the logistics information of items at a current time point, and the logistics global graph is used to represent the logistics information of items within a historical time period, wherein the historical time period is the first time period before the current time point;

[0151] The first information extraction unit 12 is used to extract the node information of the logistics process graph to obtain the process graph node information corresponding to the logistics process graph, and the process graph node information represents the order information corresponding to each process graph node in the logistics process graph;

[0152] The second information extraction unit 13 is used to extract node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, where the global graph node information represents order information corresponding to each global graph node in the logistics global graph;

[0153] A graph construction unit 14 is used to construct a logistics distribution graph of a target area within a target time period based on the graph node information, the global graph node information and the logistics global graph, where the target time period is a second time period after the current time point;

[0154] The density acquisition unit 15 is used to acquire the logistics density value of the target area within the target time period based on the logistics distribution map, so as to perform logistics density prediction based on the logistics density value.

[0155] Optionally, the graph acquisition unit 11 includes:

[0156] The first acquisition subunit 111 is used to acquire an order set of the target area at the current time point, and determine a plurality of progress graph nodes and a plurality of progress graph edges based on the starting point position information and the ending point position information of each order in the order set;

[0157] A first information determination subunit 112 is used to determine the order information in the order set as the progress graph node information of the progress graph node, and determine the order quantity and order distance in the order set as the progress graph edge information of the progress graph edge;

[0158] The first construction subunit 113 is used to construct a logistics process graph of the target area at the current time point based on the process graph nodes, the process graph node information, the plurality of process graph edges and the process graph edge information;

[0159] The second acquisition subunit 114 is used to acquire a historical order set of the target area within a historical time period, and determine a plurality of global graph nodes and a plurality of global graph edges based on a plurality of ongoing graph nodes and historical starting point position information and historical end point position information of each historical order in the historical order set;

[0160] The second information determination subunit 115 is used to determine the progress graph node information and the historical order information in the historical order set as the global graph node information of the global graph node, and determine the progress graph edge information, the number of historical orders in the historical order set, and the historical order distance as the global graph edge information of the global graph edge;

[0161] The second construction subunit 116 is used to construct a logistics global graph of the target area within a historical time period based on multiple global graph nodes, global graph node information, multiple global graph edges and global graph edge information.

[0162] Optionally, the first information extraction unit 12 includes:

[0163] The first information extraction subunit 121 is used to extract image information from the logistics graph to obtain first node information, first edge information and first graph global information of the logistics graph;

[0164] The first aggregation subunit 122 is used to aggregate the number of process graph nodes of the logistics process graph into a preset first node number based on the first node information, the first edge information and the first graph global information, so as to obtain an aggregated process graph corresponding to the logistics process graph;

[0165] The second information extraction subunit 123 is used to extract the node information of the aggregation graph to obtain the graph node information corresponding to the logistics graph.

[0166] Optionally, the second information extraction unit 13 includes:

[0167] The third information extraction subunit 131 is used to extract image information from the logistics global graph to obtain second node information, second edge information and second graph global information of the logistics global graph;

[0168] The second aggregation subunit 132 is used to aggregate the number of global graph nodes of the logistics global graph into a preset second node number based on the second node information, the second edge information and the second graph global information, so as to obtain an aggregated global graph corresponding to the logistics global graph;

[0169] The fourth information extraction subunit 133 is used to extract node information from the aggregated global graph to obtain global graph node information corresponding to the logistics global graph.

[0170] Optionally, the graph construction unit 14 includes:

[0171] The third information determination subunit 141 is used to perform attention processing on the progress graph node information and the global graph node information to obtain the logistics distribution information of each progress graph node in the historical time period of the logistics progress graph;

[0172] The third acquisition subunit 142 is used to acquire the logistics environment information of the target area, perform attention processing on the logistics distribution information and the logistics environment information, and obtain the environmental logistics distribution information of each process graph node of the logistics process graph;

[0173] The matrix conversion subunit 143 is used to perform matrix conversion on the environmental logistics distribution information to obtain an adjacency matrix, and the adjacency matrix is ​​used to represent the logistics distribution structure of the target area within the target time period;

[0174] A fourth acquisition subunit 144 is used to determine a plurality of target global graph nodes from the logistics global graph based on the adjacency matrix, and to acquire node information of each target global graph node;

[0175] The graph generation subunit 145 is used to generate a logistics distribution graph of a target area within a target time period based on the adjacency matrix, multiple target global graph nodes and node information of each target global graph node.

[0176] Optionally, the third information determination subunit 141 is also used to convert the global graph node information into a first key vector and a first value vector, and convert the progress graph node information into a first query vector; perform a dot product operation on the first key vector and the first query vector to obtain a node weight value, and the node weight value is used to characterize the similarity between the node information of each progress graph node in the logistics progress graph and the node information of each global graph node in the logistics global graph; perform weighted summation on the node weight value and the first value vector to obtain node fusion information, and the node fusion information is the node comprehensive information obtained by selecting and emphasizing the global graph node information based on similarity; fuse the node fusion information and the global graph node information to obtain the logistics distribution information of each progress graph node in the logistics progress graph within a historical time period.

[0177] Optionally, the third acquisition subunit 142 is also used to convert the logistics distribution information into a second key vector and a second value vector, and convert the logistics environment information into a second query vector; perform a dot product operation on the second key vector and the second query vector to obtain an environmental weight value, which is used to characterize the influence of the logistics environment information on the logistics distribution information; perform weighted summation on the environmental weight value and the second value vector to obtain environmental weighted logistics information, which is an information representation obtained by adjusting the logistics distribution information based on the logistics environment information; fuse the environmental weighted logistics information and the logistics distribution information to obtain environmental logistics distribution information for each process node in the logistics process graph, which characterizes the logistics distribution status of each process node in the logistics process graph under the environment indicated by the logistics environment information.

[0178] Optionally, the density acquisition unit 15 includes:

[0179] The fifth information extraction subunit 151 is used to extract node information from the logistics distribution map to obtain distribution map node information corresponding to the logistics distribution map;

[0180] A pooling subunit 152 is used to perform pooling processing on the distribution map node information to obtain logistics feature information of the logistics distribution map;

[0181] The numerical conversion subunit 153 is used to perform numerical conversion processing on the logistics characteristic information to obtain the logistics density value of the target area within the target time period, and the logistics density value is used to predict the logistics density of the target area within the target time period.

[0182] In the embodiments of the present application, the beneficial effects of the solutions are described in the introduction of the method embodiments and will not be repeated here.

[0183] See also Fig. 9 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Fig. 9 As shown, the electronic device 500 includes a processor 501 and a memory 502. The processor 501 is electrically connected to the memory 502.

[0184] The processor 501 is the control center of the electronic device 500 and may include one or more processing cores. The processor 501 uses various interfaces and lines to connect various parts of the entire electronic device 500, and executes various functions and processes data of the electronic device 500 by running or calling computer programs stored in the memory 502, and calling data stored in the memory 502, so as to control the electronic device 500 as a whole. Optionally, the processor 501 can be implemented in at least one hardware form of digital signal processing (DSP), field programmable gate array (FPGA), and programmable logic array (PLA). The processor 501 can integrate one or a combination of CPU, graphics processing unit (GPU), modem, etc. Among them, the CPU mainly processes the operating system, user pages, and applications; the GPU is responsible for rendering and drawing display content; and the modem is used to process wireless communications. It is understandable that the above-mentioned modem may not be integrated into the processor 501, but may be implemented separately through a communication chip.

[0185] The memory 502 can be used to store software programs and modules, and the processor 501 executes various functional applications and data processing by running the computer programs and modules stored in the memory 502. The memory 502 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, a computer program required for at least one function, etc.; the data storage area can store data created according to the use of the electronic device 500, etc.

[0186] In addition, the memory 502 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device. Accordingly, the memory 502 may also include a memory controller to provide the processor 501 with access to the memory 502.

[0187] In this embodiment, the processor 501 in the electronic device 500 loads instructions corresponding to the processes of one or more computer programs into the memory 502 according to the following steps, and the processor 501 runs the computer program stored in the memory 502 to implement various functions, as follows:

[0188] Obtain a logistics process map and a global logistics map of the target area. The logistics process map is used to represent the logistics information of items at the current time point, and the global logistics map is used to represent the logistics information of items in a historical time period, where the historical time period is the first time period before the current time point. Extract node information from the logistics process map to obtain process map node information corresponding to the logistics process map, where the process map node information represents the order information corresponding to each process map node in the logistics process map. Extract node information from the global logistics map to obtain global map node information corresponding to the global logistics map, where the global map node information represents the order information corresponding to each global map node in the global logistics map. Construct a logistics distribution map of the target area within a target time period based on the process map node information, the global map node information and the global logistics map, where the target time period is the second time period after the current time point. Obtain the logistics density value of the target area within the target time period based on the logistics distribution map, so as to predict the logistics density based on the logistics density value.

[0189] Optionally, when the processor 501 executes the acquisition of the logistics progress map and the logistics global map of the target area, the processor 501 specifically executes:

[0190] Obtain a set of orders for a target area at the current time point, and determine multiple progress graph nodes and multiple progress graph edges based on the starting point position information and the ending point position information of each order in the order set; determine the order information in the order set as the progress graph node information of the progress graph node, and determine the number of orders and the order distance in the order set as the progress graph edge information of the progress graph edge; construct a logistics progress graph for the target area at the current time point based on the progress graph node, the progress graph node information, the multiple progress graph edges, and the progress graph edge information; obtain a set of historical orders for the target area within a historical time period, and determine multiple global graph nodes and multiple global graph edges based on multiple progress graph nodes and the historical starting point position information and the historical ending point position information of each historical order in the historical order set; determine the progress graph node information and the historical order information in the historical order set as the global graph node information of the global graph node, and determine the progress graph edge information, the number of historical orders in the historical order set, and the historical order distance as the global graph edge information of the global graph edge; construct a logistics global graph for the target area within the historical time period based on the multiple global graph nodes, the global graph node information, the multiple global graph edges, and the global graph edge information.

[0191] Optionally, when the processor 501 extracts node information of the logistics process graph to obtain the process graph node information corresponding to the logistics process graph, the processor 501 specifically executes:

[0192] Image information of the logistics graph is extracted to obtain the first node information, the first edge information and the first graph global information of the logistics graph; based on the first node information, the first edge information and the first graph global information, the number of graph nodes of the logistics graph is aggregated into a preset first node number to obtain an aggregated graph corresponding to the logistics graph; node information of the aggregated graph is extracted to obtain graph node information corresponding to the logistics graph.

[0193] Optionally, when the processor 501 extracts node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, the processor 501 specifically executes:

[0194] Image information is extracted from the logistics global graph to obtain second node information, second edge information and second graph global information of the logistics global graph; based on the second node information, the second edge information and the second graph global information, the number of global graph nodes of the logistics global graph is aggregated into a preset second node number to obtain an aggregated global graph corresponding to the logistics global graph; node information is extracted from the aggregated global graph to obtain global graph node information corresponding to the logistics global graph.

[0195] Optionally, when the processor 501 constructs a logistics distribution map of a target area within a target time period based on the ongoing graph node information, the global graph node information, and the logistics global graph, the processor 501 specifically executes:

[0196] Perform attention processing on the process graph node information and the global graph node information to obtain the logistics distribution information of each process graph node of the logistics process graph within the historical time period; obtain the logistics environment information of the target area, perform attention processing on the logistics distribution information and the logistics environment information to obtain the environmental logistics distribution information of each process graph node of the logistics process graph; perform matrix transformation on the environmental logistics distribution information to obtain the adjacency matrix, and the adjacency matrix is ​​used to represent the logistics distribution structure of the target area within the target time period; determine multiple target global graph nodes from the logistics global graph based on the adjacency matrix, and obtain the node information of each target global graph node; generate the logistics distribution map of the target area within the target time period based on the adjacency matrix, multiple target global graph nodes and the node information of each target global graph node.

[0197] Optionally, when the processor 501 performs attention processing on the progress graph node information and the global graph node information to obtain the logistics distribution information of each progress graph node in the logistics progress graph within the historical time period, the processor 501 specifically performs:

[0198] The global graph node information is converted into a first key vector and a first value vector, and the process graph node information is converted into a first query vector; a dot product operation is performed on the first key vector and the first query vector to obtain a node weight value, and the node weight value is used to characterize the similarity between the node information of each process graph node in the logistics process graph and the node information of each global graph node in the logistics global graph; a weighted sum is performed on the node weight value and the first value vector to obtain node fusion information, and the node fusion information is the node comprehensive information obtained by selecting and emphasizing the global graph node information based on the similarity; the node fusion information is fused with the global graph node information to obtain the logistics distribution information of each process graph node in the logistics process graph within the historical time period.

[0199] Optionally, when the processor 501 performs attention processing on the logistics distribution information and the logistics environment information to obtain the environment logistics distribution information of each process graph node of the logistics process graph, specifically executes:

[0200] The logistics distribution information is converted into a second key vector and a second value vector, and the logistics environment information is converted into a second query vector; a dot product operation is performed on the second key vector and the second query vector to obtain an environmental weight value, which is used to characterize the influence of the logistics environment information on the logistics distribution information; a weighted sum is performed on the environmental weight value and the second value vector to obtain environmental weighted logistics information, which is an information representation obtained by adjusting the logistics distribution information based on the logistics environment information; the environmental weighted logistics information and the logistics distribution information are integrated to obtain environmental logistics distribution information for each process node in the logistics process graph, which characterizes the logistics distribution situation of each process node in the logistics process graph under the environment indicated by the logistics environment information.

[0201] Optionally, when the processor 501 predicts the logistics density of the target area within the target time period based on the logistics distribution map, it specifically performs:

[0202] Node information is extracted from the logistics distribution map to obtain the distribution map node information corresponding to the logistics distribution map; the distribution map node information is pooled to obtain the logistics characteristic information of the logistics distribution map; the logistics characteristic information is numerically converted to obtain the logistics density value of the target area within the target time period, and the logistics density value is used to predict the logistics density of the target area within the target time period.

[0203] In the embodiments of the present application, the beneficial effects of the solutions are described in the introduction of the method embodiments and will not be repeated here.

[0204] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program code is stored. When the computer program code is executed on a computer, the computer executes the above-mentioned related method steps to implement a logistics density prediction method provided by the above-mentioned embodiment.

[0205] This embodiment also provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute the above-mentioned related steps to implement a logistics density prediction method provided by the above embodiment.

[0206] Among them, the electronic device, computer-readable storage medium, computer program product or chip provided in this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above and will not be repeated here.

[0207] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in the above embodiments are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions. For example, the order information and historical order information involved in this specification are all obtained with full authorization.

[0208] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0209] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0210] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting logistics density, It is characterized in that The method comprises: Obtain a logistics progress map and a logistics global map of the target area, wherein the logistics progress map is used to represent the logistics information of items at the current time point, and the logistics global map is used to represent the logistics information of items in a historical time period, wherein the historical time period is the first time period before the current time point; Extracting node information of the logistics process graph to obtain process graph node information corresponding to the logistics process graph, wherein the process graph node information represents order information corresponding to each process graph node in the logistics process graph; Extracting node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, wherein the global graph node information represents order information corresponding to each global graph node in the logistics global graph; Based on the ongoing graph node information, the global graph node information and the logistics global graph, construct a logistics distribution graph of the target area within a target time period, where the target time period is a second time period after the current time point; Based on the logistics distribution map, the logistics density value of the target area within the target time period is obtained to perform logistics density prediction based on the logistics density value.

2. The method according to claim 1, It is characterized in that The step of obtaining the logistics progress map and the logistics global map of the target area includes: Obtaining an order set of a target area at a current time point, and determining a plurality of progress graph nodes and a plurality of progress graph edges based on the starting point position information and the ending point position information of each order in the order set; Determine the order information in the order set as the progress graph node information of the progress graph node, and determine the order quantity and order distance in the order set as the progress graph edge information of the progress graph edge; Based on the progress graph node, the progress graph node information, the plurality of progress graph edges, and the progress graph edge information, construct a logistics progress graph of the target area at the current time point; Acquire a historical order set of the target area within a historical time period, and determine a plurality of global graph nodes and a plurality of global graph edges based on the plurality of ongoing graph nodes and the historical starting point position information and the historical end point position information of each historical order in the historical order set; Determine the progress graph node information and the historical order information in the historical order set as the global graph node information of the global graph node, and determine the progress graph edge information, the number of historical orders in the historical order set, and the historical order distance as the global graph edge information of the global graph edge; Based on the multiple global graph nodes, the global graph node information, the multiple global graph edges and the global graph edge information, a logistics global graph of the target area within the historical time period is constructed.

3. The method according to claim 1, It is characterized in that The extracting of node information from the logistics graph to obtain graph node information corresponding to the logistics graph includes: Extracting image information from the logistics graph to obtain first node information, first edge information, and first graph global information of the logistics graph; Based on the first node information, the first edge information and the first graph global information, the number of process graph nodes of the logistics process graph is aggregated into a preset first node number to obtain an aggregated process graph corresponding to the logistics process graph; The node information of the aggregate graph is extracted to obtain the graph node information corresponding to the logistics graph.

4. The method according to claim 1, It is characterized in that The constructing a logistics distribution map of the target area within a target time period based on the ongoing graph node information, the global graph node information and the logistics global graph includes: Performing attention processing on the progress graph node information and the global graph node information to obtain logistics distribution information of each progress graph node of the logistics progress graph within the historical time period; Obtaining the logistics environment information of the target area, performing attention processing on the logistics distribution information and the logistics environment information, and obtaining the environmental logistics distribution information of each process graph node of the logistics process graph; Performing matrix conversion on the environmental logistics distribution information to obtain an adjacency matrix, wherein the adjacency matrix is ​​used to represent the logistics distribution structure of the target area within a target time period; Determining a plurality of target global graph nodes from the logistics global graph based on the adjacency matrix, and obtaining node information of each target global graph node; Based on the adjacency matrix, the multiple target global graph nodes and the node information of each target global graph node, a logistics distribution map of the target area within the target time period is generated.

5. The method according to claim 4, It is characterized in that The attention processing of the progress graph node information and the global graph node information to obtain the logistics distribution information of each progress graph node of the logistics progress graph within the historical time period includes: Convert the global graph node information into a first key vector and a first value vector, and convert the progress graph node information into a first query vector; Performing a dot product operation on the first key vector and the first query vector to obtain a node weight value, wherein the node weight value is used to represent the similarity between the node information of each process graph node in the logistics process graph and the node information of each global graph node in the logistics global graph; Performing a weighted summation on the node weight value and the first value vector to obtain node fusion information, wherein the node fusion information is node comprehensive information obtained by selecting and emphasizing the global graph node information based on the similarity; The node fusion information and the global graph node information are fused to obtain the logistics distribution information of each process graph node of the logistics process graph within the historical time period.

6. The method according to claim 4, It is characterized in that The attention processing of the logistics distribution information and the logistics environment information to obtain the environmental logistics distribution information of each process graph node of the logistics process graph includes: Converting the logistics distribution information into a second key vector and a second value vector, and converting the logistics environment information into a second query vector; Performing a dot product operation on the second key vector and the second query vector to obtain an environment weight value, where the environment weight value is used to characterize the influence of the logistics environment information on the logistics distribution information; Performing a weighted summation on the environmental weight value and the second value vector to obtain environmental weighted logistics information, wherein the environmental weighted logistics information is an information representation obtained by adjusting the logistics distribution information based on the logistics environmental information; The environmental weighted logistics information and the logistics distribution information are integrated to obtain the environmental logistics distribution information of each process node in the logistics process diagram, and the environmental logistics distribution information represents the logistics distribution of each process node in the logistics process diagram under the environment indicated by the logistics environmental information.

7. The method according to claim 1, It is characterized in that The predicting of the logistics density of the target area within the target time period based on the logistics distribution map includes: Extracting node information from the logistics distribution map to obtain distribution map node information corresponding to the logistics distribution map; Performing pooling processing on the node information of the distribution map to obtain logistics characteristic information of the logistics distribution map; The logistics characteristic information is numerically converted to obtain a logistics density value of the target area within the target time period, and the logistics density value is used to predict the logistics density of the target area within the target time period.

8. A logistics density prediction device, It is characterized in that The device comprises: A graph acquisition unit, used to acquire a logistics progress graph and a logistics global graph of a target area, wherein the logistics progress graph is used to represent the logistics information of items at a current time point, and the logistics global graph is used to represent the logistics information of items within a historical time period, wherein the historical time period is the first time period before the current time point; A first information extraction unit is used to extract node information from the logistics process graph to obtain process graph node information corresponding to the logistics process graph, wherein the process graph node information represents order information corresponding to each process graph node in the logistics process graph; A second information extraction unit is used to extract node information from the logistics global graph to obtain global graph node information corresponding to the logistics global graph, wherein the global graph node information represents order information corresponding to each global graph node in the logistics global graph; A graph construction unit, configured to construct a logistics distribution graph of the target area within a target time period based on the ongoing graph node information, the global graph node information and the logistics global graph, wherein the target time period is a second time period after the current time point; The density acquisition unit is used to acquire the logistics density value of the target area within the target time period based on the logistics distribution map, so as to perform logistics density prediction based on the logistics density value.

9. An electronic device, It is characterized in that The electronic device comprises: A memory for storing executable program codes; A processor, configured to call and run the executable program code from the memory, so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer storage medium, It is characterized in that The computer storage medium stores a plurality of instructions, which are suitable for being loaded by a processor and executing the steps of the method according to any one of claims 1 to 7.

Citation Information

Cited By

  • Scheduling method and system for collaborative distribution of processing raw materials and finished products

    CN120181540A