Data processing method and device, electronic equipment and storage medium

Abstract

The invention provides a data processing method and device, electronic equipment and a storage medium, and the method comprises the steps: determining at least one jump route between a starting node and a target node based on jump step lengths of different lengths; in each jump route, at least one node exists between two nodes; path information between at least two nodes in the plurality of intermediate nodes from the starting node to the target node is unknown; the node represents the logistics position of the order; based on known path information between at least two nodes in the at least one jump route, determining path information between any two nodes from a starting node to a target node; wherein the path information comprises at least one of the logistics time, the logistics distance, the transportation mode, the logistics priority, the logistics carrier and the weather between the two nodes.

Description

Technical Field

[0001] This disclosure relates to the field of large model technology, and in particular to a data processing method, apparatus, electronic device and storage medium. Background Technology

[0002] In logistics and transportation, predicting the estimated time of arrival (ETA) is a core issue for improving user satisfaction. Related technologies rely on historical real-world logistics data for time prediction; however, untimely recording or data loss can lead to data loss, rendering the data ineffective. Insufficient historical data also hinders the performance of various algorithms in time prediction. Summary of the Invention

[0003] This disclosure provides a data processing method, apparatus, electronic device, and storage medium to at least solve the above-mentioned technical problems existing in the prior art.

[0004] According to a first aspect of this disclosure, a data processing method is provided, the method comprising:

[0005] Based on different jump step lengths, at least one jump route is determined between the starting node and the target node; in each jump route, there are at least two nodes separated by at least one node; among the multiple intermediate nodes included between the starting node and the target node, the path information between at least two nodes is unknown; the nodes represent the logistics location of the order.

[0006] Based on the known path information between at least two nodes in the at least one jump route, determine the path information between any two nodes from the starting node to the target node;

[0007] The route information includes at least one of the following: logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather between the two nodes.

[0008] In the above scheme, determining at least one jump route from the starting node to the target node based on different jump step lengths includes:

[0009] With a jump step size of 1, the different types of intermediate nodes on at least one path from the starting node to the target node or the starting node are respectively used as the start node, and the node on at least one path that is separated from the starting node by one node is used as the end node, to determine the jump route corresponding to the different types of intermediate nodes on at least one path; in each jump route, the starting node and the end node are directly connected, and the connection relationship between other nodes on the jump route remains unchanged.

[0010] Iterate over the jump step size to determine at least one jump path with a jump step size of 2 to H;

[0011] Where H is less than or equal to the total number of intermediate nodes between the starting node and the target node.

[0012] In the above scheme, determining the path information between any two nodes from the starting node to the target node based on the known path information between at least two nodes in the at least one jump route includes:

[0013] Based on the first graph convolution module, determine the route features of at least one jump route corresponding to the same jump step size;

[0014] The multi-head attention module processes the route features of at least one jump route corresponding to the same jump step length to obtain the fused features of at least one jump route corresponding to the jump step length.

[0015] The fusion features of at least one jump path corresponding to the jump step length are processed based on the second graph convolution module to obtain the hidden layer representation corresponding to the jump step length.

[0016] Based on the hidden layer representations corresponding to all jump steps, the path information between any two nodes from the starting node to the target node is determined.

[0017] In the above scheme, determining the route features of at least one jump route corresponding to the same jump step size based on the first graph convolution module includes performing the following operations on each jump route among the at least one jump route corresponding to the same jump step size:

[0018] Based on the connection relationships between nodes in the jump route, the adjacency matrix corresponding to the jump route is determined; each element in the adjacency matrix represents the connection relationship between the node in the row where the element is located and the node in the column where the element is located in the jump route.

[0019] Based on the known path information between each node in the jump route, determine the feature matrix corresponding to the jump route;

[0020] The adjacency matrix and feature matrix corresponding to each jump route are input into the first graph convolution module, and the output of the first graph convolution module is determined to be the route feature corresponding to the jump route.

[0021] In the above scheme, the process of processing the route features of at least one jump route corresponding to the same jump step length based on the multi-head attention module to obtain the fused features of at least one jump route corresponding to the same jump step length includes performing the following operations on each jump route among the at least one jump route corresponding to the same jump step length:

[0022] The route features and the original feature matrix corresponding to the jump route are input into the multi-head attention module, and the output of the multi-head attention module is determined to be the fused feature corresponding to the jump route.

[0023] The original feature matrix is ​​determined based on the known path information of any two connected nodes between the starting node and the target node.

[0024] In the above scheme, the fusion features of at least one jump path corresponding to the jump step length are processed by the second graph convolution module to obtain the hidden layer representation corresponding to the jump step length, including:

[0025] The fused features of all jump routes corresponding to the jump step size and the original adjacency matrix are input into the second graph convolution module, and the output of the second graph convolution module is determined to be the hidden layer representation corresponding to the jump step size.

[0026] In the original adjacency matrix, each element represents the connection relationship between the node in the row where the element is located and the node in the column where the element is located in at least one path from the starting node to the target node.

[0027] In the above scheme, determining the path information between any two nodes from the starting node to the target node based on the hidden layer representations corresponding to all jump steps includes:

[0028] The hidden layer representations corresponding to all jump steps are input into the multilayer perceptron, and the output of the multilayer perceptron is determined to be the path information between any two nodes from the starting node to the target node.

[0029] According to a second aspect of this disclosure, a data processing apparatus is provided, the apparatus comprising:

[0030] The jump map determination unit is used to determine at least one jump route between the starting node and the target node based on different jump step lengths; in each jump route, there are at least two nodes separated by at least one node; among the multiple intermediate nodes included from the starting node to the target node, the path information between at least two nodes is unknown; the nodes represent the logistics location of the order.

[0031] The prediction unit is used to determine the path information between any two nodes from the starting node to the target node based on the known path information between at least two nodes in the at least one jump route.

[0032] The route information includes at least one of the following: logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather between the two nodes.

[0033] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0034] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the methods of this disclosure.

[0035] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the methods described in this disclosure.

[0036] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0037] The above and other objects, features, and advantages of this disclosure will become readily apparent from the following detailed description of exemplary embodiments, taken in conjunction with the accompanying drawings. Several embodiments of this disclosure are illustrated in the drawings by way of example and not limitation, in which:

[0038] In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0039] Figure 1 This diagram illustrates the missing logistics and transportation data.

[0040] Figure 2 A schematic diagram of a first optional flow of the data processing method provided in this embodiment of the present disclosure is shown;

[0041] Figure 3 A schematic diagram of a second optional flow of the data processing method provided in an embodiment of this disclosure is shown;

[0042] Figure 4 A schematic diagram of a third optional flow of the data processing method provided in this embodiment of the present disclosure is shown;

[0043] Figure 5 A schematic diagram of a fourth optional flow of the data processing method provided in this disclosure embodiment is shown;

[0044] Figure 6 A schematic diagram of a fifth optional process for the data processing method provided in an embodiment of this disclosure is shown;

[0045] Figure 7 A schematic diagram of the data flow of the data processing method provided in an embodiment of this disclosure is shown;

[0046] Figure 8 A schematic diagram of an optional structure of the data processing apparatus provided in an embodiment of this disclosure is shown;

[0047] Figure 9 A schematic diagram of the composition structure of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0048] To make the objectives, features, and advantages of this disclosure more apparent and understandable, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0049] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0050] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this disclosure described herein can be implemented in an order other than that illustrated or described herein.

[0051] Unless otherwise defined, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used in this disclosure is for the purpose of describing embodiments of this disclosure only and is not intended to be limiting of this disclosure.

[0052] It should be understood that in the various embodiments of this disclosure, the sequence number of each implementation process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0053] Before providing a further detailed description of the embodiments of this disclosure, the nouns and terms involved in the embodiments of this disclosure will be explained, and the nouns and terms involved in the embodiments of this disclosure shall be interpreted as follows.

[0054] 1) Node.

[0055] The logistics network can be abstracted as a graph, namely a logistics network graph. A logistics network graph includes multiple nodes, and there are paths for transporting goods between two nodes that can be directly transported via at least one mode of transportation. Each node represents a geographical location, i.e., a transportation node; the edges between nodes represent the paths for directly transporting goods between a pair of nodes.

[0056] The nodes include the logistics nodes corresponding to the order. The starting node represents the node corresponding to the initial location of the object required by the order; the target node represents the node corresponding to the destination location of the logistics of the object required by the order.

[0057] There may be at least one intermediate node between the starting node and the target node, which is the node where the required object stays and is sorted.

[0058] Optionally, starting nodes, ending nodes, and intermediate nodes can be classified into different types based on at least one of the following: node function, city, location, and mode of transportation. For example: first-level nodes, second-level nodes, third-level nodes, and fourth-level nodes.

[0059] First-level nodes include hub-level centers: core hubs of national or regional logistics networks, connecting different modes of transportation (such as air, rail, and road), undertaking large-scale cross-regional cargo distribution functions, serving as transit stations for inter-regional cargo transfer, and integrating goods from different sources; possessing comprehensive functions such as warehousing and logistics information processing. Examples include national-level logistics hubs (such as an international land port in a certain city) and regional aviation hub airports.

[0060] Secondary nodes include distribution centers, which serve as hubs for the collection and distribution of goods in specific regions (such as provinces or cities), connecting primary nodes and lower-level nodes. They are used to receive goods transported from primary nodes, perform preliminary classification according to regions (cities, districts, counties), temporarily store goods and dispatch them to lower-level sorting centers, or send them directly to large customers (such as supermarkets / enterprises).

[0061] Level 3 nodes include sorting centers, which are finely categorized nodes serving cities or specific areas, connecting distribution centers with last-mile delivery. They are used to further sort goods according to delivery addresses (such as streets or communities). Some sorting centers also have simple warehousing functions, handling temporarily delayed goods. Examples include express sorting centers near airports and e-commerce sorting centers within cities.

[0062] Level 4 nodes include last-mile delivery stations, which directly serve consumers. They are used to receive goods from sorting centers and complete final delivery through delivery personnel or smart devices (such as parcel lockers), providing reverse logistics services such as temporary storage and pickup of goods.

[0063] 2) Orders.

[0064] Contracts and documents for ordering goods; including the entire procurement process for finished products, raw materials, fuel, parts, office supplies, and services. An order includes the process of transporting the required items from the originating point to the destination point.

[0065] In related technologies, the goal is to record relevant information about each order's arrival at each node (such as duration, node weather, transportation method, logistics carrier, distance to the previous node, and logistics priority, at least one of these). However, actual recorded data may be missing arrival records for any one node or multiple consecutive nodes. Based on analysis of real logistics business data, at least 1% of the records for each node are lost, and up to 53% are lost. This renders more than half of the historical data unusable due to insufficient completeness, and also prevents various algorithms on the logistics network graph from filling in missing edges or node information, resulting in poor ETA prediction accuracy.

[0066] Figure 1 This diagram illustrates the missing logistics and transportation data.

[0067] like Figure 1 As shown, the path from the starting node GI to the target node X1 includes intermediate nodes AF, F, C, D, E, and P. During the transportation of packages 1 to 5 along the aforementioned path, for package 1, path information between node AF and node F, node F and node C, and path information between node C and node D, and node D and node F are missing. For package 2, path information between node F and node C, node C and node D, and path information between node D and node E, node E and node P, and node P and node X1 are missing. For package 3… For package 1, the path information between node GI and node AF, and between node AF and node F is missing; the path information between node D and node E, and between node E and node P is missing. For package 4, the path information between node AF and node F, and between node F and node C is missing; the path information between node E and node P, and between node P and node X1 is missing. For package 5, the path information between node GI and node AF, and between node AF and node F is missing; the path information between node F and node C, between node C and node D, and between node D and node E is missing; the path information between node E and node P, and between node P and node X1 is missing.

[0068] To address this issue, a common approach is to remove missing data. However, with extremely limited historical data, the remaining data is insufficient to support effective algorithmic model learning. General time-series processing methods use interpolation to fill in missing data. However, in logistics scenarios, the reference value between orders is limited due to differences in time, status, and external environment. Furthermore, the actual node locations of missing parts in logistics transportation routes may not be unique, and simple interpolation methods cannot capture the impact of changes in node locations. Moreover, the nodes in each transportation route vary significantly, and interpolation methods that do not consider node labels cannot adapt to the impact of node type on transportation time.

[0069] In view of the deficiencies existing in the related technologies, this disclosure provides a data processing method to at least solve the above-mentioned or all of the technical problems.

[0070] Figure 2 A schematic diagram of a first alternative flow of the data processing method provided in this disclosure embodiment is shown, and the steps will be described accordingly.

[0071] Step S201: Based on different jump step lengths, determine at least one jump route between the starting node and the target node.

[0072] In some embodiments, the route from the starting node to the target node includes at least one route; in each route, at least one intermediate node is included between the starting node and the target node. Wherein, among the multiple intermediate nodes included between the starting node and the target node, the path information between at least two nodes is unknown; in this embodiment of the disclosure, the nodes include the starting node, the target node, and any one of the at least one intermediate node. Further, the unknown path information between at least two nodes can include unknown path information between the starting node and an intermediate node, unknown path information between any two intermediate nodes, and unknown path information between an intermediate node and the target node.

[0073] In some embodiments, the route information includes at least one of the following: logistics time, logistics distance, mode of transport, logistics priority, logistics carrier, and weather between two nodes.

[0074] The known routing information between at least two nodes can involve two nodes with a direct path, such as a second-level node and a third-level node within the same city, or a third-level node and a fourth-level node within the same administrative region; or two nodes without a direct path, such as two third-level nodes in different cities, where the two third-level nodes need to pass through a second-level node and a first-level node to conduct logistics transportation. Correspondingly, the path information of two nodes without a direct path can be determined based on the path information of two nodes with a direct path included in the path with these two nodes as the starting and ending nodes.

[0075] like Figure 1 In the diagram, there is a direct path between node AF and node F, allowing logistics to be transported directly from node AF to node F via some means of transportation, without any stops or sorting. However, there is no direct path between node AF and node C, requiring a stop and / or sorting at node F. Furthermore, the path information between node AF and node C can be determined based on the path information between node AF and node F, and the path information between node F and node C.

[0076] In some embodiments, the jump step size includes the number of nodes between two connected nodes in at least one path from the starting node to the target node. In each jump path, there are two nodes separated by at least one node. Optionally, the two nodes separated by at least one node are set as the starting node and the ending node, respectively. Optionally, in each jump path, only two nodes are separated by at least one node, while other nodes are directly connected or have direct paths between them.

[0077] In some embodiments, each jump step corresponds to at least one jump path; the starting nodes of the at least one jump path corresponding to the same jump step are different.

[0078] Step S202: Based on the known path information between at least two nodes in the at least one jump route, determine the path information between any two nodes from the starting node to the target node.

[0079] In some embodiments, the carrier implementing the data processing method (hereinafter referred to as the carrier) can directly calculate the path information between two nodes based on known path information; it can also enable the target model to learn the known path information and predict the unknown path information between any two nodes.

[0080] The carrier can be computer programs, electronic circuits, databases, mobile applications, electronic devices, cloud computing platforms, distributed systems, artificial intelligence frameworks, mathematical models, automation tools, and microcontrollers, etc., which are software or hardware capable of implementing algorithms and methods.

[0081] Thus, through the data processing method provided in this embodiment, multiple jump graphs are constructed for different jump step lengths, and the path information between any two nodes from the starting node to the ending node is calculated or learned based on the known path information between two nodes with a direct path in the jump graph, providing complete learning data for subsequent ETA prediction and improving the accuracy of ETA prediction.

[0082] Figure 3 A second alternative flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0083] Step S301: Based on different jump step lengths, using different types of nodes as start nodes, determine at least one jump route between the start node and the target node.

[0084] In some embodiments, the carrier defines at least one route from the starting node to the ending node. Each route connects two nodes via a direct path, meaning that transportation between the two nodes can be achieved using any mode of transport.

[0085] In some embodiments, when the carrier determines each jump step length, at least one jump route is formed by using different types of nodes as start nodes and jumping at that jump step length. Specifically, this includes using the start node as the starting node, determining the node that is a number of jump steps away from the start node in the direction from the start node to the end node as the end node, and directly connecting the start node and the end node while keeping the connection relationships between other nodes in the route unchanged. This route is the jump route.

[0086] In some embodiments, the carrier can have a fixed jump step size. Under this jump step size, using different types of nodes as starting nodes, at least one jump route corresponding to this jump step size is determined. Then, the jump step size is iterated to determine at least one jump route corresponding to different types of nodes as starting nodes under other jump step sizes. Each node type can be identified based on node type (e.g., first-level node, second-level node, third-level node, and fourth-level node) and a character; for example, the identifier for a first-level node is 1.

[0087] Specifically, taking a jump step of 1 as an example; the carrier uses different types of intermediate nodes on at least one route from the starting node to the target node or the starting node as the starting node, and uses a node on at least one route that is one node away from the starting node as the ending node, to determine the jump route corresponding to different types of intermediate nodes on at least one route; in each jump route, the starting node and the ending node are directly connected, and the connection relationship between other nodes on the jump route remains unchanged.

[0088] by Figure 1Taking the route shown as an example, Figure 1 In the route shown, the starting node is GI, the target node is X1, and the route is GI-AF-FCDEP-X1; it is assumed that the type of each node in this route is different.

[0089] The carrier uses a jump step size of 1. Starting with a node of type 1 (assuming GI), the resulting jump route is GI-FCDEP-X1. In this jump route, there is a 1-node gap between the starting node GI and the ending node F, and node AF is "jumped over". Starting with a node of type 2 (assuming AF), the resulting jump route is GI-AF-CDEP-X1. In this jump route, there is a 1-node gap between the starting node AF and the ending node C, and node F is "jumped over". Starting with a node of type 3 (assuming F), the resulting jump route is GI-AF-FDEP-X1. In this jump route, there is a 1-node gap between the starting node F and the ending node D, and node C is "jumped over". And so on, resulting in jump routes GI-AF-FCEP-X1, GI-AF-FCDP-X1, and GI-AF-FCDE-X1. Furthermore, the jump routes corresponding to a jump step length of 1 include: GI-FCDEP-X1, GI-AF-CDEP-X1, GI-AF-FDEP-X1, GI-AF-FCEP-X1, GI-AF-FCDP-X1, and GI-AF-FCDE-X1.

[0090] Furthermore, the carrier iterates through the jump step size to determine at least one jump path corresponding to jump step sizes from 2 to H. Here, H is less than or equal to the total number of intermediate nodes between the starting node and the target node.

[0091] For example, with a jump step length of 2, the corresponding jump routes include: GI-CDEP-X1, GI-AF-DEP-X1, GI-AF-FEP-X1, GI-AF-FCP-X1, and GI-AF-FCD-X1. With a jump step length of 3, the corresponding jump routes include: GI-DEP-X1, GI-AF-EP-X1, GI-AF-FP-X1, and GI-AF-FC-X1.

[0092] Step S302: Based on the known path information between at least two nodes in the at least one jump route, determine the path information between any two nodes from the starting node to the target node.

[0093] In some embodiments, the carrier implementing the data processing method (hereinafter referred to as the carrier) can directly calculate the path information between two nodes based on known path information; it can also enable the target model to learn the known path information and predict the unknown path information between any two nodes.

[0094] Thus, through the data processing method provided in this embodiment, multiple jump graphs are constructed for different jump step lengths, and the path information between any two nodes from the starting node to the ending node is calculated or learned based on the known path information between two nodes with a direct path in the jump graph, providing complete learning data for subsequent ETA prediction and improving the accuracy of ETA prediction.

[0095] Figure 4 A third alternative flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0096] Step S401: Based on different jump step lengths, determine at least one jump route between the starting node and the target node.

[0097] The specific steps of step S401 are the same as those of step S201 or step S301, and will not be repeated here.

[0098] Step S402: Determine the route features of at least one jump route corresponding to the same jump step size based on the first graph convolution module.

[0099] In some embodiments, the carrier processes each jump path corresponding to the same jump step size separately; for each jump path, the route features corresponding to the jump path are obtained based on the first graph convolution module; the jump path and the route features are in one-to-one correspondence.

[0100] Iterate over the jump step size to determine the route features corresponding to at least one jump route with a jump step size of 2 to H.

[0101] The functionality of the first graph convolution module can be implemented based on a graph convolutional network (GCN).

[0102] Step S403: Based on the multi-head attention module, process the route features of at least one jump route corresponding to the same jump step length to obtain the fused features of at least one jump route corresponding to the jump step length.

[0103] In some embodiments, the carrier processes each route feature based on a multi-head attention block to obtain the attention feature corresponding to each route feature; the attention features corresponding to at least one jump route under the same jump step length are fused to obtain the fused feature corresponding to each jump step length. The number of fused features is the same as the number of jump steps.

[0104] Step S404: Based on the second graph convolution module, process the fusion features of at least one jump path corresponding to the jump step length to obtain the hidden layer representation corresponding to the jump step length.

[0105] In some embodiments, the carrier processes the fused features corresponding to each jump step based on the second graph convolution module to obtain the hidden layer representation corresponding to each jump step. The number of hidden layer representations is the same as the number of jump steps.

[0106] The second graph convolution module can be implemented based on a graph convolutional network (GCN).

[0107] Step S405: Based on the hidden layer representations corresponding to all jump steps, determine the path information between any two nodes from the starting node to the target node.

[0108] In some embodiments, the carrier inputs the hidden layer representations corresponding to all jump steps into a multilayer perceptron, and determines the path information between any two nodes from the starting node to the target node based on the output of the multilayer perceptron.

[0109] Thus, through the data processing method provided in this embodiment, the features of each jump route are obtained through the first graph convolution module and the multi-head attention module. The hidden layer representations of multiple jump routes under the same jump step size are learned based on the second graph convolution module. Then, the hidden layer representations of all jump step sizes are learned based on the multilayer perceptron. The path information between any two nodes from the starting node to the target node is obtained. The unknown path information between nodes is learned through the known path information, providing complete learning data for subsequent ETA prediction and improving the accuracy of ETA prediction.

[0110] Figure 5 A fourth optional flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0111] Step S501: Determine at least one route from the starting node to the target node, and the original adjacency matrix and original feature matrix corresponding to each route.

[0112] In some embodiments, there is at least one route between the starting node and the target node; the original adjacency matrix is ​​used to characterize the connection relationships between nodes in the route; each route corresponds to one original adjacency matrix. In the original adjacency matrix, the number of elements in each row and the number of elements in each column are the same, and the same as the number of nodes (including the starting node, the target node, and at least one intermediate node) included in the route. In the original adjacency matrix, any element represents whether the node corresponding to that row and the node corresponding to that column are connected in the direction from the starting node to the target node; if connected, the element value is 1; if not connected, the element value is 0. Let element a in the i-th row and j-th column of the original adjacency matrix... ij For example, it represents whether the i-th node and the j-th node are connected in the direction from the starting node to the target node. If a ij =1, then there exists "ij" on the route. If a ij =0, then there is at least one node between the i-th node and the j-th node on the route.

[0113] by Figure 1 As shown in the example, Figure 1 If the starting node is GI, the target node is X1, and the route is GI-AF-FCDEP-X1, then its original adjacency matrix consists of 8 rows and 8 columns; each row represents the path from node GI to node X1; each column represents the path from node GI to node X1. Its adjacency matrix is:

[0114]

[0115] In the direction from the starting node GI to the target node X1, node GI (first row) is connected to node AF (second column), so the value of the element in the second column of the first row is 1; in the direction from the starting node GI to the target node X1, node AF (second row) is not connected to node GI (first column), so the value of the element in the first column of the second row is 0.

[0116] In some embodiments, the carrier determines the original feature matrix corresponding to each route based on the known path information of any two connected nodes in each route. The original feature matrix includes at least one of the following: logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather between the two nodes.

[0117] In some embodiments, the carrier can determine an original feature matrix based on raw data, which includes nodes, connected preceding and following nodes, and path information between nodes of node type; each row includes an order identifier. For example:

[0118] Order ID Current Node Current Node Type Previous Node Previous Node Type Duration Distance Transportation Xxxx11111 City 1 Airport AF City 3 Factory GI 1 day 100 KM Car Xxxx11111 City 2 Airport F City 1 Airport AF 2 days 3000 KM Airplane Xxxx2222 City 1 Airport AF City 4 Factory GI 1 day 200 KM Car

[0119] Furthermore, the carrier can obtain the original feature matrix by extracting path information containing unknown data from the original data; or it can obtain the original feature matrix by extracting all path information from the original data. Each row in the original feature matrix represents the path corresponding to two connected nodes, and each column represents a type of path information (such as at least one of logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather), for example:

[0120]

[0121] For example, the node type of airport in city 1 is determined to be AF and the identifier is 3; the node type of factory in city 3 is GI and the identifier is 1; the node type of factory in city 4 is GI and the identifier is 4; and the node type of airport in city 2 is F and the identifier is 2. Combining the original data, the distance between node 1 and node 3 (from factory 3 (GI) to airport 1 (AF)) is 100 km; the distance between node 3 and node 2 (from airport 1 (AF) to airport 2 (F)) is 2000 km; and the distance between node 4 and node 3 (from factory 4 (GI) to airport 1 (AF)) is 200 km. Therefore, each row in the original feature matrix represents the path between two nodes, and each column represents the logistics distance.

[0122] Step S502: Based on different jump step lengths, determine at least one jump route between the starting node and the target node.

[0123] The specific steps of step S502 are the same as those of step S201 or step S301, and will not be repeated here.

[0124] Step S503: Determine the adjacency matrix and feature matrix corresponding to each jump route.

[0125] In some embodiments, the following operations are performed for each jump route: the original adjacency matrix is ​​updated with the connection relationships between nodes in the logistics direction of the jump route (i.e., the direction from the starting node to the target node), and the updated original adjacency matrix is ​​determined as the adjacency matrix of the jump route.

[0126] In the adjacency matrix, any element represents whether the node corresponding to the row and the node corresponding to the column are connected in the direction from the starting node to the target node; if connected, the value of the element is 1; if not connected, the value of the element is 0.

[0127] by Figure 1 As shown in the example, the route is GI-AF-FCDEP-X1. Based on a step size of 1 and a starting node of GI, the jump route is determined as GI-FCDEP-X1, and the corresponding adjacency matrix is:

[0128]

[0129] Based on the path information between nodes in each jump route, the feature matrix corresponding to the jump route is determined. Taking the airport in city 1 as the starting node for jumps, the corresponding feature matrix is:

[0130]

[0131] Step S504: Determine the route features of at least one jump route corresponding to the same jump step size based on the first graph convolution module.

[0132] In some embodiments, the carrier inputs the adjacency matrix and feature matrix corresponding to each jump step length into the same first graph convolution module to obtain the route features corresponding to the jump route.

[0133] In some embodiments, the adjacency matrix The adjacency matrix represents the jump path with a jump size of h and a starting node of k. Furthermore, the characteristic matrix... This represents the adjacency matrix corresponding to a jump path with a jump size of h and a starting node of k. h is a positive integer less than H, where H is less than or equal to the total number of intermediate nodes between the starting and target nodes. k is a positive integer less than K, where K is less than or equal to the total number of node types.

[0134] Let's take a jump step size of 1 as an example. Under this jump step size, there are jump routes with a starting node type of 1, jump routes with a starting node type of 2, and so on, up to jump routes with a starting node of type H. The adjacency matrix of the jump route with a jump step size of 1 and a starting node type of 1 is... The adjacency matrix of a jump path with a jump step size of 1 and a starting node type of 2 is: The adjacency matrix of a jump path with a jump step size of 1 and a starting node type of N is:

[0135] In some embodiments, the carrier will use an adjacency matrix and characteristic matrix The inputs are fed into the same first graph convolutional module to obtain the route features corresponding to the jump route with a stride of h and a starting node of k.

[0136] Step S505: Based on the multi-head attention module, process the route features of at least one jump route corresponding to the same jump step length to obtain the fused features of at least one jump route corresponding to the jump step length.

[0137] In some embodiments, the carrier inputs the route features corresponding to the jump route and the original feature matrix of the route corresponding to the jump route into the multi-head attention module, determines that the output of the multi-head attention module is the attention feature corresponding to the jump route, and fuses the attention features corresponding to all jump routes under the same jump step length to obtain the fused feature corresponding to the jump step length.

[0138] Specifically, the carrier will use the adjacency matrix and characteristic matrix The route features obtained by inputting into the first graph convolution module, along with the original feature matrix X corresponding to the route, are input into the multi-layer attention module to obtain the attention features corresponding to the jump route with a step size of h and a starting node of k.

[0139] Step S506: Based on the second graph convolution module, process the fusion features of at least one jump path corresponding to the jump step length to obtain the hidden layer representation corresponding to the jump step length.

[0140] In some embodiments, the carrier inputs the fused features of all jump paths corresponding to the same jump step size and the original adjacency matrix A into the second graph convolution module, and determines that the output of the second graph convolution module is the hidden layer representation Y corresponding to the jump step size. h h represents the jump step size.

[0141] Step S507: Based on the hidden layer representations corresponding to all jump steps, determine the path information between any two nodes from the starting node to the target node.

[0142] In some embodiments, the carrier represents the hidden layers corresponding to all jump steps, i.e., Y1, Y2, ..., Y... H The input is fed into a multilayer perceptron, and the path information between any two nodes from the starting node to the target node is determined based on the output of the multilayer perceptron.

[0143] Thus, through the data processing method provided in this embodiment, the features of each jump route are obtained through the first graph convolution module and the multi-head attention module. The hidden layer representations of multiple jump routes under the same jump step size are learned based on the second graph convolution module. Then, the hidden layer representations of all jump step sizes are learned based on the multilayer perceptron. The path information between any two nodes from the starting node to the target node is obtained. The unknown path information between nodes is learned through the known path information, providing complete learning data for subsequent ETA prediction and improving the accuracy of ETA prediction.

[0144] Figure 6 A schematic diagram of a fifth optional flow of the data processing method provided in this disclosure embodiment is shown. Figure 7This diagram illustrates the data flow of the data processing method provided in an embodiment of the present disclosure, and will be combined with... Figure 6 and Figure 7 Please provide an explanation.

[0145] Step S601: Determine the original feature matrix and original adjacency matrix for each route.

[0146] In some embodiments, if there is only one path between the starting node and the target node, the method for determining the original feature matrix and the original adjacency matrix is ​​the same as in step S501. If there are at least two paths between the starting node and the target node, the method for determining the original feature matrix is ​​the same as in step S501, but the method for determining the original adjacency matrix needs to be adaptively adjusted, specifically including:

[0147] Each route corresponds to an original adjacency matrix. In this matrix, the number of elements in each row and column is the same, and also the same as the number of nodes (including the starting node, the target node, and at least one intermediate node) included in the route. In the original adjacency matrix, any element represents whether the node in that row is connected to the node in that column along the direction from the starting node to the target node; if connected, the element's value is 1; otherwise, it is 0. In the original adjacency matrix, each row and column is arranged according to the path direction (i.e., the direction from the starting node to the target node) and node type, with nodes of the same type adjacent. For example, if there are two routes from the starting node GI to the target node X1: GI-AF1-FCDEP-X1 and GI-AF2-FCDEP-X1, then the elements in each row and column of the original adjacency matrix are: GI, AF1, AF2, F, C, D, E, P, X1. The corresponding original adjacency matrix is:

[0148]

[0149] Step S602: Determine the jump route corresponding to different jump step lengths.

[0150] In some embodiments, the carrier determines a jump route corresponding to different types of intermediate nodes on at least one line, with a given jump step size h, a node of different types on at least one line as the start node, and a node on at least one line that is one node away from the start node as the end node.

[0151] In some embodiments, for different nodes of the same type on the same route, each node corresponds to a jump route. For example, the route from the starting node GI to the target node X1 includes two routes: GI-AF1-FCDEP-X1 and GI-AF2-FCDEP-X1. When determining the jump route with a jump step size of 1 and GI as the starting node, since both nodes AF1 and AF2 are "jumped", only one jump route is corresponding to this case. However, when determining the jump route with a node of type 2 (both nodes AF1 and AF2 are of type 2) as the starting node, two jump routes will be obtained, namely GI-AF1-CDEP-X1 and GI-AF2-CDEP-X1.

[0152] Furthermore, if a route includes multiple nodes of the same type, then each node of that type is used as the starting node to determine the jump route; that is, the number of jump routes corresponds to the number of nodes of that type. For example, in the route GI-AF-FCDEP-X1, assuming that node F and node D are nodes of the same type, when determining the jump route with a jump step size of 1, the jump routes are determined as GI-AF-FDEP-X1 and GI-AF-FCDP-X1 respectively.

[0153] like Figure 7 As shown, the carrier has a fixed jump step length, determines at least one jump route corresponding to that jump step length, iterates the jump step length, and determines at least one jump route corresponding to each jump step length.

[0154] Step S603: Determine the adjacency matrix and feature matrix corresponding to each jump route.

[0155] The specific steps of step S603 are the same as those of step S503, and will not be repeated here.

[0156] In step S601, where there are two routes between the starting node GI and the target node X1: GI-AF1-FCDEP-X1 and GI-AF2-FCDEP-X1, with a jump step size of 1 and GI as the starting node, the two jump routes are identical, both being GI-FCDEP-X1. Therefore, the adjacency matrix is:

[0157]

[0158]

[0159] If we use 1 as the jump step size and a node of type 2 as the starting node, we get two jump routes: GI-AF1-CDEP-X1 and GI-AF2-CDEP-X1. The corresponding adjacency matrix is:

[0160]

[0161] Step S604: Learn the route features of each jump route based on the convolutional layer of the first graph.

[0162] In some embodiments, the adjacency matrix and feature matrix corresponding to the same jump route are input into the same first graph convolutional layer to learn the representation of each node and obtain the route features corresponding to the jump route.

[0163] like Figure 7 As shown, the adjacency matrix corresponding to the jump path with jump step size h=1 and starting from node GI is... and characteristic matrix The input is fed into the first graph convolutional layer to obtain the route features of the jump path; the adjacency matrix corresponding to the jump path with jump stride h=1 and starting node AF is then processed. and characteristic matrix The input is fed into the first convolutional layer to obtain the route features of the jump path. This process is repeated until the adjacency matrix of the jump path with a jump stride h = 1 and starting from node E is obtained. and characteristic matrix The input is fed into the first convolutional layer, in Figure 7 In the case where the jump step size h = 1, the maximum value of k is 5.

[0164] Furthermore, the route characteristics of at least one jump route corresponding to jump step lengths of 2 to H are determined. Figure 7 As shown, H is less than or equal to 6.

[0165] In some alternative embodiments, the first graph convolution module may be an unweighted graph convolution layer.

[0166] Step S605: Based on the multi-head attention module, process the route features and original feature matrix of at least one jump route corresponding to the same jump step length to obtain the fused features of at least one jump route corresponding to the jump step length.

[0167] In some embodiments, the route features and the original feature matrix corresponding to each jump route are input into the multi-head attention module to determine the attention features corresponding to each jump route. The carrier utilizes a multi-head attention approach, where each input (or each head) of the multi-head attention module learns the attention results of the node representations in each jump route and the node representations on the original graph, and then standardizes and fuses them through a weighted approach.

[0168] Specifically, the route features and original feature matrix corresponding to a jump path with a jump step size h=1 and starting node GI are input into the multi-head attention module to obtain the attention features of that jump path; the route features and original feature matrix corresponding to a jump path with a jump step size h=1 and starting node AF are input into the multi-head attention module to obtain the attention features of that jump path; and so on, until the route features and original feature matrix corresponding to a jump path with a jump step size h=1 and starting node E are input into the multi-head attention module to obtain the attention features of that jump path. The attention features of at least one jump path corresponding to a jump step size h=1 are fused to obtain the fused features corresponding to the case of jump step size h=1.

[0169] The jump step size is iterated to obtain the fusion features corresponding to jump step sizes from 2 to H.

[0170] Step S606: Determine the hidden layer representation corresponding to each jump step based on the fusion features and the original adjacency matrix corresponding to each jump step.

[0171] In some embodiments, the carrier inputs the fused features and the original adjacency matrix corresponding to any jump step into the second graph convolution module, and determines that the output of the second graph convolution module is the hidden layer representation corresponding to that jump step.

[0172] In some embodiments, after the carrier acquires the hidden layer representations corresponding to all jump steps, it inputs these hidden layer representations into the prediction input layer of the multilayer perceptron to obtain the prediction result output by the multilayer perceptron. Specifically, this includes: in, Representation of hidden layer representation; W o and b o These are the parameters of the multilayer perceptron; This is the output of the multilayer perceptron.

[0173] Thus, by using the data processing method provided in this embodiment to overcome the problem of data gaps in logistics networks, a jump route is first constructed for each type of node, with the jump step size iteratively increasing. After determining at least one jump route corresponding to each jump step size, the information on the jump route is learned using a first graph convolution module. Then, a multi-head attention module is used to fuse the representations on the jump routes corresponding to all types of nodes with the original feature matrix representation. Finally, a second graph convolution module is used to output the hidden layer representation, thereby predicting the transportation time between every two adjacent nodes in the original graph. Addressing the data gap problem in complex scenarios, this solution innovatively constructs jump routes, transforming the original data into multiple dimensions for learning, improving data utilization, and avoiding errors caused by interpolation methods. A separate jump route is constructed for each node type, avoiding the graph structure disconnection problem that may occur when all nodes jump together, while also strengthening the differences between node types. Utilizing a multi-head attention mechanism, where each node type corresponds to an attention input, the characteristics of each type of node are significantly added to the learning process, avoiding the impact of uneven distribution of different types of nodes. Constructing jump routes with different jump strides helps to link path information that is far apart but highly correlated in the original route, reducing the information diffusion distance limitation and generalization problem caused by graph convolution depth.

[0174] Figure 8 A schematic diagram of an optional structure of the data processing apparatus provided in an embodiment of this disclosure is shown, and the details will be described in terms of each part.

[0175] In some embodiments, the data processing apparatus 700 includes a jump route determination unit 701 and a prediction unit 702.

[0176] The jump route determination unit 701 is used to determine at least one jump route between the starting node and the target node based on different jump step lengths; in each jump route, there are at least two nodes separated by at least one node; among the multiple intermediate nodes included from the starting node to the target node, the path information between at least two nodes is unknown; the node represents the logistics location of the order.

[0177] The prediction unit 702 is used to determine the path information between any two nodes from the starting node to the target node based on the known path information between at least two nodes in the at least one jump route.

[0178] The route information includes at least one of the following: logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather between the two nodes.

[0179] The jump route determination unit 701 is specifically used to determine the jump route corresponding to the different types of intermediate nodes on at least one line, with a jump step size of 1, and taking different types of intermediate nodes on at least one line from the starting node to the target node or the starting node as the starting node, and taking a node on at least one line that is separated from the starting node by one node as the ending node, and in each jump route, the starting node and the ending node are directly connected, and the connection relationship between other nodes on the jump route remains unchanged.

[0180] Iterate over the jump step size to determine at least one jump path with a jump step size of 2 to H;

[0181] Where H is less than or equal to the total number of intermediate nodes between the starting node and the target node.

[0182] The prediction unit 702 is specifically used to determine the route features of at least one jump route corresponding to the same jump step size based on the first graph convolution module.

[0183] The multi-head attention module processes the route features of at least one jump route corresponding to the same jump step length to obtain the fused features of at least one jump route corresponding to the jump step length.

[0184] The fusion features of at least one jump path corresponding to the jump step length are processed based on the second graph convolution module to obtain the hidden layer representation corresponding to the jump step length.

[0185] Based on the hidden layer representations corresponding to all jump steps, the path information between any two nodes from the starting node to the target node is determined.

[0186] The prediction unit 702 is specifically used to perform the following operations for each jump path in at least one jump path corresponding to the same jump step length:

[0187] Based on the connection relationships between nodes in the jump route, the adjacency matrix corresponding to the jump route is determined; each element in the adjacency matrix represents the connection relationship between the node in the row where the element is located and the node in the column where the element is located in the jump route.

[0188] Based on the known path information between each node in the jump route, determine the feature matrix corresponding to the jump route;

[0189] The adjacency matrix and feature matrix corresponding to each jump route are input into the first graph convolution module, and the output of the first graph convolution module is determined to be the route feature corresponding to the jump route.

[0190] The prediction unit 702 is specifically used to perform the following operations for each jump path in at least one jump path corresponding to the same jump step length:

[0191] The route features and the original feature matrix corresponding to the jump route are input into the multi-head attention module, and the output of the multi-head attention module is determined to be the fused feature corresponding to the jump route.

[0192] The original feature matrix is ​​determined based on the known path information of any two connected nodes between the starting node and the target node.

[0193] The prediction unit 702 is specifically used to input the fusion features of all jump routes corresponding to the jump step length and the original adjacency matrix into the second graph convolution module, and determine that the output of the second graph convolution module is the hidden layer representation corresponding to the jump step length.

[0194] In the original adjacency matrix, each element represents the connection relationship between the node in the row where the element is located and the node in the column where the element is located in at least one path from the starting node to the target node.

[0195] The prediction unit 702 is specifically used to input the hidden layer representations corresponding to all jump steps into the multilayer perceptron, and determine that the output of the multilayer perceptron is the path information between any two nodes from the starting node to the target node.

[0196] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0197] Figure 9 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0198] like Figure 9As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0199] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0200] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as data processing methods. For example, in some embodiments, the data processing method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the data processing method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform data processing methods by any other suitable means (e.g., by means of firmware).

[0201] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0202] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0203] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0204] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0205] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0206] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0207] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0208] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0209] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A data processing method, the method comprising: determining at least one hop route between a start node and a target node based on hop steps of different lengths; in each hop route, there is an interval of at least one node between two nodes; in the plurality of intermediate nodes comprised between the start node and the target node, there is at least one path information unknown between two nodes; the nodes represent logistics locations of orders; determining path information between any two nodes in the start node and the target node based on known path information between at least two nodes in the at least one hop route; wherein the path information comprises at least one of logistics time, logistics distance, transportation mode, logistics priority, logistics carrier and weather between two nodes.

2. The method of claim 1, wherein the determining at least one hop route between the start node and the target node based on hop steps of different lengths comprises: determining a hop route corresponding to each type of intermediate node on at least one route between the start node and the target node, taking the start node and the target node as start nodes respectively, and taking a node on the at least one route which is one node away from the start node as an end node, with a hop step of 1; in each hop route, the start node and the end node are directly connected, and the connection relationship between other nodes on the hop route is unchanged; iterating the hop step to determine at least one hop route with a hop step of 2 to H; wherein H is less than or equal to the number of all intermediate nodes between the start node and the target node.

3. The method of claim 1, wherein the determining path information between any two nodes in the start node and the target node based on known path information between at least two nodes in the at least one hop route comprises: determining route features of at least one hop route corresponding to a same hop step based on a first graph convolution module; processing the route features of the at least one hop route corresponding to the same hop step based on a multi-head attention module to obtain fusion features of the at least one hop route corresponding to the same hop step; processing the fusion features of the at least one hop route corresponding to the same hop step based on a second graph convolution module to obtain hidden layer representations corresponding to the same hop step; determining path information between any two nodes in the start node and the target node based on hidden layer representations corresponding to all hop steps.

4. The method of claim 3, wherein the determining route features of at least one hop route corresponding to a same hop step based on a first graph convolution module comprises performing the following operations on each hop route in the at least one hop route corresponding to the same hop step: determining an adjacency matrix corresponding to the hop route based on connection relationships between nodes in the hop route; each element in the adjacency matrix represents a connection relationship between a node corresponding to a row of the element and a node corresponding to a column of the element in the hop route; determining a feature matrix corresponding to the hop route based on known path information between nodes in the hop route. The adjacency matrix and the feature matrix corresponding to each of the jump routes are input into the first graph convolution module, and an output of the first graph convolution module is determined as a route feature corresponding to the jump route.

5. The method of claim 3, wherein the same jump step corresponds to at least one jump route, and the route feature of the at least one jump route is processed by the multi-head attention module to obtain a fusion feature of the at least one jump route corresponding to the same jump step, including performing the following operations on each of the at least one jump route corresponding to the same jump step: The route feature corresponding to the jump route and the original feature matrix are input into the multi-head attention module, and an output of the multi-head attention module is determined as the fusion feature corresponding to the jump route. wherein The original feature matrix is determined based on known path information of any two nodes connected between the start node and the target node.

6. The method of claim 3, wherein the fusion feature of the at least one jump route corresponding to the same jump step is processed by the second graph convolution module to obtain a hidden layer representation corresponding to the same jump step, including: The fusion features of all the jump routes corresponding to the same jump step and an original adjacency matrix are input into the second graph convolution module, and an output of the second graph convolution module is determined as the hidden layer representation corresponding to the same jump step. Each element in the original adjacency matrix represents a connection relationship between a node corresponding to a row of the element and a node corresponding to a column of the element in at least one path between the start node and the target node.

7. The method of claim 3, wherein the hidden layer representation corresponding to all the jump steps is used to determine path information between any two nodes between the start node and the target node, including: The hidden layer representation corresponding to all the jump steps is input into a multi-layer perception, and an output of the multi-layer perception is determined as the path information between any two nodes between the start node and the target node.

8. A data processing apparatus, comprising: a jump route determination unit configured to determine at least one jump route between a start node and a target node based on different lengths of jump steps; In each of the jump routes, there are at least two nodes with at least one node in between; the path information between at least two nodes among a plurality of intermediate nodes between the start node and the target node is unknown; and the nodes represent logistics locations of orders; a prediction unit configured to determine path information between any two nodes between the start node and the target node based on known path information between at least two nodes in the at least one jump route; The path information includes at least one of logistics time, logistics distance, transportation mode, logistics priority, logistics carrier, and weather.

9. An electronic device, comprising: comprising: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium having stored thereon computer instructions, wherein, The computer instructions are for causing a computer to perform the method according to any one of claims 1-7. The computer instructions are for causing a computer to perform the method according to any one of claims 1-7.

Images