Transport time determination method and electronic equipment

Abstract

The invention discloses a transportation time determination method and electronic equipment, and the method comprises the steps: determining a target time-space sub-graph in a target graph based on the logistics feature information of a to-be-predicted target order; the target graph comprises at least one time-space sub-graph, each time-space sub-graph has an order corresponding to the time-space sub-graph, and the order represents a backtracking dimension of historical order data contained in the time-space sub-graph; determining a causal edge target weight of a causal edge matched with the target order in each target time-space sub-graph based on the target time-space sub-graph, and an order fusion weight corresponding to the target time-space sub-graph; and determining the target transportation time of the target order based on the reference transportation time parameter corresponding to the causal edge matched with the target order, the causal edge target weight and the order fusion weight.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and more specifically to a method for determining transportation time and an electronic device. Background Art

[0002] In the global logistics and transportation business, accurately predicting the arrival time of shipping orders is key to improving user experience. Currently, shipping times are typically determined based on logistics tracking information. However, this information can be out of sequence, partially missing, or updated slowly, reducing the accuracy of shipping time predictions. Summary of the Invention

[0003] In view of this, this application provides the following technical solutions:

[0004] A method for determining transportation time, comprising:

[0005] Based on the logistics characteristic information of the target order to be predicted, a target spatiotemporal subgraph is determined in the target graph; the target graph includes at least one spatiotemporal subgraph, each of the spatiotemporal subgraphs having a corresponding order, the order representing the backtracking dimension of the historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing the causal relationship between the nodes;

[0006] Based on the target space-time subgraph, determine the causal edge target weights of the causal edges in each target space-time subgraph that match the target order, and the order fusion weight corresponding to the target space-time subgraph, wherein the causal edge target weight represents the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target space-time subgraph on the transportation time of the target order; the order fusion weight represents the degree of influence of the target space-time subgraph of the corresponding order on the transportation time of the target order;

[0007] The target shipping time of the target order is determined based on the reference shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

[0008] Optionally, the target graph generation process includes:

[0009] Constructing a business geographic map based on the correspondence between business links and geographic locations in the historical order data; the business geographic map includes at least one logistics point, which indicates that the geographic location is marked with a corresponding business link;

[0010] Extracting a logistics trajectory subsequence of an order corresponding to each historical order in the historical order data according to the business geographic map; the logistics trajectory subsequence includes the logistics point matching the order;

[0011] Clustering logistics trajectory subsequences with the same order and matching logistics feature information into nodes of the spatiotemporal subgraph corresponding to the order;

[0012] Determining causal edges of nodes based on the causal relationships between nodes in the spatiotemporal subgraph;

[0013] The spatiotemporal subgraphs corresponding to various orders of the nodes and the causal edges between the nodes are combined into a target graph.

[0014] Optionally, extracting a logistics trajectory subsequence of an order corresponding to each historical order in the historical order data according to the business geographic map includes:

[0015] Extracting, according to the business geographic map, an initial logistics trajectory subsequence of a corresponding order for each historical order in the historical order data;

[0016] Based on a target statistical condition, determining a logistics trajectory subsequence to be marked in the initial logistics trajectory subsequence; the target statistical condition represents a condition for screening the logistics trajectory subsequence;

[0017] Determining target labeling information based on the time characteristics and logistics point characteristics corresponding to the logistics trajectory subsequence;

[0018] The target labeling information is labeled to the logistics trajectory subsequence to be labeled to determine the logistics trajectory subsequence.

[0019] Optionally, the node includes a first node and a second node, wherein determining the causal edge of the node according to the causal relationship between the nodes in the spatiotemporal subgraph includes:

[0020] Determining a second node according to a first node in the space-time subgraph, wherein the logistics points in the first node and the second node have a target matching relationship corresponding to the order of the space-time subgraph;

[0021] A causal edge between the first node and the second node is determined, where the causal edge is marked with an initial causal edge weight, and the initial causal edge weight can represent a probability parameter of predicting the transportation time of the second node through the first node.

[0022] Optionally, determining a target spatiotemporal subgraph in a target graph based on logistics characteristic information of a target order to be predicted includes:

[0023] Based on the logistics characteristic information of the target order to be predicted, obtaining the time dimension information and the transportation trajectory information of the target order;

[0024] Determining the order corresponding to the target order based on the time backtracking sub-dimension corresponding to the time dimension information and the distance backtracking sub-dimension corresponding to the to-be-transported trajectory information;

[0025] Extracting a preceding logistics trajectory subsequence matching each order from the target order; the preceding logistics trajectory subsequence represents a logistics trajectory subsequence before the target order arrives at the current logistics point to be predicted;

[0026] The node including the preceding logistics trajectory subsequence is searched in the spatiotemporal subgraph of the corresponding order in the target graph, and the spatiotemporal subgraph including the node is determined as the target spatiotemporal subgraph.

[0027] Optionally, determining the causal edge target weights of the causal edges matching the target order in each target spatiotemporal subgraph and the order fusion weight corresponding to the target spatiotemporal subgraph based on the target spatiotemporal subgraph includes:

[0028] Based on the target spatiotemporal subgraph, determining a target node in the target spatiotemporal subgraph corresponding to a preceding transport trajectory subsequence of the target order; the preceding logistics trajectory subsequence represents a logistics trajectory subsequence before the target order arrives at the current logistics point to be predicted;

[0029] Adjusting the initial causal edge weight of the causal edge corresponding to the target node and the logistics characteristic information of the target order to obtain a causal edge target weight of the causal edge that matches the target order;

[0030] Based on the historical order data, the corresponding relationship between the transportation time of the historical orders corresponding to the target spatiotemporal subgraph predictions of each order and the actual transportation time is calculated, and the order fusion weight corresponding to the target spatiotemporal subgraph is determined.

[0031] Optionally, determining the target shipping time of the target order based on the benchmark shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight includes:

[0032] For each target spatiotemporal subgraph, determining the initial shipping time of the target order corresponding to the target spatiotemporal subgraph based on the benchmark shipping time parameter corresponding to the causal edge matching the target order and the causal edge target weight;

[0033] The initial transportation time of the target order corresponding to each target spatiotemporal subgraph is processed according to the order fusion weight to obtain the target transportation time of the target order.

[0034] Optionally, for each target spatiotemporal subgraph, determining the initial transportation time of the target order corresponding to the target spatiotemporal subgraph based on a reference transportation time parameter corresponding to a causal edge matching the target order and the causal edge target weight includes:

[0035] For each target spatiotemporal subgraph, based on the benchmark transportation time parameter corresponding to the causal edge matching the target order and the target weight of the causal edge, the predicted transportation time of the logistics trajectory interval corresponding to the causal edge is determined; wherein the logistics trajectory interval represents a partial trajectory interval in the total logistics trajectory interval corresponding to the target order;

[0036] Based on the predicted transportation time of each of the logistics trajectory intervals, the initial transportation time of the target order corresponding to the target spatiotemporal subgraph is determined.

[0037] Optionally, determining the causal edge target weight of the causal edge matching the target order in each target spatiotemporal subgraph based on the target spatiotemporal subgraph includes:

[0038] The target space-time subgraph is processed based on the target network and the logistics characteristic information of the target order, and the causal edge target weights of the causal edges matching the target order in each target space-time subgraph are determined; wherein, the target network is trained based on the space-time subgraph corresponding to the historical order data, and the space-time subgraph is annotated with the corresponding logistics characteristic information of the historical order and the causal edge weights matching the actual transportation time.

[0039] An electronic device, comprising:

[0040] A memory, used to store applications and data generated by the execution of the applications;

[0041] A processor, configured to execute the application program to implement:

[0042] Based on the logistics characteristic information of the target order to be predicted, a target spatiotemporal subgraph is determined in the target graph; the target graph includes at least one spatiotemporal subgraph, each of the spatiotemporal subgraphs having a corresponding order, the order representing the backtracking dimension of the historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing the causal relationship between the nodes;

[0043] Based on the target space-time subgraph, determine the causal edge target weights of the causal edges in each target space-time subgraph that match the target order, and the order fusion weight corresponding to the target space-time subgraph, wherein the causal edge target weight represents the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target space-time subgraph on the transportation time of the target order; the order fusion weight represents the degree of influence of the target space-time subgraph of the corresponding order on the transportation time of the target order;

[0044] The target shipping time of the target order is determined based on the reference shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0046] Figure 1 A flow chart of a method for determining transport time provided in an embodiment of the present application;

[0047] Figure 2 A schematic diagram of a process for generating a target graph provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of a business geographic map provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of a target map provided in an embodiment of the present application;

[0050] Figure 5 A schematic diagram of an application scenario provided in an embodiment of the present application;

[0051] Figure 6 A schematic diagram of the structure of a transport time determination device provided in an embodiment of the present application;

[0052] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0054] In this application, the terms "first" and "second" are used to distinguish between different objects, not to describe a specific order. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements and may include steps or elements that are not listed.

[0055] The present application provides a method for determining shipping time, which can be used in scenarios where shipping time is predicted for shipping orders. This method can be applied to scenarios where shipping time is determined for ordinary orders, as well as scenarios where shipping time is predicted for complex logistics links. The present application provides an analysis of the business characteristics and shipping geographic location characteristics of the target order, as well as a thorough analysis of the impact of the order's prior status on the order's shipping, thereby improving the accuracy of shipping time predictions.

[0056] See also Figure 1 , which shows a flow chart of a method for determining transport time provided by an embodiment of the present application. The method may include the following steps:

[0057] S101. Determine a target spatiotemporal subgraph in a target graph based on logistics feature information of a target order to be predicted.

[0058] S102. Based on the target spatiotemporal subgraph, determine the causal edge target weight of the causal edge matching the target order in each target spatiotemporal subgraph, and the order fusion weight corresponding to the target spatiotemporal subgraph.

[0059] S103 : Determine the target transportation time of the target order based on the benchmark transportation time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

[0060] In this embodiment, the transportation time of the target order is determined. This can be the total transportation time from the target order's starting point to the destination, or the transportation time of a specific segment of the target order's transportation trajectory. For example, if the target order's shipping location is A, the receiving location is C, and the transit station is B, the transportation time of the target order can be the time from A to C, the determined time from A to B, or the estimated transportation time from B to C. This allows for the prediction of the target order's total transportation time, and also allows for the prediction of transportation time for each transportation stage based on the target order's real-time transportation status, thereby meeting the need for real-time updates of the predicted transportation time for the target order.

[0061] In step S101, the logistics characteristic information of the target order to be predicted may include static characteristics of the target order and dynamic characteristics of the target order during transportation. The static characteristics of the target order may include information such as the type, value, origin, and destination of the target order. The dynamic characteristics of the target order may include information such as weather characteristics during transportation and transportation status of existing transportation stages. The target order may be analyzed based on the logistics characteristic information of the target order, such as determining reference information corresponding to the target order that can be used to predict transportation time.

[0062] In order to quickly and accurately obtain reference information for predicting the shipping time of a target order in an embodiment of the present application, the corresponding information can be determined through a pre-generated target graph, such as determining a target spatiotemporal subgraph that matches the logistics characteristic information of the target order based on the target graph. The target graph includes at least one spatiotemporal subgraph, each spatiotemporal subgraph having an order corresponding thereto, which represents the backtracking dimension of the historical order data contained in the spatiotemporal subgraph. The spatiotemporal subgraph includes nodes and causal edges between nodes. The nodes are formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes represent the causal relationship between the nodes.

[0063] In order to accurately identify features in historical order data that can be used for transportation time prediction in the embodiment of the present application, a target spatiotemporal subgraph of the historical order data is generated. Among them, the historical order data is the relevant data that characterizes historical orders that have been transported. The historical order data includes the transportation trajectory corresponding to the historical order, business link information, environmental feature information during transportation, transportation time information, information on events affecting transportation, etc. By constructing a spatiotemporal subgraph corresponding to the historical order data, the spatial layout characteristics and time evolution laws of the logistics transportation trajectory in the historical orders can be clearly displayed, thereby providing a data basis for subsequent transportation time prediction.

[0064] When constructing a spatiotemporal subgraph, each subgraph can be assigned a specific order, thereby reflecting the transportation patterns of orders at different retrospective dimensions. For example, the order can represent the retrospective depth of the physical distance of the historical order's transportation trajectory. For example, a historical order was sent from location A, transited through location B, and finally arrived at location C. When the order is 1, the relationship between adjacent geographical locations can be analyzed, such as the impact of location A on location B's transportation time. When the order is 2, the impact of the transportation characteristics from location A to location B on the transportation time from location B to location C can be analyzed.

[0065] In order to clarify the impact characteristics of similar data in historical order data on transportation time, in an embodiment of the present application, the logistics trajectory subsequences of the same order in the historical order data are clustered to form nodes in the space-time subgraph. Further, clustering can be performed according to the commodity type, transportation method, etc. of the corresponding historical orders in the historical order data to make the clustering more accurate. For example, in a second-order space-time subgraph, the historical order data corresponding to the transportation of electronic equipment can be clustered, and the physical trajectory subsequence from place A to place B to place C in such historical orders is determined as the first node of the corresponding space-time subgraph, and the logistics trajectory subsequence from place B to place C to place D is determined as the second node. Since the first node and the second node both include place B to place C, a causal edge can be established between the first node and the second node. The causal edge represents the causal relationship between the first node and the second node, that is, the causal relationship of the impact of the previous logistics trajectory subsequence on the transportation time of the subsequent logistics trajectory subsequence. For example, if the order in the first node is delayed from place A to place B, the order in the second node will miss the scheduled flight to place B, thereby affecting the transportation time of the logistics trajectory in the second node. In order to quantify the causal relationship between nodes, the initial weight of the causal edge can be determined based on the transportation time in the historical order data, such as the probability of the initial weight being transferred from the first node to the second node. For example, if the average transportation time from the first node to the second node is 10 hours, the transportation time between the first node and the second node can be determined based on the initial weight, which can be the product of 10 hours and the initial weight.

[0066] After constructing space-time subgraphs of different orders based on historical order data, when determining the transportation time of the target order to be predicted, the corresponding target space-time subgraph can be matched in the target graph according to the logistics characteristic information of the target order. For example, the space-time subgraph including the geographical location of the proposed route in the logistics characteristic information can be determined. At the same time, the order of the corresponding space-time subgraph can be determined according to the logistics transportation complexity of the target order. For example, if the target order is intra-city transportation, the order can be 1. If the target order is an international order, the order can be 3. In this way, the target space-time subgraph corresponding to the target order can be matched according to the above information.

[0067] In step S102, the target spatiotemporal subgraph can be analyzed to obtain the spatiotemporal causal characteristics corresponding to the target order. Because the information contained in the target spatiotemporal subgraph is determined based on the historical order data corresponding to the target order, these historical order data can be used as the data basis for predicting the transportation time. After obtaining the target spatiotemporal subgraph, the causal edge target weight of the causal edge corresponding to the node can be determined based on the node matching the target order, the current environmental characteristics, the product characteristics of the target order, etc. For example, if the initial weight of the causal edge is 0.6, if the current weather continues to rain heavily, which will affect the transportation time, the target weight can be set to 0.5. The target spatiotemporal subgraph includes different orders. The order fusion weight corresponding to each order can be determined based on the transportation characteristics of the current target order. For example, if the target order is an international transportation order, the first order is usually suitable for short distances, and its order fusion weight can be set to a lower value such as 0.2. The corresponding third order is suitable for complex transportation trajectories, and the corresponding fusion weight of the third order can be 0.5.

[0068] After obtaining the corresponding causal edge target weight and order fusion weight in step S102, the target transportation time of the target order is calculated based on the corresponding benchmark transportation time parameter. For example, if the target order is shipped from location A, transferred at location B, and reaches location C, first, when the causal edge target weight corresponding to A to B is 0.8, the benchmark transportation time parameter from A to B represents the average transportation time from A to B, such as 8 hours. The predicted transportation time from A to B is 8*0.8=6.4 hours, and the corresponding calculated transportation time from B to C is 10*1.1=11 hours. The predicted transportation time from A to C is 17.4 hours. The target space-time subgraphs corresponding to different orders can all predict the corresponding transportation time, and then perform weighted calculation according to the order fusion weight to obtain the final transportation time. For example, the fusion weight corresponding to the first order is 0.3, and the predicted transportation time is 5 hours. The fusion weight corresponding to the second order is 0.5, and the corresponding transportation time is 8 hours. The fusion weight corresponding to the third order is 0.6, and the transportation time is 9 hours. At this time, the corresponding target transportation time is 0.3*5+0.5*8+0.6*9=10.9 hours.

[0069] A method for determining shipping time provided in an embodiment of the present application can determine a target spatiotemporal subgraph corresponding to a target order within a pre-created target graph. This spatiotemporal subgraph clarifies the impact of the preceding logistics trajectory sequence on the order's shipping time, thereby extracting similar and different logistics features from the matching historical order data. This makes the shipping time prediction process for the target order closer to the actual logistics transportation process. Furthermore, target spatiotemporal subgraphs of different orders can be used to determine the differences in the impact of preceding logistics trajectory sequences of different backtracking dimensions on the target order's shipping time, thereby identifying associations under different spatiotemporal relationships and improving the accuracy of shipping time prediction.

[0070] The following describes the method for determining the transportation time in an embodiment of the present application in combination with actual application scenarios.

[0071] See also Figure 2 , is a flow chart of a method for generating a target graph provided in an embodiment of the present application, the method may include the following steps:

[0072] S201. Construct a business geographic map based on the correspondence between business links and geographic locations in historical order data.

[0073] S202: Extracting a logistics trajectory subsequence of a corresponding order for each historical order in the historical order data according to the business geographic map.

[0074] S203: Clustering logistics trajectory subsequences with the same order and matching logistics feature information into nodes of the spatiotemporal subgraph corresponding to the order.

[0075] S204: Determine the causal edges of the nodes based on the causal relationships between the nodes in the spatiotemporal subgraph.

[0076] S205 , combining the spatiotemporal subgraphs corresponding to the nodes and the causal edges between the nodes at different orders into a target graph.

[0077] In this implementation, a business geography map is pre-built to identify business and geographic location information within historical order data. This map includes at least one logistics point, each of which is geographically labeled with its corresponding business process. Business processes refer to key business points during order transportation, such as shipping, transit, warehousing, and receipt; geographic locations refer to the actual locations of orders during transportation, such as the warehouse at location A, the airport at location B, and the receipt point at location C. This allows mapping business processes and geographic locations during physical transportation, resolving the issue of ignoring geographic variations in common approaches. Furthermore, the same business process can be mapped to different geographic locations, allowing subsequent transportation time predictions to be combined with different transportation trajectories, more accurately resembling real-world scenarios. This also provides a data foundation for the subsequent construction of spatiotemporal subgraphs. For example, a logistics trajectory sequence can be used to analyze the impact of the transportation time from "Shenzhen warehouse to Guangzhou transit station" on subsequent routes. In the embodiment of the present application, the geographical location points in the logistics points are marked with the corresponding business links. This can be done by annotating the corresponding business links in the form of text annotations for each geographical location point extracted from the historical orders. For example, if the airport in place A is usually used as a transit airport, then the airport in place A can be marked with transit-related business link information. Correspondingly, in order to facilitate visual display, the geographical location points and business links can also be annotated in the form of diagrams, as shown in FIG. Figure 3, is a schematic diagram of a business geographical map provided by an embodiment of the present application. The corresponding business links can be extracted from the historical order data in advance, with the data set V B Indicates that V B ={m1,m2,m3,…,m n}, in this data set V B It includes n business links; the data set corresponding to the geographical location point is V G Indicates that V G ={l1,l2,l3,…,l k}, in this data set V G The data includes k geographical locations. Then, based on the historical order data, the corresponding relationship between the two is determined to obtain the business geographical map. Figure 3 As shown, business link m1 can be executed at geographical locations such as l1 and l2.

[0078] Then, based on the logistics trajectory record information in the historical order data and the business geographic map, the logistics trajectory of each order is determined. Different logistics trajectory subsequences can be selected for different orders. For example, the logistics trajectory of an order is ABCD, where A, B, C, and D represent four different logistics points respectively. For the first order, each logistics trajectory subsequence can contain one logistics point. For example, it can be divided into four subsequences, A, B, C, and D, which can be used to analyze the independent characteristics of a single logistics point, such as the average time required for customs clearance at place B; for the second order, each subsequence can contain two consecutive logistics points, such as AB, BC, and CD, which can be used to analyze the transfer relationship between adjacent nodes, such as the transportation time from the warehouse to the airport; for the third order, each subsequence can contain three consecutive logistics points, such as ABC and BCD, which can be used to identify longer-distance spatiotemporal causal relationships, such as the overall time consumption pattern from shipment to transit to customs clearance.

[0079] Logistics trajectory subsequences with the same order and matching logistics feature information will be clustered into nodes of the space-time subgraph corresponding to the order. Among them, the matching of logistics feature information can refer to the same or similar types of corresponding transport items in historical orders, close values, or pick-up times for delivery, etc., which makes the clustering results more accurate and can provide a more accurate data basis for subsequent transportation time predictions. For high-order clustering, since the logistics trajectory subsequences of two high-order sequences will have overlapping subsequences, the two high-order sequences can be used as two nodes respectively, and there is a causal relationship between the two, thereby establishing a causal edge between the two. For example, if "ABC" and "BCD" share "BC", there is a causal edge between them. In an embodiment of the present application, low order (such as order 1) can capture local features, and high order (such as order 3) can capture long-range dependency information, which can provide an accurate reference data basis for subsequent transportation time predictions. Based on the nodes corresponding to each order and the causal edges between the nodes, the time-space subgraph corresponding to the order can be generated. Then, the time-space subgraphs corresponding to each order can be combined into a target graph. When predicting the shipping time of the target order, the target time-space subgraph corresponding to the target order can be extracted based on the target graph as reference data for subsequent shipping time prediction. Figure 4 , which shows a schematic diagram of a target graph provided by an embodiment of the present application, Figure 4 The target graph shown only shows the subgraphs corresponding to the second and third order. For example, in a second-order graph, "AB" is a node, and "BC" is a node. These two nodes share a common logistics point, B. The causal edge between them is represented by an arrow. The same applies to the other spatiotemporal subgraphs, so we will not explain them one by one.

[0080] Correspondingly, in one implementation of the embodiment of the present application, the process of extracting the logistics trajectory subsequence of the order corresponding to each historical order in the historical order data according to the business geographic map may include the following steps:

[0081] According to the business geographic map, the initial logistics trajectory subsequence of the corresponding order for each historical order in the historical order data is extracted; based on the target statistical conditions, the logistics trajectory subsequence to be labeled is determined in the initial logistics trajectory subsequence; based on the time characteristics and logistics point characteristics corresponding to the logistics trajectory subsequence, the target labeling information is determined; the target labeling information is annotated to the logistics trajectory subsequence to be labeled to determine the logistics trajectory subsequence.

[0082] Among them, the initial logistics trajectory subsequence of the corresponding order in each historical order can be extracted according to the business geographic map. In order to improve data quality and reference value, the initial logistics trajectory subsequences that do not meet the target statistical conditions can be eliminated in the embodiment of the present application. The target statistical conditions represent the conditions for screening logistics trajectory subsequences, such as the average transportation time of the corresponding logistics trajectory subsequences. If the average transportation time of a certain type of logistics trajectory subsequence is 10 hours, and the transportation time of a very small part of the logistics trajectory subsequences of this type is 40 hours, then these logistics trajectory subsequences can be eliminated because this type of logistics trajectory subsequence may have accidental influencing factors, such as forgetting to ship at the specified time.

[0083] Furthermore, in order to clearly obtain the time and space characteristics of the corresponding logistics trajectory subsequence, after extracting the relevant logistics trajectory subsequence, it can also be marked with the corresponding target annotation information to facilitate the subsequent application, backtracking and identification of data. For example, it can be marked according to time characteristics and logistics point characteristics, such as location information, environmental characteristic information, and time series information (which can be the actual generation time of each logistics point, such as the actual delivery time, actual loading time, etc., or the transportation time between adjacent logistics points, the corresponding average transportation time, etc.), so as to facilitate subsequent data application and processing.

[0084] Correspondingly, in one implementation of the embodiment of the present application, the causal edge between the nodes of the spatiotemporal subgraph in the embodiment of the present application is described by taking the establishment of a causal edge between the first node and the second node as an example. According to the first node in the spatiotemporal subgraph, the second node is determined, wherein the logistics points in the first node and the second node have a target matching relationship corresponding to the order of the spatiotemporal subgraph. The causal edge between the first node and the second node is determined, and the causal edge is marked with an initial weight of the causal edge; the initial weight of the causal edge represents the probability that the first node causes a change in the transportation time of the second node.

[0085] This implementation primarily focuses on processing second-order and higher-order spatiotemporal subgraphs. This is because the first-order subgraphs only analyze a single logistics point and can be used for short-distance or intra-city transportation scenarios. For complex logistics and transportation scenarios, second-order and higher-order spatiotemporal subgraphs are typically used. Therefore, the analysis in this application embodiment is primarily based on this subgraph. The logistics points in the first node and the second node have a target matching relationship corresponding to the order of the space-time subgraph, which means that the common logistics points in the first node and the second node are determined according to the order corresponding to the space-time subgraph. For example, in a second-order space-time subgraph, the first node is a clustered similar logistics trajectory subsequence AB, and the second node is a clustered similar logistics trajectory subsequence BC. Then the first node and the second node both include logistics point B. Then a causal edge can be established between the first node and the second node. According to the transportation time corresponding to the first node, the transportation time corresponding to the second node is estimated. Then, the initial weight of the causal edge is determined based on the actual transportation time corresponding to the second node and the environmental characteristics corresponding to the current historical order. That is, the initial weight can represent the probability parameter of predicting the transportation time of the second node through the first node, and further can represent the accuracy probability parameter of the prediction. For example, the transportation time of the corresponding logistics trajectory subsequence in the second node is 10 hours on average, and the initial weight of the causal edge is 0.8, that is, the transportation time predicted to the corresponding logistics trajectory subsequence in the second node through the first node is 8 hours. Due to the deviations in environmental characteristics or transport items in historical order data, in the subsequent application to the actual target order transportation time prediction scenario, the initial weights of the causal edges can be optimized according to the real-time scenario corresponding to the target order, thereby improving the accuracy of transportation time determination.

[0086] After the target graph is generated, in actual application scenarios, a target spatiotemporal subgraph can be determined within the target graph based on the logistics characteristics of the target order to be predicted. This target spatiotemporal subgraph can represent the reference features used in the transportation time prediction process. Accordingly, this process can include the following steps:

[0087] Based on the logistics characteristics of the target order to be predicted, the target order's time dimension information and transportation trajectory information are obtained. The order corresponding to the target order is determined based on the time backtracking sub-dimension corresponding to the time dimension information and the distance backtracking sub-dimension corresponding to the transportation trajectory information. The preceding logistics trajectory subsequences matching each order are extracted from the target order. The spatiotemporal subgraph of the corresponding stage in the target graph is searched for nodes containing the preceding logistics trajectory subsequences, and the spatiotemporal subgraph containing these nodes is determined as the target spatiotemporal subgraph. The preceding logistics trajectory subsequences guarantee the logistics trajectory subsequences of the target order before it arrives at the logistics point to be predicted.

[0088] The target orders to be predicted may have different logistics characteristics, such as different product types, shipping times, or shipping methods and requirements. This information can affect subsequent shipping times. To accurately predict the corresponding shipping time, a backtracking dimension for historical order data that matches the current logistics characteristics of the target order can be determined. Specifically, based on the target order's time dimension and the pending shipping trajectory information, a time backtracking sub-dimension and a distance backtracking sub-dimension are determined. The distance backtracking sub-dimension can refer to the number of consecutive logistics points included in the preceding logistics trajectory subsequence. For example, the distance backtracking sub-dimension is represented by i. Based on the target order's time dimension information (such as the order's shipping time and estimated arrival time), a meta-time Δt is set. This is a single-order time length, such as 4 hours, three days, one week, or two weeks. Correspondingly, the time length corresponding to the time backtracking sub-dimension can be no more than "i*Δt", meaning that the selected historical order data must fall within this time length. In this case, the target order corresponds to the order i, and the time length of the historical order data source in the space-time subgraph corresponding to this order can be no more than the time length corresponding to "i*Δt". Based on the logistics points included in the target order, the corresponding preceding logistics trajectory subsequences of each order are determined. The target space-time subgraph is then defined as the spatiotemporal subgraph containing these preceding logistics trajectory subsequences. For example, at the second order, the preceding logistics trajectory subsequence is AB, and its corresponding subsequent trajectory subsequence is BC. The target space-time subgraph is defined as the spatiotemporal subgraph with AB as the first node and BC as the second node, with a causal edge between the first and second nodes. This approach provides a precise historical order data reference for quickly and accurately predicting the shipping time of the target order.

[0089] Correspondingly, after determining the target spatiotemporal subgraph, the weight parameters for prediction can be determined. That is, in the embodiment of the present application, based on the target spatiotemporal subgraph, the process of determining the causal edge target weights of the causal edges matching the target order in each target spatiotemporal subgraph and the order fusion weight corresponding to the target spatiotemporal subgraph can include the following steps:

[0090] S301. Based on the target spatiotemporal subgraph, determine the target node in the target spatiotemporal subgraph corresponding to the target order's transportation trajectory subsequence. The preceding logistics trajectory subsequence represents the logistics trajectory subsequence before the target order arrives at the to-be-predicted logistics point. The to-be-predicted logistics point can be any logistics point in the target order, or a logistics point that has not yet been reached during the actual prediction process. For example, if the target order's planned trajectory is ABC and it has already been dispatched from location B, the to-be-predicted logistics point can be C.

[0091] S302. Adjust the initial causal edge weight according to the causal edge initial weight corresponding to the target node and the logistics characteristic information of the target order to obtain the causal edge target weight of the causal edge matching the target order.

[0092] For example, if the target order's intended trajectory is ABC and it has already been dispatched from point B, the logistics point to be predicted could be C. AB can be used as the preceding logistics trajectory subsequence, and the node that includes this AB logistics trajectory subsequence in the target spatiotemporal subgraph is used as the target node. The initial causal edge weights of the causal edge from the target node to the corresponding node, i.e., the node that includes BC's logistics subsequence, are then obtained. These initial weights are adjusted based on the target order's logistics characteristics, such as weather conditions, the value of the goods, unexpected conditions in the current transportation trajectory, or the backlog of goods at the corresponding logistics point, to determine the adjustment ratio. This results in the target causal edge weight.

[0093] S303: Based on the historical order data, the corresponding relationship between the transportation time of the historical orders corresponding to the target spatiotemporal subgraph predictions of each order and the actual transportation time is calculated, and the order fusion weight corresponding to the target spatiotemporal subgraph is determined.

[0094] For example, if the actual shipping time for a historical order is 10 hours, the time predicted by the second-order spatiotemporal subgraph is 12 hours, and the time predicted by the third-order spatiotemporal subgraph is 9 hours, then the third-order prediction is closer to the actual shipping time, and its corresponding order fusion weight will be higher than the corresponding order fusion weight of the second-order. Furthermore, the order fusion weight can be adjusted based on the logistics information characteristics of the target order. For example, if the target order's transportation process is more complex and involves many logistics points, the corresponding higher-order spatiotemporal subgraph will have a higher order fusion weight.

[0095] For example, if the first-order determined transportation time is T1, the corresponding order fusion weight is a; if the first-order determined transportation time is T2, the corresponding order fusion weight is b; if the third-order determined transportation time is T3, the corresponding order fusion weight is c, then the target time is T = T1*a+T2*b+T3*c. This allows the information of each order to be integrated, making the final transportation time closer to the actual situation.

[0096] In one implementation of the embodiment of the present application, a process of determining a target shipping time for a target order based on a reference shipping time parameter corresponding to a causal edge matching the target order, a causal edge target weight, and an order fusion weight may include the following steps:

[0097] For each target space-time subgraph, the initial transportation time of the target order corresponding to the target space-time subgraph is determined based on the benchmark transportation time parameters corresponding to the causal edge matching the target order and the causal edge target weight; the initial transportation time of the target order corresponding to each target space-time subgraph is processed according to the order fusion weight to obtain the target transportation time of the target order.

[0098] In one implementation of the embodiment of the present application, for each target spatiotemporal subgraph, a process of determining the initial shipping time of the target order corresponding to the target spatiotemporal subgraph based on the baseline shipping time parameter corresponding to the causal edge matching the target order and the target weight of the causal edge may include the following steps:

[0099] For each target spatiotemporal subgraph, based on the benchmark transportation time parameters corresponding to the causal edge matching the target order and the target weight of the causal edge, the predicted transportation time of the logistics trajectory interval corresponding to the causal edge is determined; wherein, the logistics trajectory interval represents a part of the trajectory interval in the total logistics trajectory interval corresponding to the target order; based on the predicted transportation time of each of the logistics trajectory intervals, the initial transportation time of the target order corresponding to the target spatiotemporal subgraph is determined.

[0100] Among them, the benchmark transportation time can be determined based on the average transportation time. For example, the transportation trajectory corresponding to the target order includes the logistics point "ABC". The time corresponding to the benchmark transportation time parameter of AB corresponding to the second-order target space-time subgraph is 2 days, and the time corresponding to the benchmark transportation time parameter of BC is 1 day; the benchmark time of ABC corresponding to the third-order target space-time subgraph is 3.2 days. The causal edge target weight corresponding to the second-order target space-time subgraph is AB: 0.6 (affected by the current weather, transportation may be delayed); BC: 0.9 (customs clearance efficiency is high, and the weight is close to the benchmark). The causal edge target weight corresponding to the third-order target space-time subgraph is ABC: 0.7 (the overall path time of historically similar orders is shorter).

[0101] For the second-order spacetime subgraph:

[0102] AB initial time = benchmark time * weight = 2 days * 0.6 = 1.2 days;

[0103] BC initial time = 1 day * 0.9 = 0.9 days;

[0104] Total initial time for the second stage = 1.2 + 0.9 = 2.1 days.

[0105] For the third-order spacetime subgraph:

[0106] ABC initial time = 3.2 days * 0.7 = 2.24 days.

[0107] The order fusion weights include a second-order weight of 0.4 and a third-order weight of 0.6, and the corresponding target transportation time is:

[0108] Target transportation time = 2.1×0.4+2.24×0.6=0.84+1.344=2.184 days.

[0109] In order to quickly and accurately obtain relevant weight parameters for calculating the target shipping time, such as the causal edge target weight, in the embodiments of the present application, a pre-trained deep learning network can be used to learn the causal edge features in each spatiotemporal subgraph. Accordingly, in one embodiment of the present application, based on the target spatiotemporal subgraph, the causal edge target weight of the causal edge that matches the target order in each target spatiotemporal subgraph is determined, including:

[0110] The target space-time subgraph is processed based on the target network and the logistics characteristic information of the target order, and the causal edge target weights of the causal edges matching the target order in each target space-time subgraph are determined; wherein, the target network is trained based on the space-time subgraph corresponding to the historical order data, and the space-time subgraph is annotated with the corresponding logistics characteristic information of the historical order and the causal edge weights matching the actual transportation time.

[0111] In this embodiment, the determination of the target weight of the causal edge is obtained by matching the logistics feature information of the target order with the historical pattern through a pre-trained target network (such as a spatiotemporal graph neural network, ST-GNN). The data used to train the target network can be a spatiotemporal subgraph of historical orders (such as first-order, second-order, and third-order subgraphs), and the causal edge annotation information of each spatiotemporal subgraph includes logistics characteristics (such as transportation mode, weather, commodity type, etc.) and actual transportation time (true value). The dynamic weights of the causal edges are learned so that the error between the predicted transportation time and the actual time is minimized (such as MSE loss). Among them, the network structure of the target network during the training process can adopt a spatiotemporal graph network (ST-GNN), combined with graph convolution (GCN) and time series modeling (such as LSTM) to capture spatiotemporal dependencies. In this embodiment, the target network adjusts the weights through real-time features, which solves the problem that traditional methods cannot adapt to emergencies (such as weather changes). Weights of different orders are ultimately used for weighted fusion of initial transportation time (such as second-order weight 0.4, third-order weight 0.6) to improve prediction accuracy.

[0112] See also Figure 5, is a schematic diagram of an application scenario provided by an embodiment of the present application. In this application scenario, the target graph includes multiple space-time subgraphs. When the transportation time of the target order needs to be predicted, the target space-time subgraph can be queried in the target graph based on the logistics feature information of the target order, and then the causal edge features of the target space-time subgraph are learned based on the target network to determine the causal edge target weights and the order fusion weights of target space-time subgraphs of different orders. Finally, the time of each transportation path and the corresponding total transportation time are determined in the prediction module according to the simulated transportation trajectory of the target order and the above-mentioned weight parameters.

[0113] In the process of generating the target graph in the embodiment of the present application, different order spatiotemporal subgraphs can be constructed. Each spatiotemporal subgraph can trace back the transportation trajectory information of historical order data of different time lengths and geographical distances. Therefore, the spatiotemporal information of the transportation trajectory of the historical order data that can be identified is different. The causal relationship between each node can also be characterized, and the impact of the previous state of the order on the future transportation time of the order is fully considered, thereby accurately depicting the similarities and differences between orders, and being closer to the real logistics transportation process. When learning the causal edge features, the initial weight of the causal edge can be optimized according to the real-time logistics characteristics of the target order, and then the order fusion weight can be determined, which can effectively depict the difference in the impact of the previous state of different lengths on the order transportation time, thereby identifying the correlation relationship under different spatiotemporal relationships and improving the accuracy of the transportation time prediction.

[0114] The present application also provides a transport time determination device, see Figure 6 , the device comprises:

[0115] A first determining unit 601 is configured to determine a target spatiotemporal subgraph in a target graph based on logistics feature information of a target order to be predicted; the target graph includes at least one spatiotemporal subgraph, each spatiotemporal subgraph having a corresponding order, the order representing a backtracking dimension of historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing a causal relationship between the nodes;

[0116] A second determining unit 602 is configured to determine, based on the target spatiotemporal subgraph, the causal edge target weights of the causal edges in each target spatiotemporal subgraph that matches the target order, and the order fusion weights corresponding to the target spatiotemporal subgraphs, wherein the causal edge target weights represent the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target spatiotemporal subgraph on the transportation time of the target order; and the order fusion weights represent the degree of influence of the target spatiotemporal subgraph of the corresponding order on the transportation time of the target order.

[0117] The third determining unit 603 is configured to determine the target transportation time of the target order based on the reference transportation time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

[0118] Optionally, a target graph generation unit is further included, and the target graph generation unit includes:

[0119] A construction subunit is configured to construct a business geographic map based on the correspondence between business links and geographic locations in the historical order data; the business geographic map includes at least one logistics point, which indicates that the geographic location is marked with a corresponding business link;

[0120] an extraction subunit, configured to extract, based on the business geographic map, a logistics trajectory subsequence of an order corresponding to each historical order in the historical order data; the logistics trajectory subsequence including the logistics points matching the order;

[0121] A clustering subunit, configured to cluster logistics trajectory subsequences having the same order and matching logistics feature information into nodes of a spatiotemporal subgraph corresponding to the order;

[0122] A first determining subunit, configured to determine a causal edge of a node based on a causal relationship between nodes in the spatiotemporal subgraph;

[0123] The combining sub-unit is used to combine the spatiotemporal sub-graphs corresponding to various orders of the nodes and the causal edges between the nodes into a target graph.

[0124] Optionally, the extraction subunit is configured to:

[0125] Extracting, according to the business geographic map, an initial logistics trajectory subsequence of a corresponding order for each historical order in the historical order data;

[0126] Based on a target statistical condition, determining a logistics trajectory subsequence to be marked in the initial logistics trajectory subsequence; the target statistical condition represents a condition for screening the logistics trajectory subsequence;

[0127] Determining target labeling information based on the time characteristics and logistics point characteristics corresponding to the logistics trajectory subsequence;

[0128] The target labeling information is labeled to the logistics trajectory subsequence to be labeled to determine the logistics trajectory subsequence.

[0129] Optionally, the node includes a first node and a second node, wherein the first determining subunit is configured to:

[0130] Determining a second node according to a first node in the space-time subgraph, wherein the logistics points in the first node and the second node have a target matching relationship corresponding to the order of the space-time subgraph;

[0131] A causal edge between the first node and the second node is determined, where the causal edge is marked with an initial causal edge weight, and the initial causal edge weight can represent a probability parameter of predicting the transportation time of the second node through the first node.

[0132] Optionally, the first determining unit includes:

[0133] A first acquisition subunit is configured to obtain time dimension information and to-be-transported trajectory information of a target order to be predicted based on logistics characteristic information of the target order to be predicted;

[0134] A second determining subunit is configured to determine the order corresponding to the target order based on the time backtracking sub-dimension corresponding to the time dimension information and the distance backtracking sub-dimension corresponding to the to-be-transported trajectory information;

[0135] A sequence extraction sub-unit is used to extract a preceding logistics trajectory sub-sequence matching each order from the target order; the preceding logistics trajectory sub-sequence represents a logistics trajectory sub-sequence before the target order arrives at the current logistics point to be predicted;

[0136] The query subunit is used to query the node including the preceding logistics trajectory subsequence in the spatiotemporal subgraph of the corresponding order in the target graph, and determine the spatiotemporal subgraph including the node as the target spatiotemporal subgraph.

[0137] Optionally, the second determining unit includes:

[0138] A third determining subunit is configured to determine, based on the target spatiotemporal subgraph, a target node in the target spatiotemporal subgraph corresponding to a preceding transport trajectory subsequence of the target order; the preceding logistics trajectory subsequence represents a logistics trajectory subsequence before the target order arrives at the current logistics point to be predicted;

[0139] an adjusting subunit, configured to adjust the initial causal edge weight of the causal edge corresponding to the target node and the logistics characteristic information of the target order to obtain a target causal edge weight of the causal edge that matches the target order;

[0140] The fourth determination subunit is used to calculate the corresponding relationship between the transportation time of the historical orders corresponding to the target spatiotemporal subgraph predictions of each order and the actual transportation time based on the historical order data, and determine the order fusion weight corresponding to the target spatiotemporal subgraph.

[0141] Optionally, the third determining unit includes:

[0142] a fifth determining subunit, configured to determine, for each target spatiotemporal subgraph, an initial transportation time of the target order corresponding to the target spatiotemporal subgraph based on a reference transportation time parameter corresponding to a causal edge matching the target order and a target weight of the causal edge;

[0143] The second acquisition subunit is used to process the initial transportation time of the target order corresponding to each target spatiotemporal subgraph according to the order fusion weight to obtain the target transportation time of the target order.

[0144] Optionally, the fifth determining subunit is configured to:

[0145] For each target spatiotemporal subgraph, based on the benchmark transportation time parameter corresponding to the causal edge matching the target order and the target weight of the causal edge, the predicted transportation time of the logistics trajectory interval corresponding to the causal edge is determined; wherein the logistics trajectory interval represents a partial trajectory interval in the total logistics trajectory interval corresponding to the target order;

[0146] Based on the predicted transportation time of each of the logistics trajectory intervals, the initial transportation time of the target order corresponding to the target spatiotemporal subgraph is determined.

[0147] Optionally, determining the causal edge target weight of the causal edge matching the target order in each target spatiotemporal subgraph based on the target spatiotemporal subgraph includes:

[0148] The target space-time subgraph is processed based on the target network and the logistics characteristic information of the target order, and the causal edge target weights of the causal edges matching the target order in each target space-time subgraph are determined; wherein, the target network is trained based on the space-time subgraph corresponding to the historical order data, and the space-time subgraph is annotated with the corresponding logistics characteristic information of the historical order and the causal edge weights matching the actual transportation time.

[0149] It should be noted that the specific implementation of each unit and sub-unit in this embodiment can refer to the corresponding content in the previous text and will not be described in detail here.

[0150] In another embodiment of the present application, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for determining the transportation time as described in the above embodiment is implemented.

[0151] In another embodiment of the present application, an electronic device is provided. Figure 7 ,include:

[0152] Memory 701, used to store applications and data generated by the application;

[0153] The processor 702 is configured to execute the application program to implement:

[0154] Based on the logistics characteristic information of the target order to be predicted, a target spatiotemporal subgraph is determined in the target graph; the target graph includes at least one spatiotemporal subgraph, each of the spatiotemporal subgraphs having a corresponding order, the order representing the backtracking dimension of the historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing the causal relationship between the nodes;

[0155] Based on the target space-time subgraph, determine the causal edge target weights of the causal edges in each target space-time subgraph that match the target order, and the order fusion weight corresponding to the target space-time subgraph, wherein the causal edge target weight represents the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target space-time subgraph on the transportation time of the target order; the order fusion weight represents the degree of influence of the target space-time subgraph of the corresponding order on the transportation time of the target order;

[0156] The target shipping time of the target order is determined based on the reference shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

[0157] It should be noted that the specific implementation of the processor in this embodiment can refer to the corresponding content in the previous text and will not be described in detail here.

[0158] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0159] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0161] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining transport time, comprising: Based on the logistics characteristic information of the target order to be predicted, a target spatiotemporal subgraph is determined in the target graph; The target graph includes at least one spatiotemporal subgraph, each of the spatiotemporal subgraphs having a corresponding order, the order representing the backtracking dimension of the historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing the causal relationship between the nodes; Based on the target space-time subgraph, determine the causal edge target weights of the causal edges in each target space-time subgraph that match the target order, and the order fusion weight corresponding to the target space-time subgraph, wherein the causal edge target weight represents the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target space-time subgraph on the transportation time of the target order; the order fusion weight represents the degree of influence of the target space-time subgraph of the corresponding order on the transportation time of the target order; The target shipping time of the target order is determined based on the reference shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

2. The method according to claim 1, wherein the target graph generation process comprises: Build a business geographic map based on the correspondence between business links and geographic locations in historical order data; The business geographic map includes at least one logistics point, and the logistics point represents that the geographical location point is marked with a corresponding business link; Extracting a logistics trajectory subsequence of an order corresponding to each historical order in the historical order data according to the business geographic map; the logistics trajectory subsequence includes the logistics point matching the order; Clustering logistics trajectory subsequences with the same order and matching logistics feature information into nodes of the spatiotemporal subgraph corresponding to the order; Determining causal edges of nodes based on the causal relationships between nodes in the spatiotemporal subgraph; The spatiotemporal subgraphs corresponding to various orders of the nodes and the causal edges between the nodes are combined into a target graph.

3. The method according to claim 2, wherein extracting a logistics trajectory subsequence of a corresponding order for each historical order in the historical order data based on the business geographic map comprises: Extracting, according to the business geographic map, an initial logistics trajectory subsequence of a corresponding order for each historical order in the historical order data; Based on a target statistical condition, determining a logistics trajectory subsequence to be marked in the initial logistics trajectory subsequence; the target statistical condition represents a condition for screening the logistics trajectory subsequence; Determining target labeling information based on the time characteristics and logistics point characteristics corresponding to the logistics trajectory subsequence; The target labeling information is labeled to the logistics trajectory subsequence to be labeled to determine the logistics trajectory subsequence.

4. The method according to claim 2, wherein the node comprises a first node and a second node, wherein: Determining the causal edges of the nodes according to the causal relationship between the nodes in the spatiotemporal subgraph includes: Determining a second node according to a first node in the space-time subgraph, wherein the logistics points in the first node and the second node have a target matching relationship corresponding to the order of the space-time subgraph; A causal edge between the first node and the second node is determined, where the causal edge is marked with an initial causal edge weight, and the initial causal edge weight can represent a probability parameter of predicting the transportation time of the second node through the first node.

5. The method according to claim 1, wherein determining a target spatiotemporal subgraph in a target graph based on logistics characteristic information of a target order to be predicted comprises: Based on the logistics characteristic information of the target order to be predicted, obtaining the time dimension information and the transportation trajectory information of the target order; Determining the order corresponding to the target order based on the time backtracking sub-dimension corresponding to the time dimension information and the distance backtracking sub-dimension corresponding to the to-be-transported trajectory information; Extracting a preceding logistics trajectory subsequence matching each order from the target order; the preceding logistics trajectory subsequence represents a logistics trajectory subsequence before the target order arrives at the current logistics point to be predicted; The node including the preceding logistics trajectory subsequence is searched in the spatiotemporal subgraph of the corresponding order in the target graph, and the spatiotemporal subgraph including the node is determined as the target spatiotemporal subgraph.

6. The method according to claim 1, wherein determining, based on the target spatiotemporal subgraph, the causal edge target weights of the causal edges in each target spatiotemporal subgraph that match the target order, and the order fusion weight corresponding to the target spatiotemporal subgraph comprises: Based on the target spatiotemporal subgraph, determining a target node in the target spatiotemporal subgraph corresponding to a preceding transport trajectory subsequence of the target order; The preceding logistics trajectory subsequence represents the logistics trajectory subsequence before the target order arrives at the current logistics point to be predicted; Adjusting the initial causal edge weight of the causal edge corresponding to the target node and the logistics characteristic information of the target order to obtain a causal edge target weight of the causal edge that matches the target order; Based on the historical order data, the corresponding relationship between the transportation time of the historical orders corresponding to the target spatiotemporal subgraph predictions of each order and the actual transportation time is calculated, and the order fusion weight corresponding to the target spatiotemporal subgraph is determined.

7. The method according to claim 1, wherein determining the target shipping time of the target order based on the benchmark shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight comprises: For each target spatiotemporal subgraph, determining the initial shipping time of the target order corresponding to the target spatiotemporal subgraph based on the benchmark shipping time parameter corresponding to the causal edge matching the target order and the causal edge target weight; The initial transportation time of the target order corresponding to each target spatiotemporal subgraph is processed according to the order fusion weight to obtain the target transportation time of the target order.

8. The method according to claim 7, wherein for each target spatiotemporal subgraph, determining the initial shipping time of the target order corresponding to the target spatiotemporal subgraph based on the benchmark shipping time parameter corresponding to the causal edge matching the target order and the causal edge target weight comprises: For each target spatiotemporal subgraph, based on the benchmark transportation time parameter corresponding to the causal edge matching the target order and the target weight of the causal edge, the predicted transportation time of the logistics trajectory interval corresponding to the causal edge is determined; wherein the logistics trajectory interval represents a partial trajectory interval in the total logistics trajectory interval corresponding to the target order; Based on the predicted transportation time of each of the logistics trajectory intervals, the initial transportation time of the target order corresponding to the target spatiotemporal subgraph is determined.

9. The method according to claim 1, wherein determining the causal edge target weight of each causal edge in the target spatiotemporal subgraph that matches the target order based on the target spatiotemporal subgraph comprises: The target space-time subgraph is processed based on the target network and the logistics characteristic information of the target order, and the causal edge target weights of the causal edges matching the target order in each target space-time subgraph are determined; wherein, the target network is trained based on the space-time subgraph corresponding to the historical order data, and the space-time subgraph is annotated with the corresponding logistics characteristic information of the historical order and the causal edge weights matching the actual transportation time.

10. An electronic device comprising: A memory, used to store applications and data generated by the execution of the applications; A processor, configured to execute the application program to implement: Based on the logistics characteristic information of the target order to be predicted, a target spatiotemporal subgraph is determined in the target graph; the target graph includes at least one spatiotemporal subgraph, each of the spatiotemporal subgraphs having a corresponding order, the order representing the backtracking dimension of the historical order data contained in the spatiotemporal subgraph, the spatiotemporal subgraph including nodes and causal edges between nodes, the nodes being formed by clustering logistics trajectory subsequences of the same order in the historical order data, and the causal edges between nodes representing the causal relationship between the nodes; Based on the target space-time subgraph, determine the causal edge target weights of the causal edges in each target space-time subgraph that match the target order, and the order fusion weight corresponding to the target space-time subgraph, wherein the causal edge target weight represents the degree of influence of the preceding logistics trajectory subsequence corresponding to the target order determined based on the target space-time subgraph on the transportation time of the target order; the order fusion weight represents the degree of influence of the target space-time subgraph of the corresponding order on the transportation time of the target order; The target shipping time of the target order is determined based on the reference shipping time parameter corresponding to the causal edge matching the target order, the causal edge target weight, and the order fusion weight.

Images