A time prediction method, device, computer device and storage medium

Abstract

Embodiments of the present application provide a time prediction method and device, computer equipment and storage medium, which can be applied to cloud technology, artificial intelligence, intelligent transportation, Internet of Vehicles and various scenes; the embodiments of the present application can obtain a route network of a target route from a starting point to an ending point; feature extraction is performed on a main node to obtain main road section feature information of a main road section, and feature extraction is performed on an auxiliary node to obtain surrounding road section feature information of a surrounding road section; according to the route network, the main road section feature information and the surrounding road section feature information are aggregated to obtain aggregated feature information of the main node; according to the aggregated feature information, time prediction processing is performed on the target route to obtain an arrival time of the target route from the starting point to the ending point; the embodiments of the present application can accurately predict the arrival time of the target route.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a time prediction method, apparatus, computer device, and storage medium, wherein the storage medium is a computer-readable storage medium. Background Technology

[0002] Currently, map software estimates the travel time from the origin to the destination. There are many methods used by map software to estimate arrival time. For example, map software may use human experience to estimate arrival time, but this method often fails to provide accurate estimates. Another method is to use neural network models to predict arrival time; however, neural network models cannot accurately assess road conditions, which can also lead to inaccurate arrival time predictions.

[0003] In summary, there is currently an issue with inaccurate arrival time estimates. Summary of the Invention

[0004] This application provides a time prediction method, apparatus, computer device, and storage medium that can accurately predict arrival time.

[0005] A time prediction method, comprising:

[0006] Obtain the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments.

[0007] Feature extraction is performed on the main nodes to obtain the main road segment feature information, and feature extraction is performed on the auxiliary nodes to obtain the surrounding road segment feature information.

[0008] Based on the route network, the feature information of the main road segment and the feature information of the surrounding road segment are aggregated to obtain the aggregated feature information of the main node.

[0009] Based on the aggregated feature information, time prediction processing is performed on the target route to obtain the arrival time of the target route from the starting point to the destination.

[0010] Accordingly, embodiments of this application provide a time prediction device, including:

[0011] The acquisition unit can be used to acquire the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments.

[0012] The extraction unit can be used to extract features from the main nodes to obtain the main road segment feature information, and to extract features from the auxiliary nodes to obtain the surrounding road segment feature information.

[0013] The aggregation unit can be used to aggregate the feature information of the main road segment and the feature information of the surrounding road segment according to the route network, so as to obtain the aggregated feature information of the main node.

[0014] The prediction unit can be used to perform time prediction processing on the target route based on the aggregated feature information, so as to obtain the arrival time of the target route from the starting point to the destination.

[0015] In some embodiments, the aggregation unit may be used to determine the neighboring nodes of each node according to the route network. The nodes include primary nodes and secondary nodes, and the neighboring nodes include primary nodes and secondary nodes. The aggregation unit performs aggregation processing on the trunk road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes to obtain the aggregated feature information of the primary node.

[0016] In some embodiments, the aggregation unit may be specifically used to aggregate the main road segment feature information of the main node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes of each node to obtain the aggregated feature information of the node; and to extract the aggregated feature information of the main node from the aggregated feature information of the node.

[0017] In some embodiments, the aggregation unit can be specifically used to aggregate the main road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes of the node, as well as the road segment feature information of the node itself, for each node, to obtain the aggregated feature information of the node, wherein the road segment feature information of the node is the main road segment feature information of the primary node or the surrounding road segment feature information of the secondary node; and to extract the aggregated feature information of the primary node from the aggregated feature information of the node.

[0018] In some embodiments, the acquisition unit may be specifically used to acquire at least one candidate route from the starting point to the ending point; determine the target route from the at least one candidate route; and acquire the route network of the target route based on the target route.

[0019] In some embodiments, the acquisition unit may be specifically used to acquire the main road segment of the target route and the surrounding road segments corresponding to the main road segment; determine the main node in the route network based on the main road segment; determine the auxiliary node in the route network based on the surrounding road segments; and determine the edge in the route network based on the positional relationship between the main road segments and the positional relationship between the main road segment and the surrounding road segments; and construct the route network of the target route based on the main node, auxiliary node and edge.

[0020] In some embodiments, the aggregated feature information is information obtained by aggregating the time prediction model after training; the time prediction device further includes a training unit, which can be used to acquire a route sample data set, which includes route data samples labeled with arrival time tags; based on the route data samples, time prediction is performed using the time prediction model to be trained to obtain the predicted arrival time; based on the predicted arrival time and the arrival time tags, the model parameters of the time prediction model to be trained are converged to obtain the trained time prediction model.

[0021] In some embodiments, the training unit can be specifically used to acquire historical route information, which includes historical arterial road segment information and historical surrounding road segment information; to extract features from the historical arterial road segment information to obtain historical arterial road segment feature information, and to extract features from the historical surrounding road segment information to obtain historical surrounding road segment feature information; to construct route data samples based on the arterial road segment feature information, historical surrounding road segment feature information, and historical route information; and to label the route data samples with arrival time tags to generate a route sample data set.

[0022] Furthermore, embodiments of this application also provide a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute any of the time prediction methods provided in embodiments of this application.

[0023] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program adapted for loading by a processor to execute any of the time prediction methods provided in embodiments of this application.

[0024] Furthermore, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements any of the time prediction methods provided in embodiments of this application.

[0025] Furthermore, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements any of the time prediction methods provided in embodiments of this application.

[0026] This application embodiment can obtain a route network of a target route from the starting point to the destination. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. Main nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between main nodes represent the positional relationship between the main road segments. Feature extraction is performed on the main nodes to obtain the main road segment feature information, and feature extraction is performed on the auxiliary nodes to obtain the surrounding road segment feature information. Based on the route network, the main road segment feature information and the surrounding road segment feature information are aggregated to obtain the aggregated feature information of the main nodes. Based on the aggregated feature information, time prediction processing is performed on the target route to obtain the arrival time of the target route from the starting point to the destination. Since this application embodiment can use the main road segment feature information and the surrounding road segment feature information extracted from the route network to aggregate the aggregated feature information of the main nodes, the arrival time of the target route can be accurately predicted based on the aggregated feature information of the main nodes. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of a scenario for the time prediction method provided in an embodiment of this application;

[0029] Figure 2 This is a flowchart illustrating the time prediction method provided in an embodiment of this application;

[0030] Figure 3 This is a schematic diagram provided in this application embodiment of aggregating the feature information of the main road segment and the feature information of the surrounding road segment according to the route network to obtain the aggregated feature information of the main node.

[0031] Figure 4 These are two schematic flowcharts of the time prediction method provided in the embodiments of this application;

[0032] Figure 5 This is a schematic diagram of the "fishbone diagram" provided in the embodiments of this application;

[0033] Figure 6 This is a schematic diagram of the structure of the post-training time prediction model provided in the embodiments of this application;

[0034] Figure 7This is a schematic diagram of the time prediction device provided in the embodiments of this application;

[0035] Figure 8 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation

[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] This application provides a time prediction method, apparatus, computer device, and computer-readable storage medium. The time prediction apparatus can be integrated into a computer device, which may be a server or a terminal, etc.

[0038] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) acceleration services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. The embodiments of this invention can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0039] This application relates to Artificial Intelligence (AI), which is the theory, method, technology, and application system for simulating, extending, and expanding human intelligence using digital computers or machines controlled by digital computers to perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0040] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0041] For example, see Figure 1 Taking the time prediction device integrated into a computer device as an example, the computer device can obtain the route network of the target route from the starting point to the destination; perform feature extraction on the main nodes to obtain the main road segment feature information, and perform feature extraction on the auxiliary nodes to obtain the surrounding road segment feature information; based on the route network, aggregate the main road segment feature information and the surrounding road segment feature information to obtain the aggregated feature information of the main nodes; and perform time prediction processing on the target route based on the aggregated feature information to obtain the arrival time of the target route from the starting point to the destination.

[0042] The route network includes master nodes, auxiliary nodes, edges between master nodes and auxiliary nodes, and edges between master nodes. Master nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between master nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between master nodes represent the positional relationship between the main road segments. The main road segment can refer to the main road segment from the starting point to the end point in the target route.

[0043] Among them, the main node carries the information of the main road section; the auxiliary node carries the information of the surrounding road sections.

[0044] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.

[0045] This embodiment will be described from the perspective of a time prediction device, which can be integrated into a computer device, such as a server or a terminal. The terminal can include tablet computers, laptops, personal computers (PCs), wearable devices, virtual reality devices, or other smart devices that can acquire data.

[0046] like Figure 2 As shown, the specific process of this time prediction method is as follows: steps S101 to S104:

[0047] S101. Obtain the route network of the target route from the starting point to the destination.

[0048] The route network includes master nodes, auxiliary nodes, edges between master nodes and auxiliary nodes, and edges between master nodes. Master nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between master nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between master nodes represent the positional relationship between the main road segments.

[0049] The method for obtaining the route network of the target route from the starting point to the destination in this embodiment of the application can be as follows:

[0050] For example, obtain at least one candidate route from the starting point to the destination; determine the target route from the at least one candidate route; and obtain the route network of the target route based on the target route.

[0051] For example, after a user determines a starting point and a destination, the computer device can return at least one candidate route between the starting point and the destination; in response to a selection operation for a target route, the computer device determines the target route from at least one candidate route. As another example, after a user determines a starting point and a destination, the computer device can obtain at least one candidate route between the starting point and the destination, and obtain the candidate arrival time for each candidate route from the starting point to the destination; based on the candidate arrival times, the target route is determined from at least one candidate route, wherein the candidate arrival time corresponding to the target route can be the time with the shortest arrival time among all candidate arrival times.

[0052] In this embodiment of the application, the method for obtaining the route network of the target route based on the target route can be as follows: based on the target route, obtain the main road segments of the target route and the surrounding road segments corresponding to the main road segments; based on the main road segments, determine the main nodes in the route network; based on the surrounding road segments, determine the auxiliary nodes in the route network; and based on the positional relationship between the main road segments and the positional relationship between the main road segments and the surrounding road segments, determine the edges in the route network; and construct the route network of the target route based on the main nodes, auxiliary nodes, and edges.

[0053] The main node carries information about the main road segment; the auxiliary nodes carry information about the surrounding road segments. The main road segment information can include at least one dimension, such as the basic attributes of the road segment, the speed of the road segment, and the condition of the road segment.

[0054] The basic attribute dimension of the main road segment information may include the road segment mileage, the road grade of the road segment, the road segment width, and whether there are traffic lights on the road segment. The road segment speed dimension of the main road segment information may include several speed parameters within a first preset time interval after departure. The speed parameters include the average speed and the standard deviation of the speed. The standard deviation of the speed refers to the standard deviation between the average speeds. Specifically, in this embodiment, the departure time t when the route navigation begins can be used as the basis. Within the first preset time interval [t-30min, t+30min], calculations are performed every 5 minutes to obtain 12 average speeds and 12 standard deviations of the main road segment within the first preset time interval. The main road segment information in the road segment condition dimension can include several road condition parameters within a second preset time interval after departure. These road condition parameters can include congestion parameters, several real-time speeds, and several real-time traffic flows. Specifically, in this embodiment, the departure time t when the route navigation begins can be used as a reference. Within the second preset time interval [t-30min, t], calculations are performed every 5 minutes to obtain 6 congestion parameters, 6 real-time speeds, and 6 real-time traffic flows for the main road segment within the second preset time interval.

[0055] Surrounding road information can include information on surrounding roads in at least one dimension. For example, the dimensions of surrounding road information can include the basic attributes of the road segment, the speed of the road segment, and the condition of the road segment. For details on surrounding road segment information in terms of basic attributes, speed, and condition, please refer to the description of the main road information, which will not be repeated here.

[0056] Each main road segment corresponds to a main node, and each surrounding road segment corresponds to an auxiliary node.

[0057] S102. Extract features from the main nodes to obtain the main road segment feature information, and extract features from the auxiliary nodes to obtain the surrounding road segment feature information.

[0058] In this embodiment, a neural network model can be used to extract features from the main nodes to obtain the main road segment feature information, and a neural network model can be used to extract features from the auxiliary nodes to obtain the surrounding road segment feature information.

[0059] The trunk road segment feature information in this application embodiment may include trunk road segment feature information of at least one dimension. The dimensions of the trunk road segment feature information may include a road segment basic attribute dimension, a road segment speed dimension, and a road segment condition dimension. The trunk road segment feature information may be information extracted from trunk road segment information. The trunk road segment feature information of the road segment basic attribute dimension may include road segment mileage feature information, road level feature information, road segment width feature information, and whether the road segment has traffic lights. The trunk road segment feature information of the road segment speed dimension may include feature information of several speed parameters within a first preset time interval after departure. The trunk road segment feature information of the road segment condition dimension may include feature information of several road condition parameters within a second preset time interval after departure.

[0060] The surrounding road segment feature information can include at least one dimension of surrounding road segment feature information. The dimensions of the surrounding road segment feature information can include the road segment basic attribute dimension, the road segment speed dimension, and the road segment condition dimension. For details on the surrounding road segment feature information of the road segment basic attribute dimension, the surrounding road segment feature information of the road segment speed dimension, and the surrounding road segment feature information of the road segment condition dimension, please refer to the above-mentioned main road segment feature information, which will not be repeated here.

[0061] In this embodiment of the application, the neural network model can be a graph convolutional neural network model.

[0062] S103. Based on the route network, aggregate the feature information of the main road segment and the feature information of the surrounding road segment to obtain the aggregated feature information of the main node.

[0063] like Figure 3 As shown in the embodiments of this application, the aggregation of the feature information of the main road segment and the feature information of the surrounding road segment based on the route network, to obtain the aggregated feature information of the main node, can be done in the manner shown in steps B1 to B2:

[0064] B1. Based on the route network, determine the neighboring nodes of each node.

[0065] The nodes include primary nodes and secondary nodes, and the neighboring nodes also include primary nodes and secondary nodes.

[0066] This application embodiment can determine the neighboring nodes of each node based on the topology of the route network. The neighboring nodes may include the node's first-order neighboring nodes, or the node's first-order and second-order neighboring nodes, or at least one of the node's first-order, second-order, and higher-order neighboring nodes.

[0067] In this context, a first-order neighbor node can be a node that is directly adjacent to a node; a second-order neighbor node can be a node that is directly adjacent to a first-order neighbor node of a node; a higher-order neighbor node refers to a third-order neighbor node or a node of higher order. A third-order neighbor node can be a node that is directly adjacent to a second-order neighbor node of a node, and so on.

[0068] B2. Aggregate the main road segment feature information of the main node in the neighboring nodes and the surrounding road segment feature information of the auxiliary nodes in the neighboring nodes to obtain the aggregated feature information of the main node.

[0069] In this application embodiment, there are several ways to aggregate the main road segment feature information of the primary node among neighboring nodes and the surrounding road segment feature information of the secondary nodes among neighboring nodes to obtain the aggregated feature information of the primary node:

[0070] For example, for each node, the main road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes are aggregated to obtain the aggregated feature information of the node; from the aggregated feature information of the node, the aggregated feature information of the primary node is extracted.

[0071] In other words, this application embodiment uses the main road segment feature information of the primary node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes to update the original information of the node.

[0072] For example, for each node, the main road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes, as well as the node's road segment feature information, are aggregated to obtain the node's aggregated feature information. The node's road segment feature information is either the main road segment feature information of the primary node or the surrounding road segment feature information of the secondary node. The aggregated feature information of the primary node is extracted from the node's aggregated feature information.

[0073] In other words, this application embodiment uses the main road segment feature information of the primary node among the neighboring nodes and the surrounding road segment feature information of the auxiliary nodes among the neighboring nodes, as well as the original road segment feature information of the node, to update the original road segment feature information of the node.

[0074] In this embodiment of the application, since each node has a corresponding node identifier, the method for extracting the aggregated feature information of the main node from the aggregated feature information of the nodes in this embodiment of the application can be: extracting the aggregated feature information of the main node from the aggregated feature information of the nodes according to the node identifier of the main node.

[0075] S104. Based on the aggregated feature information, perform time prediction processing on the target route to obtain the arrival time of the target route from the starting point to the destination.

[0076] For example, embodiments of this application can utilize existing neural network prediction models in related technologies to perform time prediction processing on the target route based on aggregated feature information, thereby obtaining the arrival time of the target route from the starting point to the destination.

[0077] For example, embodiments of this application can utilize a post-trained time prediction model to perform time prediction processing on a target route based on aggregated feature information, thereby obtaining the arrival time of the target route from the starting point to the destination. The post-trained time prediction model can refer to a neural network model obtained by training the time prediction model to be trained. The post-trained time prediction model can be a graph convolutional neural network model.

[0078] This application embodiment can utilize a trained time prediction model to aggregate the feature information of main road segments and surrounding road segments based on the route network. Specifically, before aggregating the feature information of main road segments and surrounding road segments, this application embodiment can train the time prediction model to be trained, as follows:

[0079] For example, obtain a route sample dataset, which includes route data samples labeled with arrival time tags; based on the route data samples, use the time prediction model to be trained to predict the arrival time; based on the predicted arrival time and arrival time tags, converge the model parameters of the time prediction model to be trained to obtain the trained time prediction model.

[0080] In this embodiment of the application, the method for obtaining the route sample data set may be as follows: obtaining historical route information, which includes historical trunk road segment information and historical surrounding road segment information; extracting features from the historical trunk road segment information to obtain historical trunk road segment feature information, and extracting features from the historical surrounding road segment information to obtain historical surrounding road segment feature information; constructing route data samples based on the trunk road segment feature information, historical surrounding road segment feature information, and historical route information; and labeling the route data samples with arrival time tags to generate the route sample data set.

[0081] In this embodiment, a loss function can be used to calculate the loss value between the predicted arrival time and the arrival time label; based on the loss value, the model parameters of the time prediction model to be trained are converged to obtain the trained time prediction model. The loss function can be a mean squared error loss function, a logarithmic loss function, an absolute value loss function, etc.

[0082] This application embodiment can obtain a route network of a target route from the starting point to the destination. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. Main nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between main nodes represent the positional relationship between the main road segments. Feature extraction is performed on the main nodes to obtain the main road segment feature information, and feature extraction is performed on the auxiliary nodes to obtain the surrounding road segment feature information. Based on the route network, the main road segment feature information and the surrounding road segment feature information are aggregated to obtain the aggregated feature information of the main nodes. Based on the aggregated feature information, time prediction processing is performed on the target route to obtain the arrival time of the target route from the starting point to the destination. Since this application embodiment can use the main road segment feature information and the surrounding road segment feature information extracted from the route network to aggregate the aggregated feature information of the main nodes, the arrival time of the target route can be accurately predicted based on the aggregated feature information of the main nodes.

[0083] Based on the method described in the above embodiments, the following examples will provide further detailed explanations.

[0084] In this embodiment, the time prediction device is specifically integrated into a computer device, which can be a server or a terminal.

[0085] like Figure 4 As shown, a time prediction method is described, with the specific process as follows: steps S201 to S207:

[0086] S201, Computer equipment acquires a set of route sample data.

[0087] The route sample data set includes route data samples labeled with arrival time tags.

[0088] The method for obtaining the route sample data set in this embodiment of the application can be as follows: obtaining historical route information, which includes historical trunk road segment information and historical surrounding road segment information; extracting features from the historical trunk road segment information to obtain historical trunk road segment feature information, and extracting features from the historical surrounding road segment information to obtain historical surrounding road segment feature information; constructing route data samples based on trunk road segment feature information, historical surrounding road segment feature information, and historical route information; and labeling the route data samples with arrival time tags to generate the route sample data set.

[0089] Among them, historical route information can refer to route information within a historical period; historical main road section information can refer to road section information within a historical period; and historical surrounding road section information can refer to road section information within a historical period.

[0090] In this embodiment, the historical route information also includes historical arterial road segments and historical surrounding road segments, as well as the positional relationships between historical arterial road segments and between historical arterial road segments and historical surrounding road segments. In this embodiment, the positional relationships between historical arterial road segments and historical surrounding road segments, and between historical arterial road segments and historical surrounding road segments, can be represented as a "fishbone diagram," as shown in the example below. Figure 5 As shown.

[0091] In this embodiment of the application, a time prediction model to be trained can be used to extract features from historical main road segment information to obtain historical main road segment feature information, and a time prediction model to be trained can be used to extract features from historical surrounding road segment information to obtain historical surrounding road segment feature information.

[0092] The historical arterial road segment feature information can include at least one dimension of historical arterial road segment feature information; the dimensions of the historical arterial road segment feature information can include the basic attribute dimension, the speed dimension, and the condition dimension. For details on the historical arterial road segment feature information in the basic attribute dimension, the speed dimension, and the condition dimension, please refer to the description of arterial road segment feature information above, which will not be repeated here.

[0093] Historical surrounding road segment feature information may include at least one dimension of historical surrounding road segment feature information; the dimensions of historical surrounding road segment feature information may include the road segment basic attribute dimension, the road segment speed dimension, and the road segment condition dimension. For details on the historical surrounding road segment feature information of the road segment basic attribute dimension, the historical surrounding road segment feature information of the road segment speed dimension, and the historical surrounding road segment feature information of the road segment condition dimension, please refer to the description of surrounding road segment feature information mentioned above, which will not be repeated here.

[0094] The arrival time tag can refer to the actual arrival time from the starting point to the destination of the route within a historical period.

[0095] In this embodiment of the application, the historical surrounding road segment feature information can be represented in the form of a vector, and the dimension of the vector can be K-dimensional.

[0096] S202. The computer equipment uses a time prediction model to be trained based on the route data sample to predict the arrival time.

[0097] In this embodiment, the time prediction model to be trained can be a graph convolutional neural network model. The time prediction model to be trained includes a feature extraction layer, a pooling layer, and an output layer.

[0098] This application embodiment can determine the neighboring road segments of each historical road segment based on the "fishbone diagram" in the historical route information. The historical road segments include historical trunk road segments and historical surrounding road segments. The neighboring road segments can include first-order neighboring road segments of the historical road segment, or first-order neighboring road segments and second-order neighboring road segments of the historical road segment, or at least one of the following: first-order neighboring road segments, second-order neighboring road segments, and higher-order neighboring road segments of the historical road segment.

[0099] Among them, a first-order neighbor segment can refer to a segment directly adjacent to a historical segment; a second-order neighbor segment can refer to a segment directly adjacent to a first-order neighbor segment of a historical segment; a higher-order neighbor segment refers to a third-order neighbor segment or higher, and a third-order neighbor segment can refer to a segment directly adjacent to a second-order neighbor segment of a historical segment, and so on.

[0100] This application embodiment can employ a feature extraction layer of the time prediction model to be trained. Based on the neighboring road segments of each historical road segment, it aggregates the historical trunk road segment feature information of the historical arterial road segments among the neighboring road segments and the historical surrounding road segment feature information of the historical peripheral road segments among the neighboring road segments to obtain the aggregated historical feature information of each historical road segment. Here, the vector dimension of the historical feature information of each historical road segment changes from K dimensions to L dimensions, where the historical feature information is either historical peripheral road segment feature information or historical arterial road segment feature information.

[0101] Based on the above, this embodiment of the application can extract the aggregated historical feature information of the main road segments from the aggregated historical feature information of the historical road segments based on the identification information of the historical main road segments. Next, this embodiment of the application can use the pooling layer of the time prediction model to be trained to perform pooling processing on the aggregated historical feature information of the main road segments, obtaining pooled historical feature information, wherein the pooled historical feature information can be represented by a fixed-length vector of length L. Then, this embodiment of the application can use the output layer of the time prediction model to be trained to predict the predicted arrival time based on the pooled historical feature information; wherein the output layer can be a fully connected layer.

[0102] S203. The computer equipment converges the model parameters of the time prediction model to be trained based on the predicted arrival time and arrival time label, and obtains the trained time prediction model.

[0103] This application embodiment can use a loss function to calculate the loss value between the predicted arrival time and the arrival time label; based on the loss value, the model parameters of the time prediction model to be trained are converged to obtain the trained time prediction model. The loss function can be a squared loss function, an absolute value loss function, etc.

[0104] The squared loss function can be found in formula (1), and the absolute value loss function can be found in formula (2).

[0105] loss(ETA, ATA) = (ETA - ATA) 2 Formula (1)

[0106] loss(ETA, ATA)=|ETA-ATA| Formula (2)

[0107] In formulas (1) and (2), loss(ETA, ATA) represents the loss value; ETA represents the predicted arrival time; and ATA represents the arrival time label.

[0108] In this embodiment, gradient descent can be used to train the entire training time prediction model end-to-end. The model parameters include the matrix of the feature extraction layer and the weight parameters of the output layer; the feature extraction layer can be understood as a graph convolutional neural network layer.

[0109] S204. The computer equipment obtains the route network of the target route from the starting point to the destination.

[0110] The route network includes master nodes, auxiliary nodes, edges between master nodes and auxiliary nodes, and edges between master nodes. Master nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between master nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between master nodes represent the positional relationship between the main road segments.

[0111] Among them, the main node carries the information of the main road section; the auxiliary node carries the information of the surrounding road sections.

[0112] In this embodiment of the application, the method for obtaining the route network of the target route from the starting point to the end point may be as follows: obtaining at least one candidate route from the starting point to the end point; determining the target route from the at least one candidate route; and obtaining the route network of the target route based on the target route.

[0113] In this embodiment of the application, the method for obtaining the route network of the target route based on the target route can be as follows: based on the target route, obtain the main road segments of the target route and the surrounding road segments corresponding to the main road segments; based on the main road segments, determine the main nodes in the route network; based on the surrounding road segments, determine the auxiliary nodes in the route network; and based on the positional relationship between the main road segments and the positional relationship between the main road segments and the surrounding road segments, determine the edges in the route network; and construct the route network of the target route based on the main nodes, auxiliary nodes, and edges.

[0114] S205. The computer equipment extracts features from the main nodes to obtain the main road segment feature information, and extracts features from the auxiliary nodes to obtain the surrounding road segment feature information.

[0115] In this embodiment, a neural network model can be used to extract features from the main nodes to obtain the main road segment feature information, and a neural network model can also be used to extract features from the auxiliary nodes to obtain the surrounding road segment feature information. The neural network model here can be a post-trained time prediction model. For details on the post-trained time prediction model, please refer to [link to relevant documentation]. Figure 6 The post-training time prediction model includes a feature extraction layer, a pooling layer, and an output layer. In this embodiment, the feature extraction layer of the post-training time prediction model extracts features from the route network to obtain the main road segment feature information and the surrounding road segment feature information.

[0116] The trunk road segment feature information in this application embodiment includes trunk road segment feature information of at least one dimension. The dimensions of the trunk road segment feature information include the road segment basic attribute dimension, the road segment speed dimension, and the road segment condition dimension. For details, please refer to the foregoing description of trunk road segment feature information, which will not be repeated here.

[0117] The surrounding road segment feature information in this application embodiment includes surrounding road segment feature information of at least one dimension. The dimensions of the surrounding road segment feature information include the road segment basic attribute dimension, the road segment speed dimension, and the road segment condition dimension. For details, please refer to the foregoing description of the surrounding road segment feature information, which will not be repeated here.

[0118] In this embodiment, the feature information of the main road segment corresponding to each main node can be represented in the form of a vector, and the dimension of the feature information of each main road segment can be K-dimensional; the feature information of the surrounding road segments corresponding to each auxiliary node can be represented in the form of a vector, and the dimension of the feature information of each auxiliary road segment can be K-dimensional.

[0119] S206. The computer equipment aggregates the feature information of the main road segment and the feature information of the surrounding road segment according to the route network to obtain the aggregated feature information of the main node.

[0120] In this embodiment, the feature extraction layer of the trained time prediction model can aggregate the feature information of the main road segment and the feature information of the surrounding road segment according to the route network to obtain the aggregated feature information of the main node.

[0121] According to the embodiments of this application, the method of aggregating the feature information of the main road segment and the feature information of the surrounding road segment based on the route network to obtain the aggregated feature information of the main node can be as follows: the computer device can determine the neighboring nodes of each node according to the route network; the main road segment feature information of the main node in the neighboring nodes and the surrounding road segment feature information of the auxiliary nodes in the neighboring nodes are aggregated to obtain the aggregated feature information of the main node.

[0122] The nodes include primary nodes and secondary nodes, and the neighboring nodes also include primary nodes and secondary nodes.

[0123] In this application embodiment, there are several ways to aggregate the main road segment feature information of the primary node among neighboring nodes and the surrounding road segment feature information of the secondary nodes among neighboring nodes to obtain the aggregated feature information of the primary node:

[0124] For example, for each node, a feature extraction layer is used to aggregate the main road segment feature information of the main node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes to obtain the aggregated feature information of the node; from the aggregated feature information of the node, the aggregated feature information of the main node is extracted.

[0125] Specifically, the aggregation of the main road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes of a node can be found in formula (3):

[0126]

[0127] in, represents the aggregated feature information of the node; v represents the node; k represents the Kth feature extraction layer in the post-trained time prediction model, which can be understood as the Kth graph convolutional layer in the post-trained time prediction model; N(v) represents the neighboring nodes of node v. This represents the vector of the target node u among the neighboring nodes extracted by the (K-1)th feature extraction layer in the time prediction model after training. The vector of the target node u among the neighboring nodes is the main road segment feature information of the main node among the neighboring nodes or the surrounding road segment feature information of the auxiliary node among the neighboring nodes.

[0128] Specifically, the meaning of formula (3) can be understood as follows: for each node, the vectors of the neighboring nodes corresponding to the node are averaged to obtain the first average value; the first average value is assigned to the node; the first average value is the aggregated feature information of the node.

[0129] For example, for each node, the main road segment feature information of the primary node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes, as well as the node's road segment feature information, are aggregated to obtain the node's aggregated feature information. The node's road segment feature information is either the main road segment feature information of the primary node or the surrounding road segment feature information of the auxiliary node. The aggregated feature information of the primary node is extracted from the node's aggregated feature information.

[0130] Specifically, the aggregation of the main road segment feature information of the primary node among the neighboring nodes of a node, the surrounding road segment feature information of the auxiliary nodes among the neighboring nodes, and the road segment feature information of the node can be found in formula (4), formula (5), or formula (6):

[0131]

[0132] In formula (4), v, k, N(v), The explanation can be found in formula (3). Specifically, the meaning of formula (4) can be understood as follows: for each node, the vectors of the neighboring nodes and the node itself are averaged to obtain the second average value; the second average value is assigned to the node; the second average value is the aggregated feature information of the node.

[0133]

[0134] In formula (5), v, k, N(v), For an explanation, please refer to formula (3); W k The matrix in the Kth feature extraction layer of the time prediction model after training represents the matrix; σ represents the activation function, which can be the sigmoid function, the ReLU function, etc. Specifically, the meaning of formula (5) can be understood as follows: for each node, the vectors of the corresponding neighboring nodes are averaged to obtain the third average value; then, the third average value is compared with the matrix W. k The average value after the first processing is obtained by multiplication. Then, the average value after the first processing is mapped by the activation function σ to obtain the first mapped value. The first mapped value is the aggregated feature information of the node.

[0135]

[0136] In formula (6), v, k, N(v), For an explanation, please refer to formula (3); in formula (6), W kThe activation function represented by σ can be found in formula (5). Specifically, formula (6) can be understood as follows: for each node, the vectors of the node's neighboring nodes and the node's own vector are averaged to obtain the fourth average value; then, the fourth average value is compared with matrix W. k The average value after the second processing is obtained by multiplying the values. Then, the average value after the second processing is mapped using the activation function σ to obtain the second mapped value. The second mapped value is the aggregated feature information of the node.

[0137] In this embodiment, the route network remains unchanged after aggregation; only the feature information of each node undergoes dimensional transformation. The feature information of a node is either the feature information of the main road segment or the feature information of surrounding road segments. This embodiment transforms the dimension of the node's feature information from K-dimensional to L-dimensional.

[0138] S207. The computer equipment performs time prediction processing on the target route based on the aggregated feature information to obtain the arrival time of the target route from the starting point to the destination.

[0139] For example, embodiments of this application can utilize a trained time prediction model to perform time prediction processing on a target route based on aggregated feature information, thereby obtaining the arrival time of the target route from the starting point to the destination.

[0140] In this embodiment, the pooling layer of the trained time prediction model can be used to pool the aggregated feature information of the master node to obtain pooled feature information. The obtained pooled feature information is a fixed-length vector of length L. In this embodiment, the pooling processing of the aggregated feature information of the master node can be performed by max pooling or average pooling. Max pooling can be performed according to formula (7), and average pooling can be performed according to formula (8):

[0141] g = max({g a , a∈b}) Formula (7)

[0142] Where g represents the feature information after pooling; g a Describe the aggregated feature information of the master nodes; 'a' represents the master node; 'b' represents the set of master nodes, which includes several master nodes.

[0143] Specifically, since the aggregated feature information of each master node can be represented as an L-dimensional vector, based on this, formula (7) can be interpreted as extracting the maximum value corresponding to the dimension from the elements corresponding to the dimension in the aggregated feature information of all master nodes for each dimension, and obtaining the maximum value corresponding to each dimension; thus, based on the maximum value corresponding to each dimension, the pooled feature information is constructed, and the pooled feature information is a vector of length L.

[0144] g = mean({g a , a∈b}) Formula (8)

[0145] Among them, g, g a a and b can be found in the description of formula (7). Specifically, since the aggregated feature information of each master node can be represented as an L-dimensional vector, formula (8) can be interpreted as calculating the average value of all elements in each dimension based on the elements corresponding to the dimension in the aggregated feature information of all master nodes, and thus obtaining the average value of the elements corresponding to each dimension; in this way, the pooled feature information is constructed based on the average value of the elements corresponding to each dimension, and the pooled feature information is a vector of length L.

[0146] In addition to the above, other pooling methods may also be used in the embodiments of this application, such as minimum pooling, geometric mean pooling, etc.

[0147] Then, in this embodiment of the application, the pooled feature information is processed by the output layer of the trained time prediction model to obtain the arrival time of the target route from the starting point to the destination.

[0148] Based on the above, the embodiments of this application utilize the feature information of surrounding road segments to provide more information to the post-trained time prediction model, thereby improving the prediction accuracy of the post-trained time prediction model.

[0149] The embodiments of this application can be applied to the following, but not limited to, situations:

[0150] (1) After the user enters the navigation state through the map software, the remaining time of the target route is calculated at regular intervals using the embodiments of this application, so as to facilitate the user to arrange the trip.

[0151] (2) Using the embodiments of this application, an isochronous reachable circle is calculated at the user's location, such as a half-hour reachable circle or a one-hour reachable circle, so that the user can understand the living radius of the location.

[0152] (3) When delivering food, the time taken for each target route can be accurately calculated using the embodiments of this application, so as to better assign orders to delivery personnel and improve delivery efficiency.

[0153] (4) In the context of ride-hailing, the time taken for each target route can be accurately calculated using the embodiments of this application, thereby better enabling drivers to accept orders and improving their passenger transport efficiency.

[0154] (5) The arrival time predicted in this application embodiment can be provided to the system's evaluation service for evaluation processing to evaluate the merits of each target route, and then push the optimal target route to the user.

[0155] This application embodiment can obtain a route network of a target route from the starting point to the destination. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. Main nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments in the target route, edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between main nodes represent the positional relationship between the main road segments. Feature extraction is performed on the main nodes to obtain the main road segment feature information, and feature extraction is performed on the auxiliary nodes to obtain the surrounding road segment feature information. Based on the route network, the main road segment feature information and the surrounding road segment feature information are aggregated to obtain the aggregated feature information of the main nodes. Based on the aggregated feature information, time prediction processing is performed on the target route to obtain the arrival time of the target route from the starting point to the destination. Since this application embodiment can use the main road segment feature information and the surrounding road segment feature information extracted from the route network to aggregate the aggregated feature information of the main nodes, the arrival time of the target route can be accurately predicted based on the aggregated feature information of the main nodes.

[0156] To better implement the above methods, embodiments of this application also provide a time prediction device, which can be integrated into a computer device, such as a server or terminal, and the terminal may include a tablet computer, a laptop computer, and / or a personal computer.

[0157] For example, such as Figure 7 As shown, the time prediction device may include an acquisition unit 301, an extraction unit 302, an aggregation unit 303, a prediction unit 304, and a training unit 305, as follows:

[0158] (1) Obtain unit 301;

[0159] The acquisition unit 301 can be used to acquire the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments.

[0160] In some embodiments, the acquisition unit 301 may be specifically used to acquire at least one candidate route from the starting point to the ending point; determine the target route from the at least one candidate route; and acquire the route network of the target route based on the target route.

[0161] In some embodiments, the acquisition unit 301 may be specifically used to acquire the main road segment of the target route and the surrounding road segments corresponding to the main road segment; determine the main node in the route network based on the main road segment; determine the auxiliary node in the route network based on the surrounding road segments; and determine the edge in the route network based on the positional relationship between the main road segments and the positional relationship between the main road segment and the surrounding road segments; and construct the route network of the target route based on the main node, auxiliary node and edge.

[0162] (2) Extraction unit 302;

[0163] The extraction unit 302 can be used to extract features from the main nodes to obtain the main road segment feature information, and to extract features from the auxiliary nodes to obtain the surrounding road segment feature information.

[0164] (3) Aggregation unit 303;

[0165] The aggregation unit 303 can be used to aggregate the feature information of the main road segment and the feature information of the surrounding road segment according to the route network, so as to obtain the aggregated feature information of the main node.

[0166] In some embodiments, the aggregation unit 303 can be specifically used to determine the neighboring nodes of each node according to the route network. The nodes include primary nodes and secondary nodes, and the neighboring nodes include primary nodes and secondary nodes. The aggregation processing is performed on the trunk road segment feature information of the primary node and the surrounding road segment feature information of the secondary node in the neighboring nodes to obtain the aggregated feature information of the primary node.

[0167] In some embodiments, the aggregation unit 303 can be specifically used to aggregate the main road segment feature information of the main node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes of each node to obtain the aggregated feature information of the node; and extract the aggregated feature information of the main node from the aggregated feature information of the node.

[0168] In some embodiments, the aggregation unit 303 can be specifically used to aggregate the main road segment feature information of the main node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes of each node, as well as the road segment feature information of the node, to obtain the aggregated feature information of the node. The road segment feature information of the node is the main road segment feature information of the main node or the surrounding road segment feature information of the auxiliary node. The aggregated feature information of the main node is extracted from the aggregated feature information of the node.

[0169] (4) Prediction unit 304;

[0170] The prediction unit 304 can be used to perform time prediction processing on the target route based on the aggregated feature information to obtain the arrival time of the target route from the starting point to the end point.

[0171] (5) Training Unit 305;

[0172] The aggregated feature information is obtained by aggregating the information using the trained time prediction model; the training unit 305 can be used to obtain a route sample data set, which includes route data samples labeled with arrival time tags; based on the route data samples, the time prediction model to be trained is used to predict the arrival time; based on the predicted arrival time and arrival time tags, the model parameters of the time prediction model to be trained are converged to obtain the trained time prediction model.

[0173] In some embodiments, the training unit 305 can be specifically used to acquire historical route information, which includes historical arterial road segment information and historical surrounding road segment information of the historical arterial road segment information; to extract features from the historical arterial road segment information to obtain historical arterial road segment feature information, and to extract features from the historical surrounding road segment information to obtain historical surrounding road segment feature information; to construct route data samples based on the arterial road segment feature information, historical surrounding road segment feature information, and historical route information; and to label the route data samples with arrival time tags to generate a route sample data set.

[0174] As can be seen from the above, the acquisition unit 301 of this application embodiment can be used to acquire the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments. The extraction unit 302 can be used to extract features from the main nodes to obtain the main road segment feature information, and to extract features from the auxiliary nodes to obtain... The aggregation unit 303 can be used to aggregate the feature information of the main road segment and the feature information of the surrounding road segment according to the route network to obtain the aggregated feature information of the main node; the prediction unit 304 can be used to perform time prediction processing on the target route according to the aggregated feature information to obtain the arrival time of the target route from the starting point to the end point; since the embodiments of this application can use the feature information of the main road segment and the feature information of the surrounding road segment extracted from the route network to aggregate the feature information of the main node, the arrival time of the target route can be accurately predicted according to the aggregated feature information of the main node.

[0175] This application also provides a computer device, such as... Figure 8 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:

[0176] The computer device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 8 The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0177] The processor 401 is the control center of the computer device, connecting various parts of the computer device through various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and computer programs, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0178] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, computer programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0179] The computer device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0180] The computer device may also include an input unit 404, which can be used to receive input digital or character information communication, and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0181] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the computer device loads the executable files corresponding to the processes of one or more computer programs into the memory 402 according to the following instructions, and the processor 401 runs the computer programs stored in the memory 402 to realize various functions, as follows:

[0182] Obtain the route network of the target route from the origin to the destination. The route network includes master nodes, auxiliary nodes, edges between master nodes and auxiliary nodes, and edges between master nodes. Master nodes represent the main road segments in the target route, auxiliary nodes represent the surrounding road segments of the main road segments, edges between master nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and edges between master nodes represent the positional relationship between the main road segments. Extract features from the master nodes to obtain the main road segment features, and extract features from the auxiliary nodes to obtain the surrounding road segment features. Based on the route network, aggregate the main road segment features and the surrounding road segment features to obtain the aggregated feature information of the master nodes. Based on the aggregated feature information, perform time prediction processing on the target route to obtain the arrival time from the origin to the destination.

[0183] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0184] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by a computer program, or by a computer program controlling related hardware. The computer program can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0185] Therefore, embodiments of this application provide a computer-readable storage medium storing a computer program that can be loaded by a processor to execute any of the time prediction methods provided in embodiments of this application.

[0186] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0187] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0188] Since the instructions stored in the computer-readable storage medium can execute the steps of any of the time prediction methods provided in the embodiments of this application, the beneficial effects that any of the time prediction methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0189] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations of the above embodiments.

[0190] The above provides a detailed description of a time prediction method, computer device, apparatus, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A time prediction method, characterized in that, include: Obtain the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments. Feature extraction is performed on the main node to obtain the main road segment feature information, and feature extraction is performed on the auxiliary node to obtain the surrounding road segment feature information. Based on the route network, the feature information of the main road segment and the feature information of the surrounding road segment are aggregated to obtain the aggregated feature information of the main node; This step includes: for each node in the route network, averaging the vectors of the node's neighboring nodes to obtain a third average value; Multiply the third average value by the matrix to obtain the first processed average value; the matrix is ​​the matrix in the feature extraction layer of the trained time prediction model; the first processed average value is mapped using an activation function to obtain a first mapped value, which is the aggregated feature information of the node; the aggregated feature information of the main node is extracted from the aggregated feature information of the node. Based on the aggregated feature information, time prediction processing is performed on the target route to obtain the arrival time of the target route from the starting point to the destination.

2. The time prediction method according to claim 1, characterized in that, The aggregation process, based on the route network, of the main road segment feature information and the surrounding road segment feature information to obtain the aggregated feature information of the main node includes: Based on the route network, the neighboring nodes of each node are determined, wherein each node includes a primary node and a secondary node, and the neighboring nodes include a primary node and a secondary node; The main road segment feature information of the main node in the neighboring nodes and the surrounding road segment feature information of the auxiliary nodes in the neighboring nodes are aggregated to obtain the aggregated feature information of the main node.

3. The time prediction method according to claim 2, characterized in that, The aggregation process of the main road segment feature information of the primary node among the neighboring nodes and the surrounding road segment feature information of the secondary nodes among the neighboring nodes to obtain the aggregated feature information of the primary node includes: For each node, the main road segment feature information of the primary node and the surrounding road segment feature information of the secondary node among the neighboring nodes are aggregated to obtain the aggregated feature information of the node. The aggregated feature information of the master node is extracted from the aggregated feature information of the node.

4. The time prediction method according to claim 2, characterized in that, The aggregation process of the main road segment feature information of the primary node among the neighboring nodes and the surrounding road segment feature information of the secondary nodes among the neighboring nodes to obtain the aggregated feature information of the primary node includes: For each node, the main road segment feature information of the main node and the surrounding road segment feature information of the auxiliary node among the neighboring nodes of the node, as well as the road segment feature information of the node itself, are aggregated to obtain the aggregated feature information of the node. The road segment feature information of the node is either the main road segment feature information of the main node or the surrounding road segment feature information of the auxiliary node. The aggregated feature information of the master node is extracted from the aggregated feature information of the node.

5. The time prediction method according to claim 1, characterized in that, The route network for obtaining the target route from the starting point to the destination includes: Obtain at least one candidate route from the starting point to the destination; Determine the target route from at least one of the candidate routes; Based on the target route, obtain the route network of the target route.

6. The time prediction method according to claim 5, characterized in that, The step of obtaining the route network of the target route based on the target route includes: Based on the target route, obtain the main road segment of the target route and the surrounding road segments corresponding to the main road segment; Based on the main road segments, determine the main nodes in the route network; based on the surrounding road segments, determine the auxiliary nodes in the route network; and based on the positional relationships between the main road segments and between the main road segments and the surrounding road segments, determine the edges in the route network. The route network of the target route is constructed based on the main node, the auxiliary node, and the edge.

7. The time prediction method according to claim 1, characterized in that, The aggregated feature information is obtained by aggregating the information using a trained time prediction model; before aggregating the feature information of the main road segment and the feature information of the surrounding road segment according to the route network, the method further includes: Obtain a route sample data set, which includes route data samples labeled with arrival time tags; Based on the route data sample, a time prediction model to be trained is used to predict the arrival time. Based on the predicted arrival time and the arrival time label, the model parameters of the time prediction model to be trained are converged to obtain the trained time prediction model.

8. The time prediction method according to claim 7, characterized in that, The acquired route sample data set includes: Obtain historical route information, which includes historical trunk road information and historical surrounding road information of the historical trunk road information; Feature extraction is performed on the historical main road segment information to obtain historical main road segment feature information, and feature extraction is performed on the historical surrounding road segment information to obtain historical surrounding road segment feature information; A route data sample is constructed based on the main road segment feature information, the historical surrounding road segment feature information, and the historical route information; The route data samples are labeled with arrival time tags to generate the route sample data set.

9. A time prediction device, characterized in that, include: The acquisition unit is used to acquire the route network of the target route from the starting point to the end point. The route network includes main nodes, auxiliary nodes, edges between main nodes and auxiliary nodes, and edges between main nodes. The main nodes represent the main road segments in the target route, the auxiliary nodes represent the surrounding road segments of the main road segments in the target route, the edges between main nodes and auxiliary nodes represent the positional relationship between the main road segments and the surrounding road segments, and the edges between main nodes represent the positional relationship between the main road segments. The extraction unit is used to extract features from the main node to obtain the main road segment feature information, and to extract features from the auxiliary node to obtain the surrounding road segment feature information. An aggregation unit is used to aggregate the feature information of the main road segment and the feature information of the surrounding road segments according to the route network to obtain the aggregated feature information of the main node. Specifically, the aggregation unit is used to: for each node in the route network, average the vectors of the node's corresponding neighboring nodes to obtain a third average value; multiply the third average value by a matrix to obtain a first processed average value; the matrix is ​​a matrix in the feature extraction layer of the trained time prediction model; map the first processed average value using an activation function to obtain a first mapped value, which is the aggregated feature information of the node; and extract the aggregated feature information of the main node from the aggregated feature information of the node. The prediction unit is used to perform time prediction processing on the target route based on the aggregated feature information to obtain the arrival time of the target route from the starting point to the destination.

10. A computer device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the time prediction method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted for loading by a processor to execute the time prediction method according to any one of claims 1 to 8.

12. A computer program product, characterized in that, The computer program product stores a computer program adapted for loading by a processor to execute the time prediction method according to any one of claims 1 to 8.

Images