Data processing method and device for driving time prediction, equipment and medium

Abstract

The application provides a data processing method, device and equipment for driving time prediction and a medium. Traffic feature maps corresponding to driving areas and road network data maps are obtained. The traffic feature maps and the road network data maps are respectively divided and clustered to obtain multi-class traffic feature target maps and road network data target maps. Source domains and target domains are determined from the traffic feature target maps and the road network data target maps. A graph convolution network is trained according to data of the source domains to obtain an initial training model. The initial training model is trained according to data of the target domains to obtain a target training model. The application realizes accurate prediction of the driving time of a mine car by training a convolution network on historical driving trajectories and road network data, thereby solving the technical problem that the prior art cannot predict the driving time of a mine car in an open-pit mine area that has not obtained driving features and road network data.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more particularly to a data processing method, apparatus, device, and medium for predicting driving time. Background Technology

[0002] As more and more open-pit mines rapidly develop towards larger equipment, more information-based management, and more intelligent production, production scheduling, as the central control link in open-pit mine production, has always received much attention. With the increasing maturity of the Global Positioning System (GPS) technology, GPS technology has been applied to newly developed truck scheduling systems, enabling open-pit mine production systems to conveniently schedule and monitor mining trucks.

[0003] Current predictions of vehicle travel time primarily focus on urban transportation systems. However, open-pit mines have complex terrain and fluctuating mining processes, making the prediction of mine car travel time in new open-pit mine areas a significant challenge. Currently, there is no existing technology capable of accurately predicting the travel time of open-pit mine cars. Summary of the Invention

[0004] To address the technical problem that existing technologies cannot accurately predict the travel time of mine cars in new open-pit mine areas, this invention provides a data processing method, apparatus, equipment, and medium for predicting travel time.

[0005] In a first aspect, the present invention provides a data processing method for predicting travel time, comprising:

[0006] Obtain traffic feature map and road network data map corresponding to the driving area, wherein the feature points in the traffic feature map include feature information related to driving time;

[0007] The traffic feature map and the road network data map are divided into traffic feature sub-maps and road network data sub-maps, respectively.

[0008] Clustering is performed on the traffic feature sub-map and the road network data sub-map respectively to obtain multi-class traffic feature target maps and road network data target maps;

[0009] A first source domain and a first target domain are determined from the traffic feature target map, and a second source domain and a second target domain are determined from the road network data target map;

[0010] The graph convolutional network is trained based on the data from the first source domain and the second source domain to obtain an initial training model;

[0011] The initial training model is trained based on the data from the first target domain and the second target domain to obtain the target training model.

[0012] Furthermore, obtain the traffic feature map and road network data map corresponding to the driving area, including:

[0013] The first feature information of the trajectory points is stored in the nodes of the traffic feature graph, and the second feature information between adjacent trajectory points is stored in the edges of the traffic feature graph; and

[0014] The sequence number of the path is stored in the node of the road network data graph, and the connectivity information between the paths is stored in the edge of the road network data graph;

[0015] The first feature information includes time, longitude, and latitude, while the second feature information includes weather information, flight schedule information, average vehicle speed, and traffic flow.

[0016] Furthermore, the traffic feature map and the road network data map are divided separately, including:

[0017] The traffic feature map is divided according to the second feature information between adjacent trajectory points; and

[0018] The road network data map is divided according to the connection information between the paths.

[0019] Furthermore, clustering is performed on the traffic feature subgraph and the road network data subgraph, respectively, including:

[0020] Clustering is performed on the traffic feature sub-map based on the first feature information of the trajectory points and the second feature information between adjacent trajectory points; and

[0021] The road network data subgraph is clustered based on the sequence number of the path and the connectivity information between the paths.

[0022] Further, determining the first source domain and the first target domain from the traffic feature target map includes:

[0023] The traffic feature target map with the largest amount of feature data in each type of traffic feature target map is determined as the first source domain, and the other traffic feature target maps in each type of traffic feature target map, excluding the first source domain, are determined as the first target domain; and

[0024] Determining the second source domain and the second target domain from the road network data target map includes:

[0025] The road network data target map containing the most connected paths in each type of road network data target map is determined as the first source domain, and the other road network data target maps in each type of road network data target map, excluding the first source domain, are determined as the first target domain.

[0026] Furthermore, the graph convolutional network is trained based on data from the first source domain and the second source domain, including:

[0027] The graph convolutional network is trained using data from the first source domain, and then trained using data from the second source domain based on an attention mechanism.

[0028] Furthermore, it also includes:

[0029] The pre-driving trajectory data of the mining truck is input into the target training model to obtain the predicted travel time of the mining truck.

[0030] In a second aspect, the present invention provides a data processing apparatus for predicting travel time, comprising:

[0031] The image acquisition module is used to acquire traffic feature maps and road network data maps corresponding to the driving area. The feature points in the traffic feature maps include feature information related to driving time.

[0032] The graph partitioning module is used to partition the traffic feature graph and the road network data graph respectively to obtain traffic feature sub-graphs and road network data sub-graphs;

[0033] The graph clustering module is used to cluster the traffic feature subgraph and the road network data subgraph respectively to obtain multiple traffic feature target maps and road network data target maps;

[0034] The region determination module is used to determine a first source region and a first target region from the traffic feature target map, and to determine a second source region and a second target region from the road network data target map;

[0035] The initial training module is used to train the graph convolutional network based on the data from the first source domain and the second source domain to obtain an initial training model;

[0036] The model generation module is used to train the initial training model based on the data from the first target domain and the second target domain to obtain the target training model.

[0037] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the data processing for travel time prediction provided in the first aspect.

[0038] Fourthly, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the data processing for travel time prediction provided in the first aspect.

[0039] This invention improves the accuracy of trip duration prediction by acquiring traffic feature maps and road network data maps corresponding to driving areas, taking into account both static road conditions and dynamic trajectory features, and selecting multiple data types. The traffic feature maps and road network data maps are divided and clustered to obtain multiple types of traffic feature target maps and road network data target maps. The initial data is preprocessed, and the data matrix is ​​stored in a graph format, preserving the correlation between data and improving the accuracy of trip duration prediction. Source and target domains are determined from the traffic feature target maps and road network data target maps, using a sub-map with a large amount of feature data as the source domain, further improving the accuracy of trip duration prediction. A graph convolutional network is trained based on the source domain data to obtain an initial training model, which fully utilizes data with more features, improving the accuracy of trip duration prediction. The initial training model is then trained based on the target domain data to obtain a target training model. Fine-tuning the model using data with fewer features significantly saves computational costs and further improves the accuracy of trip duration prediction. This invention achieves accurate prediction of mine car formation time by training a convolutional network on historical driving trajectories and road network data, thereby solving the technical problem that existing technologies cannot predict mine car travel time in open-pit mines where driving characteristics and road network data have not yet been obtained. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the data processing method for predicting travel time provided in an embodiment of the present invention.

[0042] Figure 2 This is a structural block diagram of a data processing device for predicting driving time provided in an embodiment of the present invention.

[0043] Figure 3 Block diagram of an electronic device provided in the embodiments of the present invention Detailed Implementation

[0044] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0045] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0046] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0047] As more and more open-pit mines rapidly develop towards larger equipment, more information-based management, and more intelligent production, existing technologies face the problem of being unable to predict mine car travel times in open-pit mine areas where driving characteristics and road network data have not yet been obtained. To address this issue, this invention provides a data processing method for predicting mine car travel times.

[0048] like Figure 1 The diagram shows a flowchart of a data processing method for predicting travel time, provided by an embodiment of the present invention. This data processing method for predicting travel time includes:

[0049] S101: Obtain the traffic feature map and road network data map corresponding to the driving area, wherein the feature points in the traffic feature map include feature information related to driving time;

[0050] Specifically, the traffic feature map can be obtained by extracting the historical trajectory data of the mining truck to obtain the historical trajectory. in, The time for the j-th trajectory point in the i-th trajectory. Let j be the longitude of the j-th trajectory point in the i-th trajectory. Let be the latitude of the j-th trajectory point in the i-th trajectory; for the historical trajectory Attribute information is labeled to obtain a traffic feature map, wherein the first attribute includes weather information and bus / train schedule information; a second attribute is calculated for each trajectory, the second attribute including average speed and traffic flow; and the historical trajectories are analyzed based on the first and second attributes. Attribute information is marked; the road network data map can be obtained by extracting the road network data to obtain the road network data map G = (V, E); where V is a node in the road network data map, E is an edge in the road network data map, and V = {v1, v2, ..., v...} l ,…,v m}, v m For the m-th road segment in the road network data map, e ij =(v i ,v j )∈E, e ij This indicates that the i-th road segment and the j-th road segment are connected in the road network data map.

[0051] S102: Divide the traffic feature map and the road network data map into traffic feature sub-maps and road network data sub-maps respectively;

[0052] Specifically, the graph can be partitioned based on nodes or edges. For example, when partitioning by nodes, some subgraphs may exist at the boundaries of multiple subgraphs. If a node is randomly assigned to any subgraph, it may disrupt the correlation between subgraphs, leading to poor prediction results. Therefore, the node can be assigned to different subgraphs simultaneously, thus ensuring that the structure of any subgraph is not destroyed. Although the integrity of the subgraph data is preserved, it increases computational resources. Selectively removing some nodes can reduce computational resources. A threshold for the distance between boundary nodes can be set to filter out some nodes on the boundaries of the subgraph, achieving the technical effect of reducing computational load.

[0053] S103: Cluster the traffic feature sub-map and the road network data sub-map respectively to obtain multiple traffic feature target maps and road network data target maps;

[0054] S104: Determine a first source domain and a first target domain from the traffic feature target map, and determine a second source domain and a second target domain from the road network data target map;

[0055] S105: Train the graph convolutional network based on the data from the first source domain and the second source domain to obtain an initial training model;

[0056] S106: Train the initial training model based on the data from the first target domain and the second target domain to obtain the target training model.

[0057] This invention improves the accuracy of trip duration prediction by acquiring traffic feature maps and road network data maps corresponding to driving areas, taking into account both static road conditions and dynamic trajectory features, and selecting multiple data types. The traffic feature maps and road network data maps are divided and clustered to obtain multiple types of traffic feature target maps and road network data target maps. The initial data is preprocessed, and the data matrix is ​​stored in a graph format, preserving the correlation between data and improving the accuracy of trip duration prediction. Source and target domains are determined from the traffic feature target maps and road network data target maps, using a sub-map with a large amount of feature data as the source domain, further improving the accuracy of trip duration prediction. A graph convolutional network is trained based on the source domain data to obtain an initial training model, which fully utilizes data with more features, improving the accuracy of trip duration prediction. The initial training model is then trained based on the target domain data to obtain a target training model. Fine-tuning the model using data with fewer features significantly saves computational costs and further improves the accuracy of trip duration prediction. This invention achieves accurate prediction of mine car formation time by training a convolutional network on historical driving trajectories and road network data, thereby solving the technical problem that existing technologies cannot predict mine car travel time in open-pit mines where driving characteristics and road network data have not yet been obtained.

[0058] Furthermore, obtain the traffic feature map and road network data map corresponding to the driving area, including:

[0059] The first feature information of the trajectory points is stored in the nodes of the traffic feature graph, and the second feature information between adjacent trajectory points is stored in the edges of the traffic feature graph; and

[0060] The sequence number of the path is stored in the node of the road network data graph, and the connectivity information between the paths is stored in the edge of the road network data graph;

[0061] The first feature information includes time, longitude, and latitude, and may also include information such as road length, width, number of turns, and turn coordinates. The second feature information includes weather information, bus schedule information, average vehicle speed, and traffic flow.

[0062] Furthermore, the traffic feature map and the road network data map are divided separately, including:

[0063] The traffic feature map is divided according to the second feature information between adjacent trajectory points; and

[0064] The road network data map is divided according to the connection information between the paths.

[0065] Specifically, the traffic feature map and the road network data map can be segmented using a multi-level k-way graph partitioning algorithm. Subgraph partitioning can be performed at the node level. Starting from a vertex in the original graph, a new graph can be grown along the edges from that vertex, dividing the original graph into two subgraphs with the same number of nodes. Alternatively, a threshold can be set during the subgraph growth process to obtain k subgraphs. For example, the traffic feature map and the road network data map can be partitioned k times along the edges according to their density, merging nodes in the same subgraph. The resulting subgraphs can then be concatenated to obtain the initial traffic feature map and road network data map. Subgraph partitioning can also be performed based on edges. For example, when partitioning by nodes, some subgraphs may exist at the boundaries of multiple subgraphs. If a node is randomly assigned to any subgraph, there is a risk of disrupting the correlation between subgraphs, resulting in poor prediction performance. Therefore, the node can be assigned to different subgraphs simultaneously, thus ensuring that the structure of any subgraph is not destroyed. Although the integrity of the subgraph data is preserved, it increases computational resources. In this case, selectively removing some nodes can reduce computational resources. A threshold for the distance between boundary nodes can be set to filter out some nodes on the boundaries of the subgraph, thereby achieving the technical effect of reducing the amount of computation.

[0066] Furthermore, clustering is performed on the traffic feature subgraph and the road network data subgraph, respectively, including:

[0067] Clustering is performed on the traffic feature sub-map based on the first feature information of the trajectory points and the second feature information between adjacent trajectory points; and

[0068] The road network data subgraph is clustered based on the sequence number of the path and the connectivity information between the paths.

[0069] Specifically, the traffic feature subgraph and the road network data subgraph can be clustered using the graph-level classification algorithm Graph2vec. For example, following the skipgram training process of doc2vec, the features of each subgraph are stored in a vector to obtain the vector representation of each subgraph, and then they are classified based on a vector-based clustering method.

[0070] Further, determining the first source domain and the first target domain from the traffic feature target map includes:

[0071] The traffic feature target map with the largest amount of feature data in each type of traffic feature target map is determined as the first source domain, and the other traffic feature target maps in each type of traffic feature target map, excluding the first source domain, are determined as the first target domain; and

[0072] Determining the second source domain and the second target domain from the road network data target map includes:

[0073] The road network data target map containing the most connected paths in each type of road network data target map is determined as the first source domain, and the other road network data target maps in each type of road network data target map, excluding the first source domain, are determined as the first target domain.

[0074] Specifically, the consistency between the source and target domains can also be verified as follows: obtain the normalized Laplace matrix L of the nodes in the source domain, and verify the consistency of the eigenfunctions f of this normalized Laplace matrix L. G (n) is diagonalized. Where, λ l The eigenvalues ​​of the normalized Laplace matrix L are... For the eigenfunction f G The transformation function of (n), where n is the index of the node, u l (n) represents the orthogonal eigenvectors of the subgraph, n∈[1,N-1], where N represents the total number of nodes in the graph. We can obtain... Similarly, the eigenfunctions g of the target domain corresponding to the source domain can be obtained. G (n) and its transformation function g G (n), if This indicates that the source domain and the corresponding target domain are consistent.

[0075] Furthermore, the graph convolutional network is trained based on data from the first source domain and the second source domain, including:

[0076] The graph convolutional network is trained using data from the first source domain, and then trained using data from the second source domain based on an attention mechanism.

[0077] Specifically, the data from the first and second source domains are first trained using graph convolution, and then Fourier transform is performed to obtain the frequency domain formula. Where, x k,i For input, x k+1,j For the output, j∈[1,φ k ],φ k F represents the number of neighboring nodes. k,i,j It is a diagonal matrix, and the function h represents the three layers of graph convolution: convolutional layer, pooling layer, and fully connected layer. For the convolution process of the second source domain, an attention mechanism can also be introduced to dynamically capture feature data that has a greater impact on the learning results. f represents the convolution kernel, g represents the subgraph, f*Gg represents the graph convolution process after adding attention, and A′ represents the adjoint adjacency matrix. The influence weights between nodes are dynamically adjusted by adjusting the adjoint adjacency matrix A′ and the attention matrix S. S i,j The expression represents the correlation strength between node i and node j, softmax represents the normalization process, S′ represents the final attention matrix, and the model parameter matrix of the first source domain is obtained after training. Model parameter matrix of the second source domain The trained model parameters are used as the initial parameters for training the model based on the target domain data. The fully connected layer then predicts the travel time for a given route. Among them, MLP (Concat) is a fully connected layer. This represents the predicted travel time for road segment l, and the total predicted travel time t is calculated by... l Adding them together will give the final predicted time. Where m is the number of road segments contained in a given path Q, and finally the loss function of the target model is calculated. t j Y represents the actual travel time for road segment j. i This represents the actual travel time of road segment i, ultimately resulting in the trained target model.

[0078] Furthermore, it also includes:

[0079] The pre-driving trajectory data of the mining truck is input into the target training model to obtain the predicted travel time of the mining truck.

[0080] Based on the above embodiments and their contents, as an optional embodiment, such as... Figure 2 As shown in the diagram, this invention provides a structural block diagram of a data processing device for predicting driving time. The data processing device for predicting driving time includes:

[0081] The image acquisition module 201 is used to acquire the traffic feature map and road network data map corresponding to the driving area, wherein the feature points in the traffic feature map include feature information related to driving time;

[0082] The graph partitioning module 202 is used to partition the traffic feature map and the road network data map respectively to obtain a traffic feature sub-map and a road network data sub-map;

[0083] The graph clustering module 203 is used to cluster the traffic feature subgraph and the road network data subgraph respectively to obtain multiple traffic feature target graphs and road network data target graphs;

[0084] The region determination module 204 is used to determine a first source domain and a first target domain from the traffic feature target map, and to determine a second source domain and a second target domain from the road network data target map.

[0085] The initial training module 205 is used to train the graph convolutional network based on the data from the first source domain and the second source domain to obtain an initial training model;

[0086] The model generation module 206 is used to train the initial training model based on the data of the first target domain and the second target domain to obtain the target training model.

[0087] This invention improves the accuracy of trip duration prediction by acquiring traffic feature maps and road network data maps corresponding to driving areas, taking into account both static road conditions and dynamic trajectory features, and selecting multiple data types. The traffic feature maps and road network data maps are divided and clustered to obtain multiple types of traffic feature target maps and road network data target maps. The initial data is preprocessed, and the data matrix is ​​stored in a graph format, preserving the correlation between data and improving the accuracy of trip duration prediction. Source and target domains are determined from the traffic feature target maps and road network data target maps, using a sub-map with a large amount of feature data as the source domain, further improving the accuracy of trip duration prediction. A graph convolutional network is trained based on the source domain data to obtain an initial training model, which fully utilizes data with more features, improving the accuracy of trip duration prediction. The initial training model is then trained based on the target domain data to obtain a target training model. Fine-tuning the model using data with fewer features significantly saves computational costs and further improves the accuracy of trip duration prediction. This invention achieves accurate prediction of mine car formation time by training a convolutional network on historical driving trajectories and road network data, thereby solving the technical problem that existing technologies cannot predict mine car travel time in open-pit mines where driving characteristics and road network data have not yet been obtained.

[0088] The graph acquisition module 201 includes a graph data unit.

[0089] The graph data unit is used to store the first feature information of the trajectory points to the nodes of the traffic feature graph, store the second feature information between adjacent trajectory points to the edges of the traffic feature graph; and store the sequence number of the path to the nodes of the road network data graph, and store the connectivity information between the paths to the edges of the road network data graph; wherein, the first feature information includes time, longitude and latitude, and the second feature information includes weather information, bus schedule information, average vehicle speed and traffic flow.

[0090] The graph partitioning module 202 includes graph partitioning units.

[0091] The graph partitioning unit is used to partition the traffic feature graph based on the second feature information between adjacent trajectory points; and to partition the road network data graph based on the connection information between the paths.

[0092] The graph clustering module 203 includes graph clustering units.

[0093] The graph clustering unit is used to cluster the traffic feature subgraph based on the first feature information of the trajectory points and the second feature information between adjacent trajectory points; and to cluster the road network data subgraph based on the sequence number of the path and the connectivity information between the paths.

[0094] The region determination module 204 includes a source domain determination unit and a target domain determination unit.

[0095] The source domain determination unit is used to determine the traffic feature target map with the largest amount of feature data in each type of traffic feature target map as the first source domain, and to determine the other traffic feature target maps in each type of traffic feature target map besides the first source domain as the first target domain; and

[0096] The target domain unit is used to determine the road network data target map containing the most connected paths in each type of road network data target map as the first source domain, and to determine the other road network data target maps in each type of road network data target map besides the first source domain as the first target domain.

[0097] The initial training module 205 includes an initial training unit.

[0098] The initial training unit trains the graph convolutional network based on the data from the first source domain, and trains the graph convolutional network based on the data from the second source domain using an attention mechanism.

[0099] The model generation module 206 includes a model generation unit.

[0100] The model generation unit inputs the pre-driving trajectory data of the mining truck into the target training model to obtain the predicted travel time of the mining truck.

[0101] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0102] Figure 3A block diagram of an electronic device provided in an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 301, a memory 302, and a bus 303;

[0103] The processor 301 and the memory 302 communicate with each other via the bus 303. The processor 301 is used to call program instructions in the memory 302 to execute the data processing method for travel time prediction provided in the above embodiment, including:

[0104] Obtain traffic feature map and road network data map corresponding to the driving area, wherein the feature points in the traffic feature map include feature information related to driving time;

[0105] The traffic feature map and the road network data map are divided into traffic feature sub-maps and road network data sub-maps, respectively.

[0106] Clustering is performed on the traffic feature sub-map and the road network data sub-map respectively to obtain multi-class traffic feature target maps and road network data target maps;

[0107] A first source domain and a first target domain are determined from the traffic feature target map, and a second source domain and a second target domain are determined from the road network data target map;

[0108] The graph convolutional network is trained based on the data from the first source domain and the second source domain to obtain an initial training model;

[0109] The initial training model is trained based on the data from the first target domain and the second target domain to obtain the target training model.

[0110] This invention provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of a data processing method for travel time prediction. For example, it includes:

[0111] Obtain traffic feature map and road network data map corresponding to the driving area, wherein the feature points in the traffic feature map include feature information related to driving time;

[0112] The traffic feature map and the road network data map are divided into traffic feature sub-maps and road network data sub-maps, respectively.

[0113] Clustering is performed on the traffic feature sub-map and the road network data sub-map respectively to obtain multi-class traffic feature target maps and road network data target maps;

[0114] A first source domain and a first target domain are determined from the traffic feature target map, and a second source domain and a second target domain are determined from the road network data target map;

[0115] The graph convolutional network is trained based on the data from the first source domain and the second source domain to obtain an initial training model;

[0116] The initial training model is trained based on the data from the first target domain and the second target domain to obtain the target training model.

[0117] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0118] Finally, specific embodiments have been used to illustrate the principles and implementation methods of this invention. The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of this invention. 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 invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A data processing method for predicting travel time, characterized in that, include: Obtaining a traffic feature map and a road network data map corresponding to a driving area, wherein the feature points in the traffic feature map include feature information related to driving time, and obtaining the traffic feature map and road network data map corresponding to the driving area includes: storing the first feature information of the trajectory points to the nodes of the traffic feature map, storing the second feature information between adjacent trajectory points to the edges of the traffic feature map; and storing the sequence number of the path to the nodes of the road network data map, storing the connectivity information between the paths to the edges of the road network data map; wherein the first feature information includes time, longitude, and latitude, and the second feature information includes weather information, bus schedule information, average vehicle speed, and traffic flow. The traffic feature map and the road network data map are divided into traffic feature sub-maps and road network data sub-maps, respectively. Clustering is performed on the traffic feature sub-map and the road network data sub-map respectively to obtain multi-class traffic feature target maps and road network data target maps; The first source domain and the first target domain are determined from the traffic feature target map, and the second source domain and the second target domain are determined from the road network data target map. Specifically, the traffic feature target map containing the largest amount of feature data in each type of traffic feature target map is determined as the first source domain, and the other traffic feature target maps in each type of traffic feature target map, excluding the first source domain, are determined as the first target domain. The determination of the second source domain and the second target domain from the road network data target map includes: determining the road network data target map containing the most connected paths in each type of road network data target map as the second source domain, and the other road network data target maps in each type of road network data target map, excluding the second source domain, as the second target domain. The graph convolutional network is trained based on the data from the first source domain and the second source domain to obtain an initial training model; the initial training model is trained based on the data from the first target domain and the second target domain to obtain a target training model, wherein the graph convolutional network is trained based on the data from the first source domain, and the graph convolutional network is trained based on the data from the second source domain using an attention mechanism; The pre-trajectory data of the mining truck is input into the target training model to obtain the predicted travel time of the mining truck.

2. The method according to claim 1, characterized in that, The traffic feature map and the road network data map are divided separately, including: dividing the traffic feature map according to the second feature information between adjacent trajectory points; and dividing the road network data map according to the connection information between the paths.

3. The method according to claim 1, characterized in that, Clustering is performed on the traffic feature subgraph and the road network data subgraph, respectively, including: clustering the traffic feature subgraph based on the first feature information of the trajectory points and the second feature information between adjacent trajectory points; and clustering the road network data subgraph based on the sequence number of the path and the connectivity information between the paths.

4. A data processing device for predicting travel time, characterized in that, include: The image acquisition module is used to acquire a traffic feature map and a road network data map corresponding to a driving area. The feature points in the traffic feature map include feature information related to driving time. Acquiring the traffic feature map and road network data map corresponding to the driving area includes: storing the first feature information of the trajectory points in the nodes of the traffic feature map, storing the second feature information between adjacent trajectory points in the edges of the traffic feature map; and storing the sequence number of the path in the nodes of the road network data map, storing the connectivity information between the paths in the edges of the road network data map. The first feature information includes time, longitude, and latitude, and the second feature information includes weather information, bus schedule information, average vehicle speed, and traffic flow. The graph partitioning module is used to partition the traffic feature graph and the road network data graph respectively to obtain traffic feature sub-graphs and road network data sub-graphs; The graph clustering module is used to cluster the traffic feature subgraph and the road network data subgraph respectively to obtain multiple traffic feature target maps and road network data target maps; A region determination module is used to determine a first source domain and a first target domain from the traffic feature target map, and to determine a second source domain and a second target domain from the road network data target map. Specifically, the traffic feature target map containing the largest amount of feature data in each type of traffic feature target map is determined as the first source domain, and the other traffic feature target maps in each type of traffic feature target map, excluding the first source domain, are determined as the first target domain. The module also determines the second source domain and the second target domain from the road network data target map, including: determining the road network data target map containing the most connected paths in each type of road network data target map as the second source domain, and the other road network data target maps in each type of road network data target map, excluding the second source domain, as the second target domain. The initial training module is used to train the graph convolutional network based on the data from the first source domain and the second source domain to obtain an initial training model; The model generation module is used to train the initial training model based on the data of the first target domain and the second target domain to obtain the target training model, wherein the graph convolutional network is trained based on the data of the first source domain, and the graph convolutional network is trained based on the data of the second source domain using an attention mechanism. In this process, the pre-driving trajectory data of the mining truck is input into the target training model to obtain the predicted travel time of the mining truck.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the data processing method for travel time prediction as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the data processing method for travel time prediction as described in any one of claims 1 to 3.

Images