Information processing method and device, electronic equipment and storage medium

Abstract

The invention provides an information processing method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence, in particular to the technical fields of knowledge maps, data processing and the like. According to the specific implementation scheme, the method comprises the steps that POI graphs corresponding to at least two continuous time units are obtained; obtaining a first relation mode and a first message mode between nodes in the POI graph and second-order neighbor nodes of the nodes, and carrying out constraint limitation on the first relation mode based on the first message mode to obtain a second relation mode; based on the second relation mode and the first message mode, carrying out information mining in a time unit on the POI graph to obtain a first POI representation vector of the POI; obtaining an associated POI graph of the POI graph; and performing information mining between time units on the POI graph and the associated POI graph, and obtaining a second POI representation vector of the POI in combination with the first POI representation vector.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, in particular to technical fields such as knowledge graphs and data processing, and specifically to an information processing method, device, electronic device and storage medium. Background Art

[0002] Points of interest on a map are a basic requirement for users when traveling. Since users' urban activities are different at different times, learning time-aware POI representations is of great value. Currently, most learning of POI representations in various application scenarios focuses on learning static representations, ignoring the variability of POI semantics and spatial distribution in the time dimension, resulting in insufficient information learning of POI representations. Summary of the invention

[0003] The present disclosure provides an information processing method, apparatus, electronic device and storage device.

[0004] According to one aspect of the present disclosure, there is provided an information processing method, comprising:

[0005] Obtaining POI graphs corresponding to at least two consecutive time units;

[0006] Acquire a first relationship pattern and a first message pattern of a node and its second-order neighboring nodes in the POI graph, and constrain the first relationship pattern based on the first message pattern to obtain a second relationship pattern;

[0007] Perform information mining in a time unit on the POI graph based on the second relationship model and the first message model to obtain a first POI representation vector of the POI;

[0008] Obtaining a POI graph associated with the POI graph;

[0009] Information mining between time units is performed on the POI graph and the associated POI graph, and a second POI representation vector of the POI is obtained by combining the first POI representation vector.

[0010] According to another aspect of the present disclosure, there is provided an information processing device, including:

[0011] A first acquisition module is used to acquire a POI map corresponding to at least two consecutive time units;

[0012] A second acquisition module is used to acquire a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in the POI graph, and to constrain the first relationship pattern based on the first message pattern to obtain a second relationship pattern;

[0013] A third acquisition module is used to perform information mining in a time unit on the POI graph based on the second relationship mode and the first message mode to obtain a first POI representation vector of the POI;

[0014] A fourth acquisition module is used to acquire a POI map associated with the POI map;

[0015] The information processing module is used to perform information mining between time units on the POI graph and the associated POI graph, and obtain a second POI representation vector of the POI in combination with the first POI representation vector.

[0016] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0017] at least one processor; and

[0018] a memory communicatively coupled to the at least one processor;

[0019] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the method described in the embodiment of the first aspect.

[0020] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method described in the embodiment of the first aspect.

[0021] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the steps of the method described in the embodiment of the first aspect are implemented.

[0022] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0024] Figure 1 is a schematic diagram of an information processing method provided by an embodiment of the present disclosure;

[0025] Figure 2 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure;

[0026] Figure 2A is a logic diagram for obtaining a first POI representation vector provided by an embodiment of the present disclosure;

[0027] Figure 3 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure;

[0028] Figure 3A is a logic diagram for obtaining a second POI representation vector provided by an embodiment of the present disclosure;

[0029] Figure 4 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure;

[0030] Figure 4A is a logic diagram of an information processing method provided by an embodiment of the present disclosure;

[0031] Figure 5 is a structural block diagram of an information processing device provided by an embodiment of the present disclosure;

[0032] Figure 6 is a schematic block diagram of an exemplary electronic device for implementing an embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0034] Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. It is an important part of the intelligent discipline. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. Artificial intelligence can simulate the information process of human consciousness and thinking.

[0035] Knowledge Graph is a graph-based data structure consisting of nodes and edges. Each node represents an "entity" and each edge represents the "relationship" between entities. It is essentially a semantic network that uses visualization technology to describe knowledge resources and their carriers, and mine, analyze, construct, draw and display knowledge and their mutual connections. It displays complex knowledge fields through data mining, information processing, knowledge measurement and graph drawing.

[0036] Data processing is the collection, storage, retrieval, processing, transformation and transmission of data. It runs through all areas of social production and social life. Its basic purpose is to extract and derive valuable and meaningful data for certain specific people from large amounts of disorganized and difficult to understand data.

[0037] Figure 1 is a schematic diagram of an information processing method provided by an embodiment of the present disclosure; Figure 1 As shown, the method comprises the following steps:

[0038] S101, obtaining point of interest (POI) images corresponding to at least two consecutive time units.

[0039] It is understandable that in different downstream scenarios, the features represented by the POI nodes in the Point of Interest (POI) graph are different. For example, POI nodes are the basic needs of urban residents for daily travel. Since different users have different urban activities at different times, for example, residents will eat at characteristic time periods at noon and evening, and go to bars in the early morning, etc., different POI nodes can be recommended for user travel in different time periods, and information such as the traffic of each POI node within different time ranges can be determined, so as to accurately represent and learn POIs and apply them to scenarios such as dynamic personalized POI recommendations or POI traffic prediction analysis.

[0040] Each time unit corresponds to a POI graph, which includes multiple POI nodes, and the POI nodes are connected by edges; the POI graphs of multiple consecutive time units are obtained to perform POI node feature mining. In this embodiment, the time unit can be counted in hours or days, without specific limitation.

[0041] S102, obtaining a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in a POI graph, and constraining the first relationship pattern based on the first message pattern to obtain a second relationship pattern.

[0042] Considering the complex environment around POI nodes, if only focusing on the local POI node relationship, the context information learning of POI nodes will be incomplete. In this embodiment, the first relationship pattern of the node and its second-order neighbor node in the POI graph is obtained. The first relationship pattern includes the relationship path from the node to the corresponding second-order neighbor node through its first-order neighbor node. For example, the node v i is through its first-order neighbor node v j To its second-order neighbor node v k , then the path is from node to its first-order neighbor node v i →v j(It can also be expressed as R i,j or r1) and first-order neighbor nodes to second-order neighbor nodes v j →v k (It can also be expressed as R j,k or r2) two relations, so the node v i The first relationship pattern between it and its second-order neighbor nodes can be expressed as

[0043] Furthermore, considering that the POI graph includes first relationship patterns of multiple nodes and their second-order neighbor nodes, each first relationship pattern covers several first-order neighbor nodes and second-order neighbor nodes, so the first relationship pattern can be expressed as follows:

[0044]

[0045] In some implementations, the first message mode between a node in the POI graph and its second-order neighbor nodes can also be obtained. The message mode refers to the mode of each directed message transmission, for example, k →v i The directed message passing mode is based on the first relational mode extracted from v i to v k All transmission paths of node v i Reach the second-order neighbor node v through different first-order neighbor nodes k The transmission path of , and collect the intermediate first-order nodes to get the first message mode, the first message mode can be expressed as:

[0046]

[0047] Furthermore, although constructing second-order neighbor nodes to integrate the first relationship pattern can improve the richness of information mining, the complexity of the node relationship pattern will reduce the computing efficiency during actual construction. Therefore, this embodiment can constrain the first relationship pattern based on the first message pattern to reduce the number of paths in the first relationship pattern and obtain a filtered second relationship pattern.

[0048] In some implementations, the paths with richer information can be screened out based on the message transmission paths between nodes in the first message mode, such as the path between nodes v i and node v k There are multiple paths through which messages can be transmitted. If node v i and node v k There are multiple first-order neighbor nodes between nodes, node v i and node v k The message path information between nodes is relatively rich, so node v can be retained. i and node v kThe first relationship pattern between them is taken as the second relationship pattern; and by analogy, the second-order neighbor nodes that are reached by multiple paths are retained, thereby filtering out the first relationship pattern between the node and the second-order neighbor node, and the filtered first relationship pattern is taken as the second relationship pattern.

[0049] S103: Perform information mining on the POI graph within a time unit based on the second relationship model and the first message model to obtain a first POI representation vector of the POI.

[0050] Optionally, the context vector of the second-order neighbor node and the context vector of the first-order neighbor node can be determined according to the second relationship pattern and the first message pattern respectively, and the context representation vector between the second-order neighbor node and the current POI node is obtained by fusing the context vector of the second-order neighbor node and the context vector of the first-order neighbor node. All the context representation vectors corresponding to the POI node are further aggregated to obtain the first POI representation vector of the POI node. The first POI representation vector fuses the context vector information of the first-order neighbor nodes and the second-order neighbor nodes within the time unit, and the information mining is more comprehensive.

[0051] S104, obtaining the associated POI graph of the POI graph.

[0052] In some implementations, the adjacent time period of the POI image may be determined, and the POI image in the adjacent time period may be used as the associated POI image. For example, the adjacent time period is arrive but arrive The POI graphs between are related POI graphs.

[0053] S105: Perform information mining between time units on the POI graph and the associated POI graph, and obtain a second POI representation vector of the POI in combination with the first POI representation vector.

[0054] Optionally, the context feature information of the nodes in the POI map and the associated POI map can be mined to determine the target context vector corresponding to each node, for example, the first POI representation vectors corresponding to the points in the POI map and the associated POI map can be determined, and by fusing the first POI representation vectors, a cross-time and space vector that integrates the time and space information is obtained. The cross-time and space vector includes the information of the node in different time units. The cross-time and space vector is combined with the node's own spatial vector and time vector to obtain a target context vector that fully mines the node information.

[0055] Furthermore, the target context vector is fused with the node's cross-temporal and spatial vector to obtain a second POI representation vector with more comprehensive information mining. The second POI representation vector fuses the context vector and the cross-temporal and spatial vector, integrates complex POI contextual environment information and multi-time period structural features, and outputs a more accurate POI representation.

[0056] In this embodiment, POI graphs of multiple continuous time units are obtained to perform POI feature mining and analysis, two modes, namely a first relationship mode and a first message mode, are determined based on the nodes in the POI graph and their second-order neighbor nodes, and constraints are imposed on the relationship paths in the first relationship mode based on the first message mode to avoid the influence of low-quality relationship paths on POI feature extraction, information mining is performed within the time unit based on the screened second relationship mode and the first message mode to obtain a more complete first POI representation vector for each POI node in the POI graph, and an associated POI graph is further determined based on the time unit, and information mining across time units is performed based on the POI graph and the associated POI graph, and the sufficient first POI representation vector within the time unit is combined to obtain a second POI representation vector with comprehensive time features, so as to fully mine the pattern information of time and space, improve the accuracy of POI representation vector mining, and solve the problem that the traditional model is insufficient in learning multi-time POI representations.

[0057] Figure 2 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure; Figure 2 As shown, the method comprises the following steps:

[0058] S201, obtaining POI images corresponding to at least two consecutive time units.

[0059] In the embodiments of the present disclosure, the implementation method of step 201 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0060] S202: Obtain a first relationship pattern and a first message pattern between a node and its second-order neighbor nodes in a POI graph.

[0061] In some implementations, the node v in the POI graph can be obtained i Arrival node v i The second-order neighbor node v k 'path; determine node v according to the path i To the first-order neighbor node v j The first path relationship of node v j To node v k 'The second path relationship; Based on the first path relationship and the second path relationship, generate a first relationship model; that is, determine the node v iTo the first-order neighbor node v j The first path relationship v i →v j , and node v j To node v k 'The second path relationship v j →v k ', by the first path relation v i →v j and the second path relation v j →v k 'Form the first relational model.

[0062] In some implementations, the node v may be extracted from the first relationship pattern. i Arrival node v k 'Candidate paths, and determine the nodes v that each candidate path passes through j ; generated at node v k 'To node v i Passing the message through node v j The first message mode; it can be understood that node v i Arrival node v k 'Can pass through multiple different nodes v j , the first message pattern is sent by node v k 'Pathway node v j To node v i The message path for delivering messages is generated, and the relationship-level and message-level patterns are defined for each node to facilitate more fine-grained mining of POI representation information.

[0063] In the embodiments of the present disclosure, the implementation method of step S202 can be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0064] S203, determine node v k 'To node v i The number of paths along which the message is delivered.

[0065] It is understandable that node v k 'To node v i The message can be transmitted through multiple different first-order neighbor nodes. k 'To node v i There may be multiple paths for transmitting messages, and the statistical node v k 'To node v i The number of paths along which the message is delivered.

[0066] S204: According to the number of paths for transmitting the message, node v in the first message mode is k'Filter to obtain the second message mode, the second-order neighbor node v in the second message mode k To node v i The number of paths for transmitting messages is greater than or equal to the set number.

[0067] Optionally, the number of paths for transmitting messages may be determined from the node v in the first message mode. k 'Determine the second-order neighbor node v to be deleted k "; The second-order neighbor node v to be deleted k "To node v i The number of paths for transmitting messages is less than the set number. In this embodiment, the set number is 2. Then, according to the number of paths for transmitting messages, the node v in the first message mode is k 'Determine the direction to node v i The number of paths for transmitting messages is less than the set number of second-order neighbor nodes v k ”.

[0068] Based on node v k "Update the node in the first message mode to obtain the second message mode, that is, delete the node v in the first message mode k ”, based on the remaining second-order neighbor nodes v k To node v i The message path of the transmitted message determines the second message mode, which reduces the analysis of the message path of a single message transmission and improves the analysis and processing efficiency. The second message mode can be expressed as

[0069] S205, according to node v k and the second message model, constraining the first relationship model to obtain the second relationship model.

[0070] In some implementations, you can get the node v k The weight coefficient of node v k , node v k The weight coefficient and the second message pattern generate a second relationship pattern.

[0071] Optionally, the node v may be determined according to the second message mode. k To node v i The number of paths for transmitting messages; According to the first relationship pattern, determine the node v i The total number of paths to all nodes; Determine the node v based on the number of paths for transmitting messages and the total number of paths k The weight coefficient of the calculation node v k To node v i The number of paths for delivering messages accounts for the number of nodes v iThe ratio of the total number of all surrounding paths is used as the node v k The weight coefficient of the node is larger, and the larger the weight coefficient is, the more important the node is. k To node v i The more important the message path is, the more important it is to deliver the message.

[0072] Optionally, node v k The weight coefficient can be expressed as:

[0073]

[0074] According to the node v k The weight coefficient and the second message mode generate a second relationship mode, which can be expressed as:

[0075]

[0076] in, Indicates the second relational mode; Represents node v k A weight coefficient is generated based on the weight coefficient and the second message pattern, and a filtered second relationship pattern is generated. Subsequent analysis is performed based on the second relationship pattern to improve processing efficiency.

[0077] S206: Perform information mining on the POI graph within a time unit based on the second relationship model and the first message model to obtain a first POI representation vector of the POI.

[0078] In some implementations, the node v may be determined based on the first message mode. k To node v i The node v through which the message is transmitted j , and get the node v j According to the second relationship pattern, determine the node v k , and get the node v k context vector of the node v; optionally, j The context vector of the node v can be obtained by preliminarily extracting features based on the connection information between nodes in the POI graph. j After the context vector of node v is obtained, the maximum pooling operation can be used to capture the node v k The context vector information of .

[0079] Optionally, node v k The context vector can be obtained as:

[0080]

[0081] in, Indicates that at time unit t, node v k Context vector of; Max-Pool represents the maximum pooling operation; Represents node v j The context vector of is a learnable parameter matrix used to learn node v j The context vector is transformed linearly in space.

[0082] Furthermore, based on the node v j The context vector and node v k The context vector of node v k With node v i message representation vector between nodes; for node v k With node v i Aggregate the message representation vectors between nodes to get node v i The first POI representation vector of the corresponding POI; Optionally, the node v can be calculated j The context vector and node v k The sum of the context vectors of different nodes v is used as the message representation vector; j With node v k The context vector of the node v is used to obtain multiple message representation vectors. k With node v i The message representation vectors between the two POIs are aggregated to obtain the first POI representation vector.

[0083] Optionally, you can base the node v k The weight coefficient for node v k With node v i The message representation vectors corresponding to each message path between nodes are weighted summed to obtain node v i At node v k The relationship representation vector under the path relationship.

[0084] Alternatively, the relation representation vector can be expressed as:

[0085]

[0086] in, is the message representation vector; Represents node v i At node v k The relationship representation vector under the path relationship.

[0087] Furthermore, for node v i The relationship representation vectors under each path relationship are summed to obtain the node v i The corresponding first POI representation vector can be expressed as:

[0088]

[0089] in, Represents the set of all edge relations; (r1, r2) is the combination relation pattern of any relation r1 and any relation r2; It is the first POI representation vector, which aggregates the context vector representation and weight coefficient of each node in the time unit. The obtained POI representation is more reliable and the information mining is more sufficient and comprehensive.

[0090] For example, Figure 2A As shown, for the second-order neighbor node v k1 and the second-order neighbor node v k2 , according to the second-order neighbor node and node v i The first-order neighbor nodes that transmit messages between each other obtain the context vector through the context pooling operation and The context vectors of the second-order neighbor nodes and the first-order neighbor nodes are further fused and weighted to obtain their respective relationship representation vectors. For node v i Aggregate the relationship representation vectors under each path relationship, that is, aggregate v k1 and v k2 , and obtain the final first POI representation vector.

[0091] S207, obtaining the associated POI graph of the POI graph.

[0092] In the embodiments of the present disclosure, the implementation method of step S207 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0093] S208: Perform information mining between time units on the POI graph and the associated POI graph, and obtain a second POI representation vector of the POI in combination with the first POI representation vector.

[0094] In the embodiments of the present disclosure, the implementation method of step S208 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0095] In this embodiment, a first relationship pattern between a node and its second-order neighbor nodes is generated through the path relationship between the node and the first-order neighbor nodes, and the path relationship between the first-order neighbor nodes and the second-order neighbor nodes. The first message pattern is determined according to the message transmission path between the node and its second-order neighbor nodes. Screening is performed according to the number of paths for transmitting messages between the second-order neighbor nodes, and low-frequency message transmission paths are deleted to obtain a second message pattern. The weight coefficient of each second-order neighbor node in the second message pattern is determined based on the number of paths. Based on the weight coefficient and the second message pattern, an updated second relationship pattern is determined to avoid too many low-quality paths affecting the analysis efficiency during actual construction. Further, the context vector of the first-order neighbor node and the context vector of the second-order neighbor node are obtained. A context vector is obtained, and a message representation vector is obtained by adding up the respective context vectors. The message representation vector includes more sufficient node context information. A relationship representation vector is obtained by weighted summation based on the message representation vectors and weight coefficients between nodes. Then, the relationship representation vectors under all path relationships are summed to obtain the first POI representation vector corresponding to the node. The first POI representation vector aggregates the context information of more distant second-order neighbor nodes, and the information mining in the POI graph within the time unit is richer. The first POI representation vector of the node obtained is more accurate. Subsequent analysis and processing are performed based on the more accurate first POI representation vector to conduct information mining across time units, thereby improving the accuracy of obtaining the second POI representation vector.

[0096] Figure 3 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure; Figure 3 As shown, the method comprises the following steps:

[0097] S301, obtaining point of interest (POI) images corresponding to at least two consecutive time units.

[0098] In the embodiments of the present disclosure, the implementation method of step S301 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0099] S302, obtaining a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in a POI graph, and constraining the first relationship pattern based on the first message pattern to obtain a second relationship pattern.

[0100] In the embodiments of the present disclosure, the implementation method of step S302 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0101] S303: Perform information mining on the POI graph within a time unit based on the second relationship model and the first message model to obtain a first POI representation vector of the POI.

[0102] In the embodiments of the present disclosure, the implementation method of step S303 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0103] S304, obtaining the associated POI graph of the POI graph.

[0104] In some implementations, the POI map time unit can be obtained to determine the adjacent time unit of the POI map; based on the adjacent time unit, the associated POI map of the POI map is determined; that is, the time unit of the current POI map is determined to obtain the adjacent time unit, and the POI map in the adjacent time unit is used as the associated POI map to achieve cross-time unit fusion processing and improve the POI representation capability.

[0105] In the embodiments of the present disclosure, the implementation method of step S304 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0106] S305, determining the node v according to the POI graph and the associated POI graph i Arriving at the second-order neighbor node v k The nearest N first-order neighbor nodes passed through, and determine the node v i The spatial distance and time interval between N first-order neighbor nodes.

[0107] In some implementations, the node v can be determined based on the POI graph. i Arrival node v k The first path set, that is, to determine the node v in the POI graph i Arrival node v k All paths of form the first path set; according to the associated POI graph, determine the node v i Arrival node v k The second path set, that is, to determine the node v in the associated POI graph i Arrival node v k All paths of constitute the second path set.

[0108] According to the first path set and the second path set, determine the node v i Arrival node v k The candidate first-order neighbor nodes passed by; it can be understood that the candidate first-order neighbor nodes include nodes in the POI graph and the associated POI graph, that is, the candidate first-order neighbor nodes do not belong to the same time unit; further determine the candidate first-order neighbor node and node v iThe spatial distance is calculated, and N first-order neighbor nodes are determined from the candidate first-order neighbor nodes according to the spatial distance; optionally, the spatial distance can be calculated according to the longitude and latitude coordinates between the nodes, and the N nearest candidate first-order neighbor nodes are determined according to the spatial distance as the N first-order neighbor nodes, and the context information of the N first-order neighbor nodes is used to provide rich spatiotemporal information to improve processing efficiency and accuracy.

[0109] Furthermore, the calculation node v i The time interval between N first-order neighbor nodes, for any first-order neighbor node, in this embodiment, the time interval is the difference between the time unit of the associated POI map where the first-order neighbor node is located and the time unit of the current POI map. For example, the associated POI map corresponding to the first-order neighbor node is time τ, and the current POI map is time t, then the time interval Δτ=t-τ.

[0110] S306, according to the node v i and N first-order neighbor nodes, as well as the spatial distance and time interval, determine the node v k To node v i The third messaging mode for delivering messages.

[0111] Alternatively, the third message mode can be expressed as:

[0112]

[0113] in, is the N first-order neighbor nodes determined by the K-Nearest Neighbor (KNN) algorithm based on spatial distance; Δd j,k is the spatial distance; Δτ is the temporal distance.

[0114] S307: Obtain a second POI representation vector of the POI according to the third message mode and the first POI representation vectors of the N first-order neighbor nodes.

[0115] In some implementations, the node v among the N first-order neighbor nodes may be determined according to the third message mode. j , and for node v in the POI graph j The first POI representation vector and the node v in the associated POI graph j The first POI representation vector of node v is obtained by performing cross-time fusion of the representation vectors. j The cross-time fusion vector; that is, for any node v among the N first-order neighbor nodes j , determine the first POI representation vector corresponding to each in the POI map and the associated POI map, so as to fuse the first POI representation vector in the POI map and the first POI representation vector in the associated POI map across time to obtain the node v jThe cross-time fusion vector.

[0116] Alternatively, the cross-temporal fusion vector can be expressed as:

[0117]

[0118] in, Represents node v j The cross-time fusion vector of is the adjacent time unit range of the POI graph; W t represents the time fusion transformation parameters; Represents node v j The corresponding first POI representation vector includes the first POI representation vector in the POI map and the first POI representation vector in the associated POI map.

[0119] Furthermore, according to the third relationship pattern, the node v j The time vector and space vector of the node v j and the second-order neighbor node v k The time interval between the nodes is obtained, and the space vector is obtained by j and the second-order neighbor node v k The spatial distance between them is obtained; for node v j The time vector and space vector of node v are fused to obtain j The space-time vector integrates time information and space information, captures multi-dimensional information for representation, and improves the POI representation capability; Optionally, the node v j The time vector and space vector of the node v are concatenated to obtain j The space-time vector of .

[0120] Furthermore, according to the spatiotemporal vector and the cross-time fusion vectors of the N first-order neighbor nodes, the node v is determined k To node v i The target context vector when delivering a message; optionally, the node v j The spatiotemporal vector of the node v is pooled to obtain the pooled spatiotemporal representation. j The cross-time fusion vector is pooled to obtain the pooled node representation; the pooled spatiotemporal representation and the pooled node representation are multiplied to obtain the node v i The corresponding target context vector is obtained by integrating spatiotemporal features and node features, and the context information mining is more comprehensive.

[0121] Optionally, node v i The corresponding target context vector can be expressed as:

[0122]

[0123] in, Represents node v i The corresponding target context vector; d j,k Represents node v j The spatial vector of node v j The time vector of For node v j The space-time vector of M represents the parameter matrix of the fusion time vector and space vector; Represents a set of N first-order neighbor nodes; For node v j The cross-time fusion vector; pooling is a pooling operation; To pool spatiotemporal representations; Represents the pooling node.

[0124] Furthermore, for the target context vector and node v k The cross-time fusion vector of the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v i The second POI representation vector of the corresponding POI.

[0125] In some implementations, the node v k The cross-time fusion vector and node v i With node v k The spatial vectors between the nodes are fused to obtain the node v k The cross-temporal fusion vector of node v k The cross-temporal fusion vector and the target context vector are concatenated to obtain the node v i At node v k To node v i Node v of the transitive relationship i The corresponding splicing vector; for node v i Sum the concatenated vectors under each transfer relationship to get the node v i The corresponding second POI representation vector.

[0126] Optionally, the acquisition of the second POI representation vector can be expressed as:

[0127]

[0128] in, and G (t) is a learnable weight parameter; d i,k Represents node v i With node v k The space vector between is node v k The cross-time fusion vector, the cross-time fusion vector and the spatial vector d i,k After fusion, we get node v k The cross-temporal fusion vector It is the target context vector, which is spliced ​​with the cross-temporal and spatial fusion vector to obtain the splicing vector, and then the second POI representation vector is obtained, which fully mines the pattern information in time and space, captures multi-dimensional information, and improves the representation ability of the POI representation vector.

[0129] For example, Figure 3A As shown in the figure, determine the associated POI graphs of multiple time units, such as Morningτ, Middayt and Nightτ′, the time intervals are represented as τ-t, tt and t-τ′ respectively, and select N first-order neighbor nodes (corresponding to K-Nearest Context in the figure) from the POI graph and the associated POI graph respectively, for example, v j1 、v j2 、v j3 and v j4 , the corresponding space-time vector is determined by the space vector and time vector of N first-order neighbor nodes. The space vector and time vector can be expressed as: (v j1 ,d k,j1 )、(v j2 ,d k,j2 )、(v j1 ,τ-t)、(v j2 ,τ-t)、(v j3 ,tt) and (v j4 ,t-τ′); pooling operations are performed on the spatiotemporal vector and the target context vector to obtain the node v i The target context vector of node v is determined according to the target context vector i The second POI representation vector.

[0130] In this embodiment, after determining the first POI representation vector of each POI node in the POI graph, the associated POI graph in the adjacent time unit is obtained, and the nearest first-order neighbor node is screened according to the POI graph and the associated POI graph, that is, the spatial distance is determined according to the latitude and longitude coordinates of the node, and then N first-order neighbor nodes are screened according to the spatial distance, and the time interval between the node and its N first-order neighbor nodes is further determined, and the third message mode is determined according to the spatial interval and the time interval, and the node v in the N first-order neighbor nodes is sent according to the third message mode. jThe first POI representation vector in the POI map and the associated POI map is fused across time to obtain a cross-time fusion vector that integrates multiple time units, and then the target context vector is obtained based on the cross-time fusion vector and the corresponding space-time vector. The target context vector fully combines the multi-time neighbor information defined by the context pattern. The target context vector and the cross-time node representation vector are aggregated to obtain the second POI representation vector. The second POI representation vector fully mines the pattern information in time and space, captures multi-dimensional information, improves the representation ability of the POI representation vector, and solves the problem of insufficient learning ability of traditional models for multi-time POI representation.

[0131] Figure 4 is a schematic diagram of another information processing method provided by an embodiment of the present disclosure; Figure 4 As shown, the method comprises the following steps:

[0132] S401, obtaining POI images corresponding to at least two consecutive time units.

[0133] In the embodiments of the present disclosure, the implementation method of step S401 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0134] S402: Obtain a first relationship pattern and a first message pattern between a node and its second-order neighbor nodes in a POI graph.

[0135] In the embodiments of the present disclosure, the implementation method of step S402 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0136] S403, determine node v k 'To node v i The number of paths along which the message is delivered.

[0137] In the embodiments of the present disclosure, the implementation method of step S403 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0138] S404: According to the number of paths for transmitting the message, node v in the first message mode is k 'Filter to obtain the second message mode, the second-order neighbor node v in the second message mode k To node v i The number of paths for transmitting messages is greater than or equal to the set number.

[0139] In the embodiments of the present disclosure, the implementation method of step S404 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0140] S405, according to the node v k and the second message model, constraining the first relationship model to obtain the second relationship model.

[0141] In the embodiments of the present disclosure, the implementation method of step S405 can be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0142] S406: Perform information mining on the POI graph within a time unit based on the second relationship model and the first message model to obtain a first POI representation vector of the POI.

[0143] In the embodiments of the present disclosure, the implementation method of step S406 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0144] S407, obtaining the associated POI graph of the POI graph.

[0145] In the embodiments of the present disclosure, the implementation method of step S407 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0146] S408, determining the node v according to the POI graph and the associated POI graph i Arriving at the second-order neighbor node v k The nearest N first-order neighbor nodes passed through, and determine the node v i The spatial distance and time interval between N first-order neighbor nodes.

[0147] In the embodiments of the present disclosure, the implementation method of step S408 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0148] S409, according to node v i and N first-order neighbor nodes, as well as the spatial distance and time interval, determine the node v k To node v i The third messaging mode for delivering messages.

[0149] In the embodiments of the present disclosure, the implementation method of step S409 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0150] S410: Obtain a second POI representation vector of the POI according to the third message mode and the first POI representation vectors of N first-order neighbor nodes.

[0151] In the embodiments of the present disclosure, the implementation method of step S410 may be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0152] For example, Figure 4A As shown, in this embodiment, the POI images of multiple time units are Conduct POI feature mining, determine the first POI representation vector within the time unit through operations such as long-view pattern mining, first-order neighbor context pooling, and pattern aggregation within time, and further conduct wide-view pattern mining. Based on spatial K-nearest neighbor context pooling and the first POI representation vector, perform pattern aggregation between time units to obtain the second POI representation vector, which is the dynamic point of interest representation. Apply accurate dynamic point of interest representation to downstream tasks, such as advertising systems, recommendation systems, and travel planning, to improve POI representation accuracy and adapt to more application scenarios.

[0153] In this embodiment, a first message mode and a first relationship mode are determined according to the path relationship and the message transmission path between a node and its second-order neighbor nodes in a POI graph, the number of paths for transmitting messages is determined based on the first message mode and constraints are imposed based on the number of paths to obtain a second relationship mode, in which second-order neighbor nodes with richer information are retained, a message representation vector is further acquired based on the context vectors of the second-order neighbor nodes and the context vectors of the first-order neighbor nodes, a first POI representation vector of the node is obtained by aggregation based on the message representation vector and the weight coefficient of the node, and a cross-time fusion vector is further obtained based on the first POI representation vectors of multiple time units, and a target context vector of the comprehensive context environment and the spatiotemporal environment is determined by combining the cross-time fusion vector with the spatial vector and the time vector, and a second POI representation vector is obtained based on the target context vector and the cross-time fusion vector, thereby improving the representation capability of the POI representation vector and solving the problem that the traditional model has insufficient learning capability for multi-time POI representation.

[0154] Figure 5 is a structural block diagram of an information processing device provided by an embodiment of the present disclosure; Figure 5 As shown, the information processing device 500 includes:

[0155] The first acquisition module 501 is used to acquire the POI graphs corresponding to at least two consecutive time units;

[0156] The second acquisition module 502 is used to acquire a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in the POI graph, and constrain the first relationship pattern based on the first message pattern to obtain a second relationship pattern;

[0157] A third acquisition module 503 is used to perform information mining in a time unit on the POI graph based on the second relationship model and the first message model to obtain a first POI representation vector of the POI;

[0158] The fourth acquisition module 504 is used to acquire the associated POI map of the POI map;

[0159] The information processing module 505 is used to perform information mining between time units on the POI graph and the associated POI graph, and obtain a second POI representation vector of the POI in combination with the first POI representation vector.

[0160] In some implementations, the second acquisition module 502 includes:

[0161] Get the node v in the POI graph i Arrival node v i The second-order neighbor node v k 'path;

[0162] Determine the node v based on the path i To the first-order neighbor node v j The first path relationship of node v j To node v k 'The second path relationship;

[0163] Based on the first path relationship and the second path relationship, a first relationship pattern is generated.

[0164] In some implementations, the second acquisition module 502 includes:

[0165] Extract node v from the first relational schema i Arrival node v k 'Candidate paths, and determine the nodes v that each candidate path passes through j ;

[0166] Generated at node v k 'To node v i Passing the message through node v j First message mode.

[0167] In some implementations, the second acquisition module 502 includes:

[0168] Determine the node v k 'To node v i The number of paths along which the message is delivered;

[0169] According to the number of paths for transmitting messages, for the node v in the first message mode k 'Filter to obtain the second message mode, the second-order neighbor node v in the second message mode k To node vi The number of paths for transmitting messages is greater than or equal to the set number;

[0170] According to the node v k and the second message model, constraining the first relationship model to obtain the second relationship model.

[0171] In some implementations, the second acquisition module 502 includes:

[0172] According to the number of paths for delivering messages, from node v in the first message pattern k 'Determine the second-order neighbor node v to be deleted k ”;

[0173] Based on node v k "Perform node updates on the first message mode to obtain a second message mode.

[0174] In some implementations, the second acquisition module 502 includes:

[0175] Get node v k The weight coefficient of

[0176] According to the node v k , node v k The weight coefficient and the second message pattern generate a second relationship pattern.

[0177] In some implementations, the second acquisition module 502 includes:

[0178] According to the second message mode, determine the node v k To node v i The number of paths along which the message is delivered;

[0179] According to the first relationship mode, determine the node v i The total number of paths to all nodes;

[0180] Determine the node v based on the number of paths for transmitting messages and the total number of paths k The weight coefficient of .

[0181] In some implementations, the third acquisition module 503 includes:

[0182] According to the first message mode, determine the node v k To node v i The node v through which the message is transmitted j , and get the node v j The context vector of

[0183] According to the second relationship mode, determine the node v k , and get the node v k The context vector of

[0184] Based on node v j The context vector and node v k The context vector of node v k With node v i The message representation vector between

[0185] For node v k With node v i Aggregate the message representation vectors between nodes to get node v i The first POI representation vector of the corresponding POI.

[0186] In some implementations, the third acquisition module 503 includes:

[0187] Based on node v k The weight coefficient for node v k With node v i The message representation vectors corresponding to each message path between nodes are weighted summed to obtain node v i At node v k The relationship representation vector under the path relationship;

[0188] For node v i The relationship representation vectors under each path relationship are summed to obtain the node v i The corresponding first POI representation vector.

[0189] In some implementations, the information processing module 505 includes:

[0190] According to the POI graph and the associated POI graph, determine the node v i Arriving at the second-order neighbor node v k The nearest N first-order neighbor nodes passed through, and determine the node v i The spatial distance and time interval between N first-order neighbor nodes;

[0191] According to the node v i and N first-order neighbor nodes, as well as the spatial distance and time interval, determine the node v k To node v i The third messaging mode for delivering messages;

[0192] A second POI representation vector of the POI is obtained according to the third message mode and the first POI representation vectors of each of the N first-order neighbor nodes.

[0193] In some implementations, the information processing module 505 includes:

[0194] According to the third message mode, determine the node v among the N first-order neighbor nodes j, and for node v in the POI graph j The first POI representation vector and the node v in the associated POI graph j The first POI representation vector of node v is obtained by performing cross-time fusion of the representation vectors. j The cross-time fusion vector of

[0195] According to the third message mode, the time-space vectors of N first-order neighbor nodes are determined, and the node v is determined according to the time-space vector and the respective cross-time fusion vectors of the N first-order neighbor nodes. k To node v i The target context vector when delivering the message;

[0196] For the target context vector and node v k The cross-time fusion vector of the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v i The second POI representation vector of the corresponding POI.

[0197] In some implementations, the information processing module 505 includes:

[0198] For node v j The spatiotemporal vector is pooled to obtain a pooled spatiotemporal representation;

[0199] For node v j Perform a pooling operation on the cross-time fusion vector to obtain a pooled node representation;

[0200] Perform dot multiplication on the pooled spatiotemporal representation and the pooled node representation to obtain the node v i The corresponding target context vector.

[0201] In some implementations, the information processing module 505 includes:

[0202] For node v k The cross-time fusion vector and node v i With node v k The spatial vectors between the nodes are fused to obtain the node v k The cross-temporal and spatial fusion vector;

[0203] For node v k The cross-temporal fusion vector and the target context vector are concatenated to obtain the node v i At node v k To node v i Node v of the transitive relationship i The corresponding concatenation vector;

[0204] For node v iSum the concatenated vectors under each transfer relationship to get the node v i The corresponding second POI representation vector.

[0205] In some implementations, the information processing module 505 includes:

[0206] According to the third relationship mode, determine the node v j The time vector and space vector of ;

[0207] For node v j The time vector and space vector of node v are fused to obtain j The space-time vector of .

[0208] In some implementations, the information processing module 505 includes:

[0209] According to the POI graph, determine the node v i Arrival node v k The first path set of

[0210] According to the associated POI graph, determine the node v i Arrival node v k The second path set of

[0211] According to the first path set and the second path set, determine the node v i Arrival node v k Candidate first-order neighbor nodes passed through;

[0212] Determine the candidate first-order neighbor nodes and node v i , and according to the spatial distance, determine N first-order neighbor nodes from the candidate first-order neighbor nodes.

[0213] In some implementations, the information processing module 505 includes:

[0214] Obtain the POI map time unit and determine the adjacent time units of the POI map;

[0215] According to the adjacent time units, the associated POI graph of the POI graph is determined.

[0216] In this embodiment, POI graphs of multiple continuous time units are obtained to perform POI feature mining and analysis, and two modes, namely a first relationship mode and a first message mode, are determined based on the nodes in the POI graph and their second-order neighbor nodes. Based on the first message mode, constraints are imposed on the relationship paths in the first relationship mode to avoid the influence of low-quality relationship paths on POI feature extraction. Information mining is performed within the time unit based on the screened second relationship mode and the first message mode to obtain a more complete first POI representation vector for each POI node in the POI graph, and an associated POI graph is further determined based on the time unit. Information mining across time units is performed based on the POI graph and the associated POI graph, and the sufficient first POI representation vector within the time unit is combined to obtain a second POI representation vector with comprehensive time characteristics, so as to fully mine the pattern information of time and space and improve the accuracy of mining the POI representation vector.

[0217] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0218] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0219] Figure 6 600 is a schematic block diagram of an example electronic device for implementing an embodiment of the present disclosure. The electronic device 600 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0220] like Figure 6 As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0221] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0222] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as information processing methods. For example, in some embodiments, the information processing method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the information processing method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the information processing method in any other appropriate manner (e.g., by means of firmware).

[0223] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0224] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0225] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0226] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0227] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0228] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0229] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0230] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An information processing method, wherein: The method comprises: Obtaining POI graphs corresponding to at least two consecutive time units; Acquire a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in the POI graph, and constrain the first relationship pattern based on the first message pattern to obtain a second relationship pattern; Perform information mining in a time unit on the POI graph based on the second relationship model and the first message model to obtain a first POI representation vector of the POI; Obtaining a POI graph associated with the POI graph; Information mining between time units is performed on the POI graph and the associated POI graph, and a second POI representation vector of the POI is obtained by combining the first POI representation vector.

2. The method according to claim 1, wherein: The process of acquiring the first relationship model includes: Get the node v in the POI graph i Arriving at the node v i The second-order neighbor node v k 'path; Determine the node v according to the path i To the first-order neighbor node v j The first path relationship, and the node v j to the node v k 'The second path relationship; The first relationship pattern is generated based on the first path relationship and the second path relationship.

3. The method according to claim 2, wherein: The process of acquiring the first message mode includes: Extract the node v from the first relationship pattern i Arriving at the node v k 'Candidate paths, and determine the node v that each candidate path passes through j ; Generate at the node v k 'To node v i Deliver the message through the node v j First message mode.

4. The method according to claim 2, wherein: The constraining and restricting the first relationship model based on the first message model to obtain the second relationship model includes: Determine the node v k 'To the node v i The number of paths along which the message is delivered; According to the number of paths for transmitting the message, the node v in the first message mode is k 'Filter to obtain the second message mode, the second-order neighbor node v in the second message mode k To the node v i The number of paths for transmitting messages is greater than or equal to the set number; According to the node v k and the second message pattern, constraining the first relationship pattern to obtain the second relationship pattern.

5. The method according to claim 4, wherein: The node v in the first message mode is k 'Filter to get the second message mode, including: According to the number of paths for transmitting the message, from the node v in the first message mode k 'Determine the second-order neighbor node v to be deleted k ”; Based on the node v k "Perform node updates on the first message mode to obtain the second message mode.

6. The method according to claim 4, wherein: According to the node v k and the second message mode, constraining the first relationship mode to obtain the second relationship mode, including: Get the node v k The weight coefficient of According to the node v k , the node v k The weight coefficient and the second message pattern are used to generate the second relationship pattern.

7. The method according to claim 6, wherein: The obtaining of the node v k The weight coefficients include: According to the second message mode, determine the node v k To the node v i The number of paths along which the message is delivered; According to the first relationship mode, determine the node v i The total number of paths to all nodes; Determine the node v according to the number of paths for transmitting the message and the total number of paths k The weight coefficient of .

8. The method according to claim 4, wherein: The performing information mining within a time unit on the POI graph based on the second relationship mode and the first message mode to obtain a first POI representation vector of the POI includes: According to the first message mode, determine the node v k To node v i The node v through which the message is transmitted j , and get the node v j The context vector of According to the second relationship mode, determine the node v k , and get the node v k The context vector of Based on the node v j The context vector and the node v k The context vector of the node v is obtained k With node v i The message representation vector between For the node v k With node v i Aggregate the message representation vectors between nodes to obtain the node v i The first POI representation vector of the corresponding POI.

9. The method according to claim 8, wherein: The pair of nodes v k With node v i Aggregate the message representation vectors between nodes to obtain the node v i The first POI representation vector of the corresponding POI includes: Based on the node v k The weight coefficient for the node v k With node v i The message representation vectors corresponding to each message path between the nodes are weighted summed to obtain the node v i At node v k The relationship representation vector under the path relationship; For the node v i The relationship representation vectors under each path relationship are summed to obtain the node v i The corresponding first POI representation vector.

10. The method according to any one of claims 1 to 9, wherein: The performing information mining between time units on the POI graph and the associated POI graph, and obtaining a second POI representation vector of the POI in combination with the first POI representation vector, includes: Determine the node v according to the POI graph and the associated POI graph i Arriving at the second-order neighbor node v k The nearest N first-order neighbor nodes passed through, and determine the node v i The spatial distance and time interval between N first-order neighbor nodes; According to the node v i and N first-order neighbor nodes, as well as the spatial distance and time interval, determine the node v k To node v i The third message mode for delivering messages; A second POI representation vector of the POI is obtained according to the third message mode and the first POI representation vectors of each of the N first-order neighbor nodes.

11. The method according to claim 10, wherein: The obtaining, according to the third message mode and the first POI representation vectors of each of the N first-order neighbor nodes, a second POI representation vector of the POI includes: According to the third message mode, determine the node v among the N first-order neighbor nodes j , and for the node v in the POI graph j The first POI representation vector and the node v in the associated POI graph j The first POI representation vector of the node v is obtained by performing cross-time fusion of the representation vectors. j The cross-time fusion vector of According to the third message mode, the space-time vectors of the N first-order neighbor nodes are determined, and according to the space-time vector and the respective cross-time fusion vectors of the N first-order neighbor nodes, the node v is determined. k To node v i The target context vector when delivering the message; For the target context vector and the node v k The cross-time fusion vector, and the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v i The second POI representation vector of the corresponding POI.

12. The method according to claim 11, wherein: The node v i The process of determining the corresponding target context vector includes: For the node v j The spatiotemporal vector is pooled to obtain a pooled spatiotemporal representation; For the node v j Perform a pooling operation on the cross-time fusion vector to obtain a pooled node representation; Perform a dot multiplication on the pooled spatiotemporal representation and the pooled node representation to obtain the node v i The corresponding target context vector.

13. The method according to claim 11, wherein: The node v i The process of determining the second POI representation vector of the corresponding POI includes: For the node v k The cross-time fusion vector and the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v k The cross-temporal and spatial fusion vector; For the node v k The cross-temporal and spatial fusion vector is concatenated with the target context vector to obtain the node v i At the node v k To node v i Node v of the transitive relationship i The corresponding concatenation vector; For the node v i The concatenated vectors under each transfer relationship are summed to obtain the node v i The corresponding second POI representation vector.

14. The method according to claim 12, wherein: The node v is determined according to the third relationship mode. j The space-time vector of , including: According to the third relationship mode, determine the node v j The time vector and space vector of ; For the node v j The time vector and space vector are fused to obtain the node v j The space-time vector of .

15. The method according to claim 10, wherein: The step of determining the node v according to the POI graph and the associated POI graph i Arrival node v k The nearest N first-order neighbor nodes passed through include: According to the POI graph, determine the node v i Arrival node v k The first path set of According to the associated POI graph, determine the node v i Arrival node v k The second path set of Determine the node v according to the first path set and the second path set i Arrival node v k Candidate first-order neighbor nodes passed through; Determine the candidate first-order neighbor node and the node v i and determining the N first-order neighbor nodes from the candidate first-order neighbor nodes according to the spatial distance.

16. The method according to claim 10, wherein: The obtaining of the associated POI graph of the POI graph includes: Acquire the POI map time unit, and determine the adjacent time units of the POI map; According to the adjacent time units, a related POI graph of the POI graph is determined.

17. An information processing device, comprising: A first acquisition module is used to acquire a point of interest (POI) map corresponding to at least two consecutive time units; A second acquisition module is used to acquire a first relationship pattern and a first message pattern between a node and its second-order neighboring nodes in the POI graph, and to constrain the first relationship pattern based on the first message pattern to obtain a second relationship pattern; A third acquisition module is used to perform information mining in a time unit on the POI graph based on the second relationship mode and the first message mode to obtain a first POI representation vector of the POI; A fourth acquisition module is used to acquire a POI map associated with the POI map; The information processing module is used to perform information mining between time units on the POI graph and the associated POI graph, and obtain a second POI representation vector of the POI in combination with the first POI representation vector.

18. The device according to claim 17, wherein: The second acquisition module includes: Get the node v in the POI graph i Arriving at the node v i The second-order neighbor node v k 'path; Determine the node v according to the path i To the first-order neighbor node v j The first path relationship, and the node v j to the node v k 'The second path relationship; The first relationship pattern is generated based on the first path relationship and the second path relationship.

19. The device according to claim 18, wherein: The second acquisition module includes: Extract the node v from the first relationship pattern i Arriving at the node v k 'Candidate paths, and determine the node v that each candidate path passes through j ; Generate at the node v k 'To node v i Deliver the message through the node v j First message mode.

20. The device according to claim 18, wherein The second acquisition module includes: Determine the node v k 'To the node v i The number of paths along which the message is delivered; According to the number of paths for transmitting the message, the node v in the first message mode is k 'Filter to obtain the second message mode, the second-order neighbor node v in the second message mode k To the node v i The number of paths for transmitting messages is greater than or equal to the set number; According to the node v k and the second message pattern, constraining the first relationship pattern to obtain the second relationship pattern.

21. The device according to claim 20, wherein: The second acquisition module includes: According to the number of paths for transmitting the message, from the node v in the first message mode k 'Determine the second-order neighbor node v to be deleted k ”; Based on the node v k "Perform node updates on the first message mode to obtain the second message mode.

22. The device according to claim 20, wherein: The second acquisition module includes: Get the node v k The weight coefficient of According to the node v k , the node v k The weight coefficient and the second message pattern are used to generate the second relationship pattern.

23. The device according to claim 22, wherein: The second acquisition module includes: According to the second message mode, determine the node v k To the node v i The number of paths along which the message is delivered; According to the first relationship mode, determine the node v i The total number of paths to all nodes; Determine the node v according to the number of paths for transmitting the message and the total number of paths k The weight coefficient of .

24. The device according to claim 20, wherein: The third acquisition module includes: According to the first message mode, determine the node v k To node v i The node v through which the message is transmitted j , and get the node v j The context vector of According to the second relationship mode, determine the node v k , and get the node v k The context vector of Based on the node v j The context vector and the node v k The context vector of the node v is obtained k With node v i The message representation vector between For the node v k With node v i Aggregate the message representation vectors between nodes to obtain the node v i The first POI representation vector of the corresponding POI.

25. The device according to claim 24, wherein: The third acquisition module includes: Based on the node v k The weight coefficient for the node v k With node v i The message representation vectors corresponding to each message path between the nodes are weighted summed to obtain the node v i At node v k The relationship representation vector under the path relationship; For the node v i The relationship representation vectors under each path relationship are summed to obtain the node v i The corresponding first POI representation vector.

26. The device according to any one of claims 17 to 25, wherein: The information processing module comprises: Determine the node v according to the POI graph and the associated POI graph i Arriving at the second-order neighbor node v k The nearest N first-order neighbor nodes passed through, and determine the node v i The spatial distance and time interval between N first-order neighbor nodes; According to the node v i and N first-order neighbor nodes, as well as the spatial distance and time interval, determine the node v k To node v i The third message mode for delivering messages; A second POI representation vector of the POI is obtained according to the third message mode and the first POI representation vectors of each of the N first-order neighbor nodes.

27. The device according to claim 26, wherein: The information processing module comprises: According to the third message mode, determine the node v among the N first-order neighbor nodes j , and for the node v in the POI graph j The first POI representation vector and the node v in the associated POI graph j The first POI representation vector of the node v is obtained by performing cross-time fusion of the representation vectors. j The cross-time fusion vector of According to the third message mode, the space-time vectors of the N first-order neighbor nodes are determined, and according to the space-time vector and the respective cross-time fusion vectors of the N first-order neighbor nodes, the node v is determined. k To node v i The target context vector when delivering the message; For the target context vector and the node v k The cross-time fusion vector, and the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v i The second POI representation vector of the corresponding POI.

28. The device according to claim 27, wherein The information processing module comprises: For the node v j The spatiotemporal vector is pooled to obtain a pooled spatiotemporal representation; For the node v j Perform a pooling operation on the cross-time fusion vector to obtain a pooled node representation; Perform a dot multiplication on the pooled spatiotemporal representation and the pooled node representation to obtain the node v i The corresponding target context vector.

29. The device according to claim 27, wherein: The information processing module comprises: For the node v k The cross-time fusion vector and the node v i With node v k The spatial vectors between the nodes are fused to obtain the node v k The cross-temporal and spatial fusion vector; For the node v k The cross-temporal and spatial fusion vector is concatenated with the target context vector to obtain the node v i At the node v k To node v i Node v of the transitive relationship i The corresponding concatenation vector; For the node v i The concatenated vectors under each transfer relationship are summed to obtain the node v i The corresponding second POI representation vector.

30. The device according to claim 28, wherein The information processing module comprises: According to the third relationship mode, determine the node v j The time vector and space vector of ; For the node v j The time vector and space vector are fused to obtain the node v j The space-time vector of .

31. The device according to claim 26, wherein: The information processing module comprises: According to the POI graph, determine the node v i Arrival node v k The first path set of According to the associated POI graph, determine the node v i Arrival node v k The second path set of Determine the node v according to the first path set and the second path set i Arrival node v k Candidate first-order neighbor nodes passed through; Determine the candidate first-order neighbor node and the node v i and determining the N first-order neighbor nodes from the candidate first-order neighbor nodes according to the spatial distance.

32. The apparatus of claim 26, wherein: The information processing module comprises: Acquire the POI map time unit, and determine the adjacent time units of the POI map; According to the adjacent time units, a related POI graph of the POI graph is determined.

33. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 16.

34. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-16.

35. A computer program product comprising a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 16.

Images