A platform user interest recommendation method and system based on artificial intelligence

Abstract

The present application relates to the field of information push technology, and in particular to a method and system for recommending user interests on a platform based on artificial intelligence. The method includes: obtaining user action data and environmental perception data, analyzing the user action data and environmental perception data, and constructing a unified spatiotemporal semantic network; analyzing the characteristics of the scene in which the user is located based on the unified spatiotemporal semantic network, and determining a multi-dimensional dynamic scene feature set; analyzing the multi-dimensional dynamic scene feature set, determining the current user's composite scene demand information, and retrieving and pushing the user's real-time demand information based on the composite scene demand information. The present application improves the perception ability of push information for user demand scenarios, realizes multi-dimensional dynamic adaptation of push information to platform users, improves the accuracy of information push for platform users, and meets users' composite information needs in complex scenarios.

Description

Technical Field

[0001] The present application relates to the field of information push technology, and in particular to a method and system for recommending platform user interests based on artificial intelligence. Background Art

[0002] As the core carrier of urban digital transformation, the smart city service platform integrates real-time data across fields such as transportation, energy, environmental protection, emergency response, and public services, aiming to improve urban governance efficiency and residents' living experience through intelligent means.

[0003] However, the information push strategy push logic of existing smart city service platforms usually relies on static rules or single-dimensional labels, lacks the ability to accurately perceive user demand scenarios, and does not fully consider the information coupling needs brought about by scenario intersections, resulting in a misalignment between the pushed information content and the user's real-time needs. Summary of the Invention

[0004] This application provides an artificial intelligence-based platform user interest recommendation method and system to solve the above technical problems.

[0005] In a first aspect, the present application provides a method for recommending user interests on a platform based on artificial intelligence, the method comprising:

[0006] Acquire user action data and environmental perception data, analyze the user action data and environmental perception data, and construct a unified spatiotemporal semantic network;

[0007] Analyzing the scene characteristics of the user based on the unified spatiotemporal semantic network to determine a multi-dimensional dynamic scene feature set;

[0008] The multi-dimensional dynamic scene feature set is analyzed to determine the composite scene demand information of the current user, and the user's real-time demand information is retrieved and pushed based on the composite scene demand information.

[0009] Through this solution, user action data and environmental perception data are jointly analyzed, and a unified spatiotemporal semantic network is constructed to conduct multi-dimensional dynamic scene feature analysis on the user's scene. Based on the obtained multi-dimensional dynamic scene feature set, the current user's complex scene demand information is analyzed, and the user's real-time demand information is retrieved and pushed based on the complex scene demand information, thereby improving the perception of the user's demand scenario by the pushed information, realizing multi-dimensional dynamic adaptation of the pushed information to the platform users, improving the accuracy of information push for platform users, and meeting the user's complex information needs in complex scenarios.

[0010] Optionally, the user action data includes user stay action information and historical search records;

[0011] The environmental perception data includes real-time traffic status data, meteorological status data, public facility operation status data and public event data.

[0012] This solution clarifies and constrains the specific content of user action data and environmental perception data. By integrating user action data (stop actions, search records) with environmental perception data (traffic, weather, facilities, events), a "user-environment" two-way interaction model is constructed. This aligns user actions with real-time environmental status, identifies the environmental sensitivity of user needs, and significantly improves the matching degree between the push information obtained from subsequent analysis and the user's real-time scenario.

[0013] Optionally, analyzing the user action data and environmental perception data to construct a unified spatiotemporal semantic network includes:

[0014] Analyzing the environmental perception data based on the user's stay action information, and screening and determining the local environmental perception data corresponding to the area where the user is located;

[0015] Performing timestamp alignment processing on the real-time traffic status data, the meteorological status data, the public facility operation status data, and the public event data in the local environment perception data to determine the local environment perception synchronization data under the same time axis;

[0016] A graph neural network algorithm based on the fusion attention mechanism uses the user action data as the attention target node and the data corresponding to the local environment perception synchronization data as the neighbor nodes to construct a basic association graph;

[0017] Determine an attention graph neural network based on the basic association graph, using the real-time relative relationship between the entity corresponding to the neighbor node and the entity corresponding to the attention target node as the attention weight between the neighbor node and the corresponding attention target node;

[0018] According to the dynamic attention update strategy, the attention graph neural network is dynamically updated to construct the unified spatiotemporal semantic network in real time.

[0019] Through this solution, the user's stay action information and the corresponding environmental perception data are aligned in time and space to obtain local environmental perception synchronization data, avoiding the spatiotemporal misalignment between subsequent push information and user needs. Based on the graph neural network algorithm, the user action data is used as the attention target node, and the various data in the corresponding local environmental perception synchronization data are used as neighbor nodes. The basic association graph of the relationship between the user and the environment is represented by the graph structure of the fusion attention mechanism, and the attention weight used to characterize the degree of association between the user and the environment is introduced to construct an attention graph neural network. The attention graph neural network is dynamically updated through a dynamic attention update strategy to realize the construction of a unified spatiotemporal semantic network, so that the subsequent analysis and push information analysis process can be quickly and dynamically adapted to changes in the user's scene.

[0020] Optionally, dynamically updating the attention graph neural network comprises at least the following steps:

[0021] Extracting, based on the environmental perception data, an influence range of a physical perception device corresponding to each item of data in the environmental perception data;

[0022] Drawing user movement trajectory information based on the user's stay action information;

[0023] The dynamic attention update strategy includes a passive update mechanism and an active update mechanism;

[0024] The passive update mechanism: when the user movement trajectory information changes, the neighbor nodes corresponding to the current attention target node in the attention graph neural network are dynamically updated according to the relative position relationship between the influence range and the change in the user's stop action;

[0025] The active update mechanism, based on a long-short-term memory collaborative algorithm, predicts the user's subsequent movement trajectory based on the user's movement trajectory information and historical search records. Based on the predicted movement trajectory, the subsequent environmental perception data nodes involved are preloaded as pre-update nodes corresponding to several of the neighboring nodes. This solution utilizes a dynamic attention update strategy that includes passive and active update mechanisms, and a dual-channel update strategy to automatically and rapidly update node relationships in the attention graph neural network as the user moves. Simultaneously, by predicting the user's movement trajectory, the efficiency of subsequent user information demand analysis is improved, reducing data push latency and hardware computing pressure.

[0026] Optionally, analyzing the scene characteristics of the user according to the unified spatiotemporal semantic network to determine a multi-dimensional dynamic scene feature set includes:

[0027] Extracting interaction features between user actions and their environment based on the dynamic attention association relationship between the attention target node and the corresponding neighbor nodes in the unified spatiotemporal semantic network;

[0028] extracting time-sensitive features based on the temporal continuity of the interactive features;

[0029] Extracting spatial correlation features based on the spatial topological relationship of the interactive features;

[0030] Extracting event response features according to the event triggering conditions of the interactive features;

[0031] Analyzing the spatiotemporal coupling strength among the time-sensitive features, the spatial correlation features, and the event-responsive features to identify scene cross-combination patterns;

[0032] Constructing the multi-dimensional dynamic scene feature set according to the scene cross-combination pattern;

[0033] The cross-combination mode includes a time-space superposition scenario, a time-event triggering scenario, a space-event linkage scenario, and a multi-dimensional linkage scenario.

[0034] Through this solution, based on the interaction characteristics between user actions and their environment, the impact of environmental factors on user information needs is divided into three dimensions: time-sensitive, space-related, and event-responsive. These reflect the impact of the environment on user information needs from the perspectives of time, space, and events. Through the cross-features between different features, the user's current scenario is identified, thereby avoiding the situation where the pushed information does not fully match the user's needs caused by single-dimensional feature analysis and preset scenario modes.

[0035] Optionally, extracting a time-sensitive feature according to the temporal continuity of the interaction feature includes:

[0036] Based on the user's stay action information and the historical search records, the user's action time segments are divided, and the action frequency and duration in different time periods are marked;

[0037] Based on the real-time traffic status data and the meteorological status data in the same time segment in the environmental perception data, the action frequency and the duration are analyzed, the influence intensity of the internal and external environment on the user action in the current time segment is identified, and a time-sensitive feature vector is generated as the time-sensitive feature.

[0038] This solution uses machine learning algorithms to identify the time period characteristics of user actions. On this basis, it constructs the correlation characteristics between user actions and environmental factors in different time periods. In this way, it analyzes the impact of external conditions on user information needs in the time dimension and generates corresponding time-sensitive feature vectors as time-sensitive features. By establishing a refined time-sensitive feature analysis system, the time dimension analysis process for user information needs has been significantly upgraded from "mechanical time period processing" to "intelligent time series perception."

[0039] Optionally, extracting spatial correlation features based on the spatial topological relationship of the interactive features includes:

[0040] Based on the user's stay action information, a dynamic spatial grid is constructed with the user's stay location as the center, and the grid is divided into a core area and a peripheral radiation area.

[0041] Analyzing the local environment perception synchronization data, extracting the location distribution and operation status of public facilities, combining the real-time traffic status data, identifying the spatial topological relationship between the public facilities and the dynamic spatial grid, and generating a facility accessibility map;

[0042] According to the geographic location keywords involved in the historical search records, high-frequency interest spatial nodes are extracted, the high-frequency interest spatial nodes are marked in the facility accessibility map, and a heat distribution map of interest points is generated as the spatial correlation feature.

[0043] Through this solution, a dynamic spatial grid centered on the user's location is constructed, so that the spatial dimension analysis process can adapt to the user's action scenario. The core area is used to focus on the user's real-time location, and the peripheral radiation area is used to expand the user's potential needs. Combined with real-time traffic status data, a facility accessibility map is generated. Based on the geographic location keywords that reflect the user's spatial dimension needs, high-frequency spatial nodes of interest are extracted and annotated into the facility accessibility map. A heat distribution map of points of interest is generated as a spatial correlation feature, which enables the spatial dimension analysis process for user information needs to complete a leapfrog upgrade from "mechanical location matching" to "scenario-based spatial services."

[0044] Optionally, extracting event response features according to the event triggering condition of the interaction feature includes:

[0045] Analyze the public event data to determine whether the corresponding public event is a strong push demand event;

[0046] If the public event is a strong push demand event, analyzing the spatial overlap between the current user movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event; when the spatial overlap is greater than a preset overlap threshold, determining the current public event data as the event response feature;

[0047] If the public event does not belong to the strong push demand event, the user's event interest in the current public event is analyzed based on the historical search records. When the event interest is greater than the preset interest threshold, the current public event data is determined as the event response feature.

[0048] Through this solution, public events are divided into events with strong push demand and events with non-strong push demand. The impact range of events with strong push demand is used as the dominant factor. The spatial overlap between the current user's movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event is analyzed to determine whether the corresponding public event data should be used as an event response feature. The user's event interest in events with non-strong push demand is used as the dominant factor to determine whether the corresponding public event data should be used as an event response feature. This enables the event dimension analysis process for user information needs to complete a leapfrog upgrade from "mechanical location matching" to "scenario-based spatial services."

[0049] Optionally, analyzing the multi-dimensional dynamic scene feature set to determine the current user's complex scene demand information includes:

[0050] Based on the dynamic priority adjustment strategy, according to the scene cross-combination pattern represented in the multi-dimensional dynamic scene feature set, the primary and secondary driving dimensions of the current user's scene are identified, and a scene influence weight set is constructed;

[0051] According to the scenario impact weight set, in combination with the multidimensional dynamic scenario feature set, multimodal feature fusion is performed on the time-sensitive feature vector, the spatial correlation feature, and the event response feature to generate a composite demand feature vector as the composite scenario demand information;

[0052] The dynamic priority adjustment strategy includes: when the time-space superposition scenario is included in the scenario cross-combination pattern, dynamically allocating scenario priorities based on the urgency of the user's time segment and the accessibility of core area facilities;

[0053] When the event response feature corresponding to the strong push demand event exists in the scenario cross-combination mode, the event response feature is used as the dominant dimension, and the influence weight of the event response feature is increased according to the urgency of the strong push demand event;

[0054] When the public event corresponding to the event response feature in the scene cross-combination mode does not belong to the strong push demand event, the influence weight corresponding to the event response feature is adjusted according to the current event interest.

[0055] This solution solves the problem of ambiguous identification of user information needs under multi-scenario coupling through dynamic priority adjustment and multimodal fusion. At the same time, the dynamic weight mechanism is used to enable the information demand analysis process to quickly respond to dynamic changes in scenarios, so that the analysis process of user information needs can be upgraded from "mechanical feature matching" to "scenario-based spatial services", thereby improving the matching degree between pushed information and users' actual information needs.

[0056] In a second aspect, the present application provides an artificial intelligence-based platform user interest recommendation system, the system comprising:

[0057] An action perception module is used to obtain user action data and environmental perception data, analyze the user action data and environmental perception data, and construct a unified spatiotemporal semantic network;

[0058] A scene analysis module, configured to analyze the characteristics of the scene in which the user is located based on the unified spatiotemporal semantic network and determine a multi-dimensional dynamic scene feature set;

[0059] The push module is used to analyze the multi-dimensional dynamic scene feature set, determine the current user's composite scene demand information, and retrieve and push the user's real-time demand information based on the composite scene demand information. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0061] Figure 1 A schematic diagram of an application scenario provided in one embodiment of the present application;

[0062] Figure 2 A flowchart of a method for recommending user interests on a platform based on artificial intelligence provided in one embodiment of the present application;

[0063] Figure 3 A schematic diagram of the structure of an artificial intelligence-based platform user interest recommendation system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0064] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0065] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document, unless otherwise specified, generally indicates an "or" relationship between the related objects.

[0066] The embodiments of the present application are described in further detail below with reference to the accompanying drawings.

[0067] The information push strategy push logic of existing smart city service platforms usually relies on static rules or single-dimensional labels, lacks the ability to accurately perceive user demand scenarios, and does not fully consider the information coupling needs brought about by scenario intersections, resulting in a misalignment between the pushed information content and the user's real-time needs.

[0068] Based on this, the present application provides a platform user interest recommendation method and system based on artificial intelligence. It conducts a joint analysis of user action data and environmental perception data, constructs a unified spatiotemporal semantic network, and performs a multi-dimensional dynamic scene feature analysis on the user's scene. Based on the obtained multi-dimensional dynamic scene feature set, it analyzes the current user's complex scene demand information, retrieves and pushes the user's real-time demand information based on the complex scene demand information, improves the perception of the user's demand scenario by the pushed information, realizes the multi-dimensional dynamic adaptation of the pushed information to the platform users, improves the accuracy of information push for the platform users, and meets the users' complex information needs in complex scenarios.

[0069] Figure 1 This is a schematic diagram of an application scenario provided by this application. In the process of pushing information to platform users on a smart city service platform, the method provided by this application is applied to improve the perception of pushed information based on user demand scenarios and achieve multi-dimensional dynamic adaptation of pushed information to platform users.

[0070] Specifically, the method of the present application is applied to any server, which communicates with the platform user terminal device and the Internet of Things environmental sensor, obtains the user behavior data provided by the platform user terminal device and the environmental perception data provided by the Internet of Things environmental sensor through the server, performs a joint analysis of the user action data and the environmental perception data, and uses the constructed unified spatiotemporal semantic network to perform a multi-dimensional dynamic scene feature analysis on the user's scene. Based on the obtained multi-dimensional dynamic scene feature set, the current user's complex scene demand information is analyzed, and the user's real-time demand information is retrieved and pushed based on the complex scene demand information, thereby improving the perception ability of the pushed information to the user's demand scene, realizing multi-dimensional dynamic adaptation of the pushed information to the platform users, improving the accuracy of information push for platform users, and meeting the user's complex information needs in complex scenarios.

[0071] For specific implementation methods, please refer to the following embodiments.

[0072] Figure 2 This is a flowchart of a method for recommending user interests on a platform based on artificial intelligence provided by an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0073] S201. Obtain user action data and environmental perception data, analyze the user action data and environmental perception data, and build a unified spatiotemporal semantic network.

[0074] User action data can be the dynamic interaction information of users on the smart city service platform. User action data comes from the positioning module and log system of the platform user terminal device.

[0075] Environmental perception data can be the real-time status information of the user's external environment. Environmental perception data can be obtained through various IoT environmental sensors deployed in the city.

[0076] The unified spatiotemporal semantic network can be a dynamic graph structure based on spatiotemporal correlation, which is used to represent the dynamic interactive relationship between user actions and the environment.

[0077] Specifically, in the information push technology of existing smart city service platforms, user action data and environmental perception data are processed independently, and lack a unified association in the time and space dimensions, resulting in a disconnect between the pushed information and the user's real-time scenario, making it difficult to accurately reflect the user's complex information needs for the smart city service platform under the cross-influence of factors such as time, space, and events. By analyzing user action data and environmental perception data, user actions and environmental perception data are aligned in a unified time and space framework to ensure the time and space correlation between user actions and corresponding environmental perception data, and avoid association misjudgments caused by time and space dislocations. According to the unified time and space framework, based on the attention mechanism of graph neural networks, a graph structure is constructed to characterize the dynamic interactive relationship between user actions and the environment in which they are located, and a unified time and space semantic network is obtained, which provides a basic framework for subsequent analysis of users' complex needs in complex scenarios.

[0078] S202: Analyze the scene characteristics of the user based on the unified spatiotemporal semantic network and determine a multi-dimensional dynamic scene feature set.

[0079] The multi-dimensional dynamic scene feature set may be a set of feature information related to the user's information needs in the scene in which the user is located, starting from the perspectives of time, space, and events.

[0080] Specifically, user needs are often driven by the intersection of multiple scenarios. For example, if a subway failure (event trigger) occurs during commuting (time-space superposition), an alternative route and a late arrival warning need to be pushed simultaneously. However, the information push mechanism of the existing smart city platform is difficult to identify such coupled needs. This solution decomposes the complex physical scenario in which the user is located into basic dimensional features such as time, space, and events, thereby characterizing the impact of the user's complex environment on the user's complex information needs at the data level. The integrated multi-dimensional dynamic scenario feature set provides an information basis for the subsequent user demand analysis process.

[0081] S203: Analyze the multi-dimensional dynamic scene feature set to determine the current user's composite scene demand information, and retrieve and push the user's real-time demand information based on the composite scene demand information.

[0082] The composite scenario demand information may be user demand vector feature information generated by weighted fusion of multi-dimensional dynamic scenario features.

[0083] The user's real-time demand information can be matched push content retrieved from the platform database based on the composite scenario demand information.

[0084] Specifically, based on the multi-dimensional dynamic scene feature set, the scene features under different dimensions are converted into feature vectors and fused and spliced, and the spliced ​​feature vector information is used as the composite scene demand information. Based on the corresponding feature vectors, the push data map in the smart city platform is targetedly searched to obtain the user's real-time demand information, and the user's real-time demand information is pushed to the corresponding user through the platform user device terminal, such as a smart phone.

[0085] Through this solution, user action data and environmental perception data are jointly analyzed, and a unified spatiotemporal semantic network is constructed to conduct multi-dimensional dynamic scene feature analysis on the user's scene. Based on the obtained multi-dimensional dynamic scene feature set, the current user's complex scene demand information is analyzed, and the user's real-time demand information is retrieved and pushed based on the complex scene demand information, thereby improving the perception of the user's demand scenario by the pushed information, realizing multi-dimensional dynamic adaptation of the pushed information to the platform users, improving the accuracy of information push for platform users, and meeting the user's complex information needs in complex scenarios.

[0086] In some embodiments, user action data includes user stay action information and historical search records; environmental perception data includes real-time traffic status data, weather status data, public facility operation status data and public event data.

[0087] User stay action information can be a spatiotemporal record of the user's stay at a specific location in the urban space, including the coordinates of the stay location, the start time of the stay, the end time of the stay, and the duration of the stay. This data is collected through the built-in GPS module or base station positioning service of the platform user's mobile terminal (such as a smartphone, a car-mounted navigation device), and is obtained in combination with the background positioning permission of the platform application.

[0088] Historical search records may refer to the keywords that users actively enter through the search box, voice assistant or quick tags on the platform, as well as their associated interaction logs such as search time, frequency, click results, etc. This data comes from the platform server's persistent storage of user session records and is extracted from the user action log database through a standardized API interface.

[0089] Real-time traffic status data can be used as a quantitative indicator of road network efficiency, including average speed, congestion index, and traffic accident locations. The data is sourced from the city traffic management center's real-time traffic flow monitoring system and synchronized to the smart city service platform via the data open platform interface.

[0090] Meteorological status data can be a set of parameters that describe the current and predicted weather conditions in the corresponding area. The data comes from the meteorological monitoring stations and satellite remote sensing systems of the Meteorological Bureau and is synchronized to the smart city service platform through the data open platform interface.

[0091] Public facility operation status data can be dynamic data reflecting the service status of urban infrastructure. The data is collected through IoT sensors, transmitted to the urban IoT platform via the IoT communication protocol, and then connected to the smart city service platform through the data bus.

[0092] Public event data can refer to information on sudden or planned events that affect urban operations, including traffic control notices, large-scale event arrangements, emergency warning notifications, etc. This data comes from information disclosure platforms, social media monitoring systems and third-party event aggregation services, and is structured and processed using natural language processing technology before being stored in the platform event knowledge base.

[0093] Specifically, when analyzing the information needs of users in smart city platforms, if recommendations are made solely based on explicit user actions, without situating user actions within the context of a dynamic environment (such as real-time traffic and facility status), it is very easy for the recommendation results to be disconnected from the user's actual scenario needs. For example, if a user searches for "indoor parking" during heavy rain, traditional systems will only return a static list of parking lots, ignoring the impact of real-time flooded roads on parking accessibility. By integrating user action data (stop actions, search records) with environmental perception data (traffic, weather, facilities, events), a "user-environment" two-way interaction model is constructed, which aligns user actions with real-time environmental status, identifies the environmental sensitivity of user needs, and significantly improves the match between the push information obtained from subsequent analysis and the user's real-time scenario.

[0094] This solution clarifies and constrains the specific content of user action data and environmental perception data. By integrating user action data (stop actions, search records) with environmental perception data (traffic, weather, facilities, events), a "user-environment" two-way interaction model is constructed. This aligns user actions with real-time environmental status, identifies the environmental sensitivity of user needs, and significantly improves the matching degree between the push information obtained from subsequent analysis and the user's real-time scenario.

[0095] In some embodiments, based on the user's stay action information, the environmental perception data is analyzed, and the local environmental perception data corresponding to the user's area is screened and determined; the real-time traffic status data, meteorological status data, public facility operation status data and public event data in the local environmental perception data are timestamp aligned to determine the local environmental perception synchronization data under the same time axis; based on the graph neural network algorithm of the fusion attention mechanism, the user action data is used as the attention target node, and the various data in the corresponding local environmental perception synchronization data are used as neighbor nodes to construct a basic association graph; according to the basic association graph, the real-time relative relationship between the corresponding entity of the neighbor node and the corresponding entity of the attention target node is used as the attention weight between the neighbor node and the corresponding attention target node to determine the attention graph neural network; according to the dynamic attention update strategy, the attention graph neural network is dynamically updated to construct a unified spatiotemporal semantic network in real time.

[0096] The local environment perception data may be a subset of environment perception data within a preset geographical range (eg, within a radius of 200 meters) around the current location of the user.

[0097] The timestamp alignment process may be a standardized process of uniformly mapping the timing information of heterogeneous data sources to a reference time axis.

[0098] The local environment awareness synchronization data may be local environment awareness data that has completed timestamp alignment processing.

[0099] The graph neural network algorithm integrating the attention mechanism can be a mathematical algorithm that combines graph structure modeling and attention weight calculation to establish a dynamic association between user actions and environmental perception data.

[0100] The attention target node can be a graph node representing the user action entity.

[0101] Neighbor nodes can be graph nodes representing context-aware entities.

[0102] The basic association graph can be a graph structure used to represent the association between the attention target node and the corresponding neighbor nodes.

[0103] The real-time relative relationship may be an indicator used to characterize the degree of correlation between the user and the corresponding environmental perception features, including real-time spatial distance (Euclidean distance) and semantic relevance (cosine similarity).

[0104] The attention weight can be an indicator obtained by quantifying the real-time relative relationship and used to characterize the degree of influence of different neighbor nodes on their corresponding attention target nodes.

[0105] The attention graph neural network can be a structured data network used to characterize the dynamic association relationship between the attention target node and several corresponding neighbor nodes.

[0106] The dynamic attention update strategy can be a mechanism for adjusting the corresponding graph neural network structure in real time according to the user's movement trajectory and environmental changes.

[0107] Specifically, user action data and environmental perception data are stored in independent systems (such as traffic data platforms and meteorological data platforms), lacking a unified spatiotemporal alignment mechanism. If user action data and environmental perception data are processed separately, it is easy to cause spatiotemporal misalignment between the subsequent pushed information and user needs. For example, when providing users with "rain shelter" recommendations, due to the spatiotemporal misalignment between the associated data, it is easy to fail to synchronize the real-time rainfall data with the opening status of surrounding facilities, resulting in the recommendation results including closed facilities. In the absence of a spatiotemporal alignment mechanism, the pushed information cannot be quickly and dynamically adapted to the changes in the user's scene, resulting in the recommended information lagging behind the user's needs and causing information push misalignment. This solution generates a buffer area with a preset radius, such as a 200-meter radius, centered on the user's location. It performs range queries on the spatial database to extract the status data of all traffic, meteorological, public facilities, and public events in the area, aligns the user's stay time period (such as 14:00-14:30), filters out environmental perception data that is not in this time window (such as the traffic snapshot at 13:50), and uses NTP (Network Time Protocol). The algorithm calibrates the clocks of each data source using the Network Time Protocol (NTP) to obtain local environmental perception synchronization data, achieving spatiotemporal alignment between user action data and environmental perception data. A graph neural network algorithm based on the fused attention mechanism encodes feature vectors for user action data and the corresponding local environmental perception synchronization data, respectively. A basic association graph is constructed with the user action data as the attention target node and the various data items in the corresponding local environmental perception synchronization data as neighbor nodes. The attention weights between the attention target node and the neighboring nodes are quantified based on the weighted sum of the real-time spatial distance (Euclidean distance) and semantic relevance (cosine similarity) between the attention target node and the corresponding neighboring nodes. The attention weights between different nodes are introduced into the basic association graph, and an attention graph neural network is constructed to characterize the overall association relationship between the user and the environment. Based on the user's subsequent actions, the dynamic attention update strategy is used to dynamically update the neighbor nodes and corresponding attention weights in the attention graph neural network, thus constructing a unified spatiotemporal semantic network. This allows the subsequent analysis and push information analysis process to be quickly and dynamically adapted to changes in the user's scenario.

[0108] Through this solution, the user's stay action information and the corresponding environmental perception data are aligned in time and space to obtain local environmental perception synchronization data, avoiding the spatiotemporal misalignment between subsequent push information and user needs. Based on the graph neural network algorithm, the user action data is used as the attention target node, and the various data in the corresponding local environmental perception synchronization data are used as neighbor nodes. The basic association graph of the relationship between the user and the environment is represented by the graph structure of the fusion attention mechanism, and the attention weight used to characterize the degree of association between the user and the environment is introduced to construct an attention graph neural network. The attention graph neural network is dynamically updated through a dynamic attention update strategy to realize the construction of a unified spatiotemporal semantic network, so that the subsequent analysis and push information analysis process can be quickly and dynamically adapted to changes in the user's scene.

[0109] In some embodiments, based on the environmental perception data, the influence range of the entity perception device corresponding to each data in the environmental perception data is extracted; based on the user's stay action information, the user's movement trajectory information is drawn; the dynamic attention update strategy includes a passive update mechanism and an active update mechanism; the passive update mechanism: when the user's movement trajectory information changes, according to the relative position relationship between the influence range and the change in the user's stay action, the several neighbor nodes corresponding to the current attention target node in the attention graph neural network are dynamically updated; the active update mechanism: based on the long-short-term memory collaborative algorithm, according to the user's movement trajectory information and the historical search records, the user's subsequent movement trajectory is predicted, and according to the predicted movement trajectory, the subsequent environmental perception data nodes involved are pre-loaded as pre-update nodes corresponding to the several neighbor nodes.

[0110] The physical sensing device can be a physical IoT device used to collect corresponding environmental sensing data.

[0111] User movement trajectory information may be path data that records the changes in the user's spatial position in a continuous time series.

[0112] The impact range may be the effective range of the environmental perception data collected by the physical perception device.

[0113] The relative position relationship may be a relative spatial position relationship between the user's stay position and the influence range of the corresponding entity perception device.

[0114] The long short-term memory collaborative algorithm can be a mathematical algorithm based on the LSTM (Long Short-Term Memory) model to predict the user's subsequent movement trajectory based on the user's historical trajectory.

[0115] The predicted movement trajectory may be a subsequent movement trajectory of the user predicted by a long short-term memory collaborative algorithm.

[0116] The pre-update node may be a node that pre-loads data as a subsequent neighbor node.

[0117] Specifically, as the user moves, the environment in which he is located is in a dynamic change process, which in turn brings about dynamic changes in user demand information. This dynamic change is mapped by adjusting the relationship between each node in the attention graph neural network. By establishing a spatial impact model for each environmental perception device (such as a traffic camera), its coverage range is converted into GeoJSON polygon data and stored in a spatial database, the data quantification processing of the influence range of each entity perception device is realized. When the user's movement trajectory information changes, the relative spatial position between the user and the corresponding environmental influencing factors changes. Since the degree of association between the attention target node representing the user and the neighbor node representing the environmental influencing factors is determined by the real-time spatial distance (Euclidean distance) and semantic relevance (cosine similarity) between the nodes, when the user's movement trajectory information changes, the attention weight needs to be adjusted according to the relative spatial position between the user and the corresponding environmental influencing factors. When the relative distance reflected by the relative spatial position exceeds the preset geographical range centered on the user, the impact of the corresponding neighbor node on the user's information needs is minimized. At this time, the corresponding neighbor node is removed from the corresponding graph structure, and the new neighbor node currently located in the preset geographical range is added to the corresponding graph structure. The attention weight between the new neighbor node and the target attention node is calculated to achieve dynamic local update of the attention graph neural network; the user's historical action trajectory is output to the LSTM model, and a predicted path probability distribution map for a preset update period (such as 10 minutes) is output. Several high-probability waypoints are extracted from the predicted path probability distribution map to construct a predicted movement trajectory. According to the predicted movement trajectory, environmental perception data located around the predicted movement trajectory and within the preset geographical range centered on the user position is extracted as a pre-update node for subsequent neighbor nodes, thereby improving the efficiency of subsequent user information demand analysis and reducing data push delay and hardware computing pressure.

[0118] Through this solution, a dynamic attention update strategy that includes passive update mechanism and active update mechanism is utilized, and a dual-channel update strategy is used to automatically and quickly update the node relationships in the attention graph neural network as the user takes action. At the same time, by predicting the user's action trajectory, the efficiency of subsequent user information demand analysis is improved, and the data push delay and hardware computing pressure are reduced.

[0119] In some embodiments, based on the dynamic attention association relationship between the attention target node and the corresponding neighbor node in the unified spatiotemporal semantic network, the interaction characteristics between the user action and the environment in which it is located are extracted; based on the temporal continuity of the interaction characteristics, time-sensitive features are extracted; based on the spatial topological relationship of the interaction characteristics, spatial correlation features are extracted; based on the event triggering conditions of the interaction characteristics, event response features are extracted; the spatiotemporal coupling strength between time-sensitive features, spatial correlation features and event response features is analyzed to identify scene cross-combination patterns; based on the scene cross-combination patterns, a multidimensional dynamic scene feature set is constructed; the cross-combination patterns include time-space overlay scenarios, time-event triggering scenarios, space-event linkage scenarios and multi-dimensional linkage scenarios.

[0120] Dynamic attention associations can be attention weights that change with user actions.

[0121] The interaction feature can be feature data used to characterize the dynamic correlation between user actions and environmental perception data, that is, the attention weight during the change process.

[0122] Temporal continuity can be the timestamp of the change corresponding to the attention weight.

[0123] Time-sensitive features may be feature information that affects user information needs in the time dimension.

[0124] The spatial topological relationship may be a topological relationship map that describes the spatial location of the user and the locations of surrounding facilities.

[0125] Spatial correlation features may be feature information that affects user information needs in a spatial dimension.

[0126] The event triggering condition may be a triggering condition of the impact of a public event on the user's information needs.

[0127] Event response features may be feature information that has an impact on user information needs in the event dimension.

[0128] The spatiotemporal coupling strength can be an indicator used to characterize the degree of interaction between time, space, and event characteristics.

[0129] The scenario cross-combination pattern can be a typical user demand scenario classification formed by the multi-dimensional interweaving of time, space, and event characteristics.

[0130] The time-space overlay scenario can be a scenario where time-sensitive features and spatial correlation features intersect, such as the route information recommendation triggered by the morning rush hour on weekdays (time) and the road congestion (space) around the company.

[0131] Time-event triggering scenarios can be scenarios where time-sensitive features and event-responsive features intersect, such as route information recommendations triggered by the morning rush hour (time) on weekdays and sudden traffic accidents (events).

[0132] Space-event linkage scenarios can be scenarios where space-related features and event-responsive features intersect, such as emergency channel navigation push triggered by a sudden traffic incident (event) when a user approaches a hospital (space).

[0133] Multi-dimensional linkage scenarios can be scenarios where time-sensitive features, spatial correlation features, and event response features intersect, such as the recommendation of multi-modal transportation connection information triggered by the suspension of subway service (space) during the evening rush hour (time) due to heavy rain (event).

[0134] Specifically, the impact of environmental factors on user information needs is divided into three dimensions: time-sensitive, space-related and event-responsive, reflecting the characteristics of the impact of the environment on user information needs from the perspectives of time, space and events respectively. Through the cross-features between different features, the user's current scene is identified, so as to avoid the situation where the push information does not fully match the user's needs caused by single-dimensional feature analysis and preset scene mode. Based on the time continuity of interactive features, the user's actions are clustered in time segments. According to the user action characteristics under the corresponding environmental perception data in different time segments, the impact of environmental perception data on user information needs in different time segments is identified, and the time-sensitive type features; with the user's real-time location as the center, a spatial grid with a preset geographic radius as the benchmark is generated, the user's accessibility to each facility in the spatial grid is analyzed, and spatial correlation features are obtained; through the event knowledge graph matching mechanism, event response features are generated according to the specific type of public events and the potential association between events and user actions; then a time-space-event three-dimensional feature tensor is constructed, and the potential correlation strength between each dimension is extracted through tensor decomposition to obtain the spatiotemporal coupling strength between different features. The time-sensitive features, spatial correlation features, and event response features with current spatiotemporal coupling strength are matched with historical scene data for similarity, and the corresponding scene cross-combination patterns are identified.

[0135] Through this solution, based on the interaction characteristics between user actions and their environment, the impact of environmental factors on user information needs is divided into three dimensions: time-sensitive, space-related, and event-responsive. These reflect the impact of the environment on user information needs from the perspectives of time, space, and events. Through the cross-features between different features, the user's current scenario is identified, thereby avoiding the situation where the pushed information does not fully match the user's needs caused by single-dimensional feature analysis and preset scenario modes.

[0136] In some embodiments, based on the user's stay action information and historical search records, the user's action time segments are divided, and the action frequency and duration in different time periods are marked; according to the real-time traffic status data and meteorological status data in the same time segment in the environmental perception data, the action frequency and duration are analyzed, and the impact intensity of the internal and external environment on the user's action in the current time segment is identified, and a time-sensitive feature vector is generated as a time-sensitive feature.

[0137] The user action time segment may be a time interval divided into user activity cycles.

[0138] Action frequency and duration can be two-dimensional indicators used to quantify the intensity of a user's activity within a period of time.

[0139] The time-sensitive feature vector may be vector information that represents the characteristics of user information needs changing over time.

[0140] Specifically, user stay data is cleaned, and positioning drift points (such as stays lasting less than 30 seconds) are removed. Time series clustering algorithms (such as K-Shape) are used to aggregate behaviors with similar time distributions into typical time periods, completing the division of user action time segments. Using search analysis engines such as Elasticsearch, a time period-behavior type aggregation index is established, and the frequency distribution of each new time period is calculated in real time to obtain the action frequency in different time segments. At the same time, a streaming computing framework (such as Flink) is used to perform sliding window statistics on the duration to obtain the duration of actions in different time segments. Grey correlation analysis is used to quantify the correlation between real-time traffic status data and meteorological status data and the corresponding action frequency and duration, so as to map the impact intensity of the external environment in the current time segment on user actions. Vector splicing of action characteristics, environmental characteristics and the impact intensity of the corresponding time segment is performed to generate time-sensitive feature vectors as time-sensitive features.

[0141] This solution uses machine learning algorithms to identify the time period characteristics of user actions. On this basis, it constructs the correlation characteristics between user actions and environmental factors in different time periods. In this way, it analyzes the impact of external conditions on user information needs in the time dimension and generates corresponding time-sensitive feature vectors as time-sensitive features. By establishing a refined time-sensitive feature analysis system, the time dimension analysis process for user information needs has been significantly upgraded from "mechanical time period processing" to "intelligent time series perception."

[0142] In some embodiments, based on the user's stay action information, a dynamic spatial grid centered on the user's stay location is constructed, and the core area and the peripheral radiation area are divided; the local environmental perception synchronization data is analyzed, the location distribution and operation status of public facilities are extracted, and combined with the real-time traffic status data, the spatial topological relationship between public facilities and the dynamic spatial grid is identified, and a facility accessibility map is generated; based on the geographic location keywords involved in the historical search records, high-frequency interest spatial nodes are extracted, and the high-frequency interest spatial nodes are marked in the facility accessibility map, and a heat distribution map of interest points is generated as a spatial correlation feature.

[0143] The dynamic spatial grid can be an adaptive geographical partition grid centered on the user's real-time location.

[0144] The core area may be an area covering the user's immediate activity range.

[0145] The peripheral radiation area may be an area covering the potential movement range of the user.

[0146] The location distribution and operation status may be the distribution location and operation status of public facilities.

[0147] The facility accessibility map can be a spatial topological relationship network that quantifies the service capacity of public facilities.

[0148] The geographic location keyword may be a keyword in historical search records that is highly relevant to the geographic location.

[0149] High-frequency interest spatial nodes can be geographic location tags that appear repeatedly in historical search records.

[0150] The POI heat distribution map may be graph structure data that reflects the distribution of user spatial demand information.

[0151] Specifically, the Kalman filter algorithm integrates GPS, WiFi fingerprint, and base station positioning data to determine the user's precise coordinates. A quadtree structure is used to divide the spatial grid, with the user's location as the origin. The grid accuracy of the core area (e.g., 10 meters × 10 meters) and the grid accuracy of the radiation area (e.g., 20 meters × 20 meters) are set. Semantic analysis is performed on the synchronized data of local environmental perception to extract the location distribution and operating status of public facilities (e.g., charging station occupancy rate, subway congestion, etc.). Combined with navigation data mapped by real-time traffic status data, a multidimensional scoring model (e.g., a traffic weight of 40% + a real-time facility status weight of 30% + a user distance weight of 30%) is constructed to generate a comprehensive accessibility index for each facility. A facility accessibility map is generated based on the facility location. A pre-trained language representation model, such as the BERT model, is used to analyze the implicit geographic location features of historical search records and identify geographic location keywords. Based on the frequency and timeliness of different keywords, high-frequency spatial nodes of interest are extracted. By annotating these high-frequency spatial nodes of interest into the facility accessibility map, a heat distribution map of points of interest is generated as a spatial correlation feature.

[0152] Through this solution, a dynamic spatial grid centered on the user's location is constructed, so that the spatial dimension analysis process can adapt to the user's action scenario. The core area is used to focus on the user's real-time location, and the peripheral radiation area is used to expand the user's potential needs. Combined with real-time traffic status data, a facility accessibility map is generated. Based on the geographic location keywords that reflect the user's spatial dimension needs, high-frequency spatial nodes of interest are extracted and annotated into the facility accessibility map. A heat distribution map of points of interest is generated as a spatial correlation feature, which enables the spatial dimension analysis process for user information needs to complete a leapfrog upgrade from "mechanical location matching" to "scenario-based spatial services."

[0153] In some embodiments, public event data is analyzed to determine whether the corresponding public event is a strong push demand event; if the public event is a strong push demand event, the spatial overlap between the current user movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event is analyzed, and when the spatial overlap is greater than a preset overlap threshold, the current public event data is determined to be an event response feature; if the public event is not a strong push demand event, the user's event interest in the current public event is analyzed based on historical search records, and when the event interest is greater than a preset interest threshold, the current public event data is determined to be an event response feature.

[0154] Strong push demand events can be high-priority public events that need to be forced to reach users.

[0155] Spatial overlap can be a quantitative indicator of the geographical intersection between user activity trajectories and event impact areas.

[0156] The preset overlap threshold may be a spatial overlap critical value for determining whether event information needs to be pushed.

[0157] The event interest level may be a quantitative value of the user's attention to the non-mandatory event type.

[0158] The preset interest threshold may be a minimum interest benchmark value for triggering the push of a non-forced demand event.

[0159] Specifically, whether to push information related to public events to users depends mainly on the type of event and the user's interest in the event. When the event type is an emergency disaster, traffic control, or other event that will directly affect users within the affected range, it is necessary to actively push relevant event information to platform users within its affected range. This type of event is classified as a strong push demand event, which does not consider user interest but focuses on the impact range of the event. Through a range overlap quantification algorithm, such as the Jaccard coefficient formula, the spatial overlap between the current user's movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event is analyzed. When the spatial overlap is greater than the preset overlap threshold, , indicating that the user is within the influence range of the current strong push demand event, and the public event data corresponding to the current strong push demand event is determined as an event response feature; if the relevant public event is not a strong push demand event, the user's interest is dominant (for example, if the user has a high interest in public welfare events, then the public welfare event information within the push range is pushed). Based on the historical search records, a TF-IDF weight matrix is ​​constructed to match the event keyword similarity and quantify the user's interest in the public events within the current range. When the event interest is greater than the preset interest threshold, it indicates that the user is interested in the current non-strong push demand event, and the current public event data is determined as an event response feature.

[0160] Through this solution, public events are divided into events with strong push demand and events with non-strong push demand. The impact range of events with strong push demand is used as the dominant factor. The spatial overlap between the current user's movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event is analyzed to determine whether the corresponding public event data should be used as an event response feature. The user's event interest in events with non-strong push demand is used as the dominant factor to determine whether the corresponding public event data should be used as an event response feature. This enables the event dimension analysis process for user information needs to complete a leapfrog upgrade from "mechanical location matching" to "scenario-based spatial services."

[0161] In some embodiments, based on a dynamic priority adjustment strategy, according to the scene cross-combination pattern represented in the multidimensional dynamic scene feature set, the primary and secondary driving dimensions of the current user's scene are identified, and a scene impact weight set is constructed; according to the scene impact weight set, combined with the multidimensional dynamic scene feature set, multimodal feature fusion is performed on the time-sensitive feature vector, the spatial correlation feature and the event response feature to generate a composite demand feature vector as composite scene demand information; the dynamic priority adjustment strategy includes: when the scene cross-combination pattern contains a time-space superposition scene, the scene priority is dynamically allocated based on the urgency of the user's time segment and the accessibility of core area facilities; when there is an event response feature corresponding to a strong push demand event in the scene cross-combination pattern, the event response feature is used as the dominant dimension, and the influence weight of the corresponding event response feature is increased according to the urgency of the corresponding strong push demand event; when the public event corresponding to the event response feature in the scene cross-combination pattern does not belong to a strong push demand event, the influence weight of the corresponding event response feature is adjusted according to the current event interest.

[0162] The dynamic priority adjustment strategy can be a decision-making mechanism that dynamically allocates the weights of features in each dimension according to the combined characteristics of the real-time scenario.

[0163] The primary and secondary driving dimensions can be the main and secondary dimensional features that affect the user's information needs in the current scenario.

[0164] The scenario impact weight set may be a parameter set that quantifies the degree of impact of each scenario dimension on user needs.

[0165] Multimodal feature fusion can be a deep learning process that integrates heterogeneous scene features.

[0166] The composite demand feature vector may be multi-dimensional vector feature information representing the user's comprehensive information demand.

[0167] Specifically, the weight of each dimension is initialized according to the scenario type (such as the initial weight of the time-space overlay scenario: time 45%, space 40%, event 15%). When multiple scenario modes are effective at the same time, a weight overlay mechanism is adopted, but a weight upper limit (such as 75%) is set to prevent the monopoly of a single dimension. When the time-space overlay scenario is included in the scenario cross-combination mode, the benchmark is adjusted in units of 1% based on the urgency of the user's time segment and the accessibility of facilities in the core area, and the weights of the time and space dimensions are adjusted proportionally. For scenarios with strong push demand events, the weight of event response features is exponentially increased based on the urgency of the event (such as the remaining effective time of traffic control), and the weights of other features are compressed. For scenarios with strong push demand events, the weight of event response features is proportionally increased based on the user's interest in the event. The time-sensitive feature vectors, spatial correlation features, and event response features with dynamic impact weights are vector spliced ​​and fused to generate a composite demand feature vector as the composite scenario demand information.

[0168] This solution solves the problem of ambiguous identification of user information needs under multi-scenario coupling through dynamic priority adjustment and multimodal fusion. At the same time, the dynamic weight mechanism is used to enable the information demand analysis process to quickly respond to dynamic changes in scenarios, so that the analysis process of user information needs can be upgraded from "mechanical feature matching" to "scenario-based spatial services", thereby improving the matching degree between pushed information and users' actual information needs.

[0169] Figure 3 This is a structural diagram of an artificial intelligence-based platform user interest recommendation system provided in one embodiment of the present application, such as Figure 3 As shown, an artificial intelligence-based platform user interest recommendation system 300 of this embodiment includes: an action perception module 301, a scene analysis module 302 and a push module 303.

[0170] Action perception module 301, used to obtain user action data and environmental perception data, analyze the user action data and environmental perception data, and construct a unified spatiotemporal semantic network;

[0171] A scene analysis module 302 is configured to analyze the scene characteristics of the user based on the unified spatiotemporal semantic network and determine a multi-dimensional dynamic scene feature set;

[0172] The push module 303 is used to analyze the multi-dimensional dynamic scene feature set, determine the composite scene demand information of the current user, and retrieve and push the user's real-time demand information based on the composite scene demand information.

[0173] Optionally, in the action perception module 301, the user action data includes user stay action information and historical search records;

[0174] The environmental perception data includes real-time traffic status data, meteorological status data, public facility operation status data and public event data.

[0175] Optionally, the action perception module 301 is specifically configured to:

[0176] Analyzing the environmental perception data based on the user's stay action information, and screening and determining the local environmental perception data corresponding to the area where the user is located;

[0177] Performing timestamp alignment processing on the real-time traffic status data, the meteorological status data, the public facility operation status data, and the public event data in the local environment perception data to determine the local environment perception synchronization data under the same time axis;

[0178] A graph neural network algorithm based on the fusion attention mechanism uses the user action data as the attention target node and the data corresponding to the local environment perception synchronization data as the neighbor nodes to construct a basic association graph;

[0179] Determine an attention graph neural network based on the basic association graph, using the real-time relative relationship between the entity corresponding to the neighbor node and the entity corresponding to the attention target node as the attention weight between the neighbor node and the corresponding attention target node;

[0180] According to the dynamic attention update strategy, the attention graph neural network is dynamically updated to construct the unified spatiotemporal semantic network in real time.

[0181] Optionally, when dynamically updating the attention graph neural network, the action perception module 301 is specifically configured to:

[0182] Extracting, based on the environmental perception data, an influence range of a physical perception device corresponding to each item of data in the environmental perception data;

[0183] Drawing user movement trajectory information based on the user's stay action information;

[0184] The dynamic attention update strategy includes a passive update mechanism and an active update mechanism;

[0185] The passive update mechanism: when the user movement trajectory information changes, the neighbor nodes corresponding to the current attention target node in the attention graph neural network are dynamically updated according to the relative position relationship between the influence range and the change in the user's stop action;

[0186] The active update mechanism: based on the long-short-term memory collaborative algorithm, predicts the user's subsequent movement trajectory according to the user movement trajectory information and the historical search records, and preloads the subsequent environmental perception data nodes as pre-update nodes corresponding to several of the neighbor nodes according to the predicted movement trajectory.

[0187] Optionally, the scene analysis module 302 is specifically configured to:

[0188] Extracting interaction features between user actions and their environment based on the dynamic attention association relationship between the attention target node and the corresponding neighbor nodes in the unified spatiotemporal semantic network;

[0189] extracting time-sensitive features based on the temporal continuity of the interactive features;

[0190] Extracting spatial correlation features based on the spatial topological relationship of the interactive features;

[0191] Extracting event response features according to the event triggering conditions of the interactive features;

[0192] Analyzing the spatiotemporal coupling strength among the time-sensitive features, the spatial correlation features, and the event-responsive features to identify scene cross-combination patterns;

[0193] Constructing the multi-dimensional dynamic scene feature set according to the scene cross-combination pattern;

[0194] The cross-combination mode includes a time-space superposition scenario, a time-event triggering scenario, a space-event linkage scenario, and a multi-dimensional linkage scenario.

[0195] Optionally, when extracting time-sensitive features based on the temporal continuity of the interaction features, the scene analysis module 302 is specifically configured to:

[0196] Based on the user's stay action information and the historical search records, the user's action time segments are divided, and the action frequency and duration in different time periods are marked;

[0197] Based on the real-time traffic status data and the meteorological status data in the same time segment in the environmental perception data, the action frequency and the duration are analyzed, the influence intensity of the internal and external environment on the user action in the current time segment is identified, and a time-sensitive feature vector is generated as the time-sensitive feature.

[0198] Optionally, when extracting spatial correlation features based on the spatial topological relationship of the interaction features, the scene analysis module 302 is specifically configured to:

[0199] Based on the user's stay action information, a dynamic spatial grid is constructed with the user's stay location as the center, and the grid is divided into a core area and a peripheral radiation area.

[0200] Analyzing the local environment perception synchronization data, extracting the location distribution and operation status of public facilities, combining the real-time traffic status data, identifying the spatial topological relationship between the public facilities and the dynamic spatial grid, and generating a facility accessibility map;

[0201] According to the geographic location keywords involved in the historical search records, high-frequency interest spatial nodes are extracted, the high-frequency interest spatial nodes are marked in the facility accessibility map, and a heat distribution map of interest points is generated as the spatial correlation feature.

[0202] Optionally, when extracting event response features based on the event triggering conditions of the interaction features, the scenario analysis module 302 is specifically configured to:

[0203] Analyze the public event data to determine whether the corresponding public event is a strong push demand event;

[0204] If the public event is a strong push demand event, analyzing the spatial overlap between the current user movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event; when the spatial overlap is greater than a preset overlap threshold, determining the current public event data as the event response feature;

[0205] If the public event does not belong to the strong push demand event, the user's event interest in the current public event is analyzed based on the historical search records. When the event interest is greater than the preset interest threshold, the current public event data is determined as the event response feature.

[0206] Optionally, the push module 303 is specifically configured to:

[0207] Based on the dynamic priority adjustment strategy, according to the scene cross-combination pattern represented in the multi-dimensional dynamic scene feature set, the primary and secondary driving dimensions of the current user's scene are identified, and a scene influence weight set is constructed;

[0208] According to the scenario impact weight set, in combination with the multidimensional dynamic scenario feature set, multimodal feature fusion is performed on the time-sensitive feature vector, the spatial correlation feature, and the event response feature to generate a composite demand feature vector as the composite scenario demand information;

[0209] The dynamic priority adjustment strategy includes: when the time-space superposition scenario is included in the scenario cross-combination pattern, dynamically allocating scenario priorities based on the urgency of the user's time segment and the accessibility of core area facilities;

[0210] When the event response feature corresponding to the strong push demand event exists in the scenario cross-combination mode, the event response feature is used as the dominant dimension, and the influence weight of the event response feature is increased according to the urgency of the strong push demand event;

[0211] When the public event corresponding to the event response feature in the scene cross-combination mode does not belong to the strong push demand event, the influence weight corresponding to the event response feature is adjusted according to the current event interest.

[0212] The system of this embodiment can be used to execute the method of any of the above embodiments. Its implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. A platform user interest recommendation method based on artificial intelligence, characterized in that: include: Acquire user action data and environmental perception data, analyze the user action data and environmental perception data, and construct a unified spatiotemporal semantic network; Analyzing the scene characteristics of the user based on the unified spatiotemporal semantic network to determine a multi-dimensional dynamic scene feature set; Analyze the multi-dimensional dynamic scene feature set to determine the current user's composite scene demand information, and retrieve and push the user's real-time demand information based on the composite scene demand information; The user action data includes user stay action information and historical search records; The environmental perception data includes real-time traffic status data, meteorological status data, public facility operation status data and public event data; The analyzing the user action data and the environment perception data to construct a unified spatiotemporal semantic network includes: Analyzing the environmental perception data based on the user's stay action information, and screening and determining the local environmental perception data corresponding to the area where the user is located; Performing timestamp alignment processing on the real-time traffic status data, the meteorological status data, the public facility operation status data, and the public event data in the local environment perception data to determine the local environment perception synchronization data under the same time axis; A graph neural network algorithm based on the fusion attention mechanism uses the user action data as the attention target node and the data corresponding to the local environment perception synchronization data as the neighbor nodes to construct a basic association graph; Determine an attention graph neural network based on the basic association graph, using the real-time relative relationship between the entity corresponding to the neighbor node and the entity corresponding to the attention target node as the attention weight between the neighbor node and the corresponding attention target node; According to the dynamic attention update strategy, the attention graph neural network is dynamically updated to construct the unified spatiotemporal semantic network in real time; The dynamically updating the attention graph neural network includes: Extracting, based on the environmental perception data, an influence range of a physical perception device corresponding to each item of data in the environmental perception data; Drawing user movement trajectory information based on the user's stay action information; The dynamic attention update strategy includes a passive update mechanism and an active update mechanism; The passive update mechanism: when the user movement trajectory information changes, the neighbor nodes corresponding to the current attention target node in the attention graph neural network are dynamically updated according to the relative position relationship between the influence range and the change in the user's stop action; The active update mechanism: based on the long-short-term memory collaborative algorithm, predicts the user's subsequent movement trajectory according to the user movement trajectory information and the historical search records, and preloads the subsequent environmental perception data nodes as pre-update nodes corresponding to several of the neighbor nodes according to the predicted movement trajectory.

2. The method according to claim 1, characterized in that The step of analyzing the scene characteristics of the user based on the unified spatiotemporal semantic network and determining a multi-dimensional dynamic scene feature set includes: Extracting interaction features between user actions and their environment based on the dynamic attention association relationship between the attention target node and the corresponding neighbor nodes in the unified spatiotemporal semantic network; extracting time-sensitive features based on the temporal continuity of the interactive features; Extracting spatial correlation features based on the spatial topological relationship of the interactive features; Extracting event response features according to the event triggering conditions of the interactive features; Analyzing the spatiotemporal coupling strength among the time-sensitive features, the spatial correlation features, and the event-responsive features to identify scene cross-combination patterns; Constructing the multi-dimensional dynamic scene feature set according to the scene cross-combination pattern; The cross-combination mode includes a time-space superposition scenario, a time-event triggering scenario, a space-event linkage scenario, and a multi-dimensional linkage scenario.

3. The method according to claim 2, characterized in that The extracting of time-sensitive features according to the temporal continuity of the interactive features includes: Based on the user's stay action information and the historical search records, the user's action time segments are divided, and the action frequency and duration in different time periods are marked; Based on the real-time traffic status data and the meteorological status data in the same time segment in the environmental perception data, the action frequency and the duration are analyzed, the influence intensity of the internal and external environment on the user action in the current time segment is identified, and a time-sensitive feature vector is generated as the time-sensitive feature.

4. The method according to claim 3, characterized in that The extracting of spatial correlation features according to the spatial topological relationship of the interactive features includes: Based on the user's stay action information, a dynamic spatial grid is constructed with the user's stay location as the center, and the grid is divided into a core area and a peripheral radiation area. Analyzing the local environment perception synchronization data, extracting the location distribution and operation status of public facilities, combining the real-time traffic status data, identifying the spatial topological relationship between the public facilities and the dynamic spatial grid, and generating a facility accessibility map; According to the geographic location keywords involved in the historical search records, high-frequency interest spatial nodes are extracted, the high-frequency interest spatial nodes are marked in the facility accessibility map, and a heat distribution map of interest points is generated as the spatial correlation feature.

5. The method according to claim 4, characterized in that The extracting event response features according to the event triggering conditions of the interaction features includes: Analyze the public event data to determine whether the corresponding public event is a strong push demand event; If the public event is a strong push demand event, analyzing the spatial overlap between the current user movement trajectory information and the predicted movement trajectory and the impact range corresponding to the public event; when the spatial overlap is greater than a preset overlap threshold, determining the current public event data as the event response feature; If the public event does not belong to the strong push demand event, the user's event interest in the current public event is analyzed based on the historical search records. When the event interest is greater than the preset interest threshold, the current public event data is determined as the event response feature.

6. The method according to claim 5, characterized in that The analyzing the multi-dimensional dynamic scene feature set to determine the current user's complex scene demand information includes: Based on the dynamic priority adjustment strategy, according to the scene cross-combination pattern represented in the multi-dimensional dynamic scene feature set, the primary and secondary driving dimensions of the current user's scene are identified, and a scene influence weight set is constructed; According to the scenario impact weight set, in combination with the multidimensional dynamic scenario feature set, multimodal feature fusion is performed on the time-sensitive feature vector, the spatial correlation feature, and the event response feature to generate a composite demand feature vector as the composite scenario demand information; The dynamic priority adjustment strategy includes: when the time-space superposition scenario is included in the scenario cross-combination pattern, dynamically allocating scenario priorities based on the urgency of the user's time segment and the accessibility of core area facilities; When the event response feature corresponding to the strong push demand event exists in the scenario cross-combination mode, the event response feature is used as the dominant dimension, and the influence weight of the event response feature is increased according to the urgency of the strong push demand event; When the public event corresponding to the event response feature in the scene cross-combination mode does not belong to the strong push demand event, the influence weight corresponding to the event response feature is adjusted according to the current event interest.

7. An artificial intelligence-based platform user interest recommendation system, characterized in that: include: An action perception module is used to obtain user action data and environmental perception data, analyze the user action data and environmental perception data, and construct a unified spatiotemporal semantic network; A scene analysis module, configured to analyze the characteristics of the scene in which the user is located based on the unified spatiotemporal semantic network and determine a multi-dimensional dynamic scene feature set; Push module for analyzing the multi-dimensional dynamic scene feature set, determining the current user's composite scene demand information, and retrieving and pushing the user's real-time demand information based on the composite scene demand information; The user action data includes user stay action information and historical search records; The environmental perception data includes real-time traffic status data, meteorological status data, public facility operation status data and public event data; The action perception module is specifically used to: Analyzing the environmental perception data based on the user's stay action information, and screening and determining the local environmental perception data corresponding to the area where the user is located; Performing timestamp alignment processing on the real-time traffic status data, the meteorological status data, the public facility operation status data, and the public event data in the local environment perception data to determine the local environment perception synchronization data under the same time axis; A graph neural network algorithm based on the fusion attention mechanism uses the user action data as the attention target node and the data corresponding to the local environment perception synchronization data as the neighbor nodes to construct a basic association graph; Determine an attention graph neural network based on the basic association graph, using the real-time relative relationship between the entity corresponding to the neighbor node and the entity corresponding to the attention target node as the attention weight between the neighbor node and the corresponding attention target node; According to the dynamic attention update strategy, the attention graph neural network is dynamically updated to construct the unified spatiotemporal semantic network in real time; When dynamically updating the attention graph neural network, the action perception module is specifically used to: Extracting, based on the environmental perception data, an influence range of a physical perception device corresponding to each item of data in the environmental perception data; Drawing user movement trajectory information based on the user's stay action information; The dynamic attention update strategy includes a passive update mechanism and an active update mechanism; The passive update mechanism: when the user movement trajectory information changes, the neighbor nodes corresponding to the current attention target node in the attention graph neural network are dynamically updated according to the relative position relationship between the influence range and the change in the user's stop action; The active update mechanism: based on the long-short-term memory collaborative algorithm, predicts the user's subsequent movement trajectory according to the user movement trajectory information and the historical search records, and preloads the subsequent environmental perception data nodes as pre-update nodes corresponding to several of the neighbor nodes according to the predicted movement trajectory.

Images