Urban public security risk chain identification method and system based on graph neural network

Abstract

The invention provides an urban public security risk chain identification method and system based on a graph neural network, and relates to the technical field of public security risk management.The method comprises the steps that an event association topological graph is constructed, and an optimal sampling path is determined by adopting a self-adaptive two-way jump sampling algorithm; identifying logic breaking points, searching potential transition events to construct a complete risk chain, and finally calculating credibility scores based on conduction strength characteristics to screen the risk chain. According to the invention, the implicit association in the risk chain can be automatically found, the risk early warning precision is improved, and a scientific basis is provided for urban public safety management decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of public safety risk management, and in particular to a method and system for identifying urban public safety risk chains based on graph neural networks. Background Art

[0002] With the acceleration of urbanization and the increase in urban population density, urban public safety risk incidents are becoming more frequent, posing serious threats to social management and the safety of people's lives and property. Urban public safety incidents often do not occur in isolation, but rather are interconnected and transmitted layer by layer in the form of risk chains, forming a complex risk network. A risk chain refers to a series of interconnected risk events, where the occurrence of a previous event triggers or exacerbates the risk of subsequent events. Accurately identifying these risk chains is crucial for preventing and mitigating the harm of public safety incidents.

[0003] Traditional methods for identifying urban public safety risks rely primarily on expert experience and statistical analysis, employing methods such as event frequency statistics, correlation analysis, and case studies to conduct risk assessments. In recent years, with the development of data science and artificial intelligence technologies, some studies have begun to explore the use of data mining and machine learning to analyze public safety incident data in order to uncover correlation patterns and risk transmission pathways.

[0004] However, the existing urban public safety risk chain identification methods have the following defects and shortcomings: the existing methods lack the ability to comprehensively model the complex correlations between risk events, and it is difficult to effectively capture the multi-dimensional correlation characteristics and nonlinear transmission relationships between events, resulting in the identified risk chains being inaccurate and incomplete; the existing risk chain identification methods are not effective when facing data sparsity problems, and it is difficult to identify those hidden risk chains that are not obvious in historical data but have higher actual risks, especially for those risk paths that exist logically but are insufficiently represented in the data, and cannot be effectively supplemented and inferred; the existing methods lack a scientific evaluation system for the credibility of the risk chain, and it is difficult to quantify the reliability of the identification results, and cannot provide a reliable basis for risk priority ranking for urban public safety management decisions, which limits the accuracy and effectiveness of risk prevention and control measures. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for identifying urban public safety risk chains based on graph neural networks, which can solve the problems in the existing technology.

[0006] A first aspect of an embodiment of the present invention provides a method for identifying urban public safety risk chains based on a graph neural network, comprising: Obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and the relationships between public safety events as edges to construct an initial event correlation topology graph; Based on the initial event correlation topology, an adaptive bidirectional jump sampling algorithm is used to determine the optimal sampling path set through dynamic step size control and timing matching; multi-dimensional correlation strength is calculated based on the event sequence in the optimal sampling path set, and an initial candidate risk chain set is screened; Based on the initial candidate risk chain set, a state feature matrix is ​​constructed through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; Based on the logically complete risk chain, conduction coefficient matrix analysis and state transition matrix decomposition are used to extract conduction strength characteristics, calculate credibility scores, and obtain a risk chain credibility score set; Based on the risk chain credibility score set, the risk chain with a credibility score exceeding a preset credibility threshold is selected, and the final identification result is determined and output.

[0007] In an optional embodiment, based on the initial event correlation topology graph, an adaptive bidirectional jump sampling algorithm is adopted to determine the optimal sampling path set through dynamic step size control and timing matching, including: Obtain the influence calculation parameters of the public safety event node from the initial event correlation topology graph, calculate the propagation influence based on the node correlation and distance attenuation factor, calculate the spatial influence based on the ratio of the public safety event node to the affected area, and calculate the temporal influence based on the event time attenuation function. A weighted combination of the propagation influence, spatial influence, and temporal influence is used to obtain the node influence metric. The node influence metric and the baseline sampling step are used to calculate the initial adaptive step size. The adaptive control threshold is obtained based on the historical difference of the node influence metric. The dynamic sampling step size is obtained by combining the initial adaptive step size and the adaptive control threshold. Bidirectional sampling is performed using a dynamic sampling step size to calculate the forward propagation probability and backward verification probability respectively, and then weighted fusion is used to obtain the time series matching score; The path continuity index and length constraint index are calculated based on the timing matching score. The timing matching score is nonlinearly weighted with the path continuity index and length constraint index to obtain the path evaluation score. The public safety event sequences with path evaluation scores greater than the preset evaluation threshold are screened to obtain the optimal sampling path set.

[0008] In an optional embodiment, the multi-dimensional correlation strength is calculated based on the event sequence in the optimal sampling path set, and the initial candidate risk chain set obtained by screening includes: Acquire event sequences from the optimal sampling path set to construct a temporal evolution graph, extract the time intervals between adjacent events and the event-intensive periods, identify key time nodes, calculate the time delay characteristics of event propagation by comparing the moments of adjacent events, extract the periodicity of event occurrence through Fourier transform, and calculate the temporal correlation strength of adjacent events based on the time delay characteristics and periodicity. Based on the geographic coordinate information of each event in the event sequence, a spatial distribution map is constructed, the spatial distance between adjacent events is calculated, and the spatial correlation strength is obtained by combining the spatial distance with the spatial impact range of the event; The event semantic vector is constructed using the description text of each event in the event sequence. The cosine similarity of the semantic vectors of adjacent events is calculated and combined with the event type matching matrix to obtain the semantic association strength. Based on the influence measurement value of each event in the event sequence, the influence correlation degree of adjacent events is calculated, and the influence correlation degree is combined with the influence transfer attenuation function to obtain the influence correlation strength; The multi-dimensional correlation strength is obtained by weighted fusion of temporal correlation strength, spatial correlation strength, semantic correlation strength and influence correlation strength; The event sequences whose multi-dimensional correlation strength is greater than a preset strength threshold are screened to obtain an initial candidate risk chain set.

[0009] In an optional embodiment, based on the initial candidate risk chain set, constructing a state feature matrix through eigenvalue decomposition to identify logical breakpoints, searching for potential transition events, and iteratively optimizing to obtain a logically complete risk chain includes: Constructing the multi-dimensional correlation strength of the event sequence in the initial candidate risk chain set into a state feature matrix, wherein each element of the state feature matrix corresponds to the correlation state between adjacent events; Performing eigenvalue decomposition on the state feature matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs whose fracture probability values ​​are greater than an adaptive fracture threshold as logical fracture points; For the events before and after the logical breakpoint, a valid time window is constructed based on the corresponding timestamp, and a reachable range ellipse is constructed based on the corresponding geographical location; Search for potential transition events within the valid time window and reachable range ellipse, calculate the time dimension contribution, spatial dimension contribution, and semantic dimension contribution of the potential transition event to the previous and next events, and perform weighted combination to obtain the event transition evaluation value; The potential transition event with the highest event transition evaluation value is used as a supplementary event and inserted between the preceding and following events corresponding to the logical breakpoint to generate an optimized event sequence; The state feature matrix of the optimized event sequence is reconstructed and the process is repeated until there are no adjacent event pairs marked as logical breakpoints in the optimized event sequence, thereby obtaining a logically complete risk chain.

[0010] In an optional embodiment, performing eigenvalue decomposition on the state feature matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs having a fracture probability value greater than an adaptive fracture threshold as logical fracture points includes: Performing eigenvalue decomposition on the state characteristic matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, and constructing them into a third-order tensor; performing multi-scale tensor decomposition on the third-order tensor to obtain an eigenvalue component matrix, an eigenvector component matrix, and a time scale component matrix at different time scales; constructing a multi-scale characteristic spectrum of the event sequence based on the eigenvalue component matrix, the eigenvector component matrix and the time scale component matrix, and mapping the energy distribution of the multi-scale characteristic spectrum at different time scales into a characteristic entropy matrix; Calculate the characteristic state difference of adjacent event pairs according to the eigenvalue diagonal matrix and the eigenvector matrix, and calculate the entropy difference vector of the adjacent event pairs according to the characteristic entropy matrix; extracting the local fluctuation features of the characteristic state difference and the cross-scale correlation features of the entropy difference vector respectively, fusing the local fluctuation features with the cross-scale correlation features to obtain a comprehensive fracture feature, and calculating the fracture probability value of the adjacent event pair based on the comprehensive fracture feature; An adaptive fracture threshold is constructed based on the variance of the characteristic state difference and the standard deviation of the entropy difference vector of the adjacent event pairs, and the adjacent event pairs whose fracture probability values ​​are greater than the adaptive fracture threshold are marked as logical fracture points.

[0011] In an optional embodiment, based on the logically complete risk chain, the conduction coefficient matrix analysis and state transition matrix decomposition are used to extract the conduction strength characteristics, calculate the credibility score, and obtain the risk chain credibility score set including: For a logically complete risk chain, a transmission coefficient matrix is ​​constructed based on the fluctuation amplitude, propagation distance, and impact range of adjacent events. The risk transmission intensity sequence is calculated through the temporal evolution of the transmission coefficient matrix. Constructing a state transfer matrix for the risk conduction intensity sequence, performing eigenvalue decomposition on the state transfer matrix, calculating and determining eigenvalues, determining events for which the product of the eigenvalue and the impact range exceeds a preset conduction threshold as diffusion conduction nodes, and calculating the energy diffusion index of the diffusion conduction nodes; Dividing the risk transmission intensity sequence into multiple time scales, calculating the fluctuation characteristic sequence at each time scale, performing segmented iterative reconstruction on the fluctuation characteristic sequence to obtain a fractal sequence, and calculating the complexity index of risk transmission through the self-similarity of the fractal sequence; Based on the risk transmission intensity sequence and the fluctuation characteristic sequence, a three-dimensional phase space including transmission intensity, fluctuation amplitude and time delay is constructed, the phase space trajectory is extracted, and the Lyapunov exponent of the phase space trajectory at different time scales is calculated; A three-dimensional feature vector is constructed based on the energy diffusion index, complexity index and Lyapunov index. The conduction stability score of the logically complete risk chain is calculated using a pre-trained adaptive weight network. The conduction stability score is transformed into a probability distribution to obtain a risk chain credibility score set.

[0012] In an optional embodiment, performing segmented iterative reconstruction on the fluctuation characteristic sequence to obtain a fractal sequence includes: Obtaining a fluctuation characteristic sequence, processing the fluctuation characteristic sequence using an overlapping segmentation method, determining the number of overlapping points of adjacent segments by a preset segment length and overlapping rate, and generating a plurality of overlapping sequence segments; Extracting fluctuation characteristics of each of the overlapping sequence segments, the fluctuation characteristics including fluctuation amplitude and fluctuation trend, and pairing adjacent overlapping sequence segments in pairs according to the fluctuation characteristics; For each pair of adjacent overlapping sequence segments, a segmented mapping rule is constructed based on the fluctuation characteristics. The fluctuation amplitude ratio of the adjacent overlapping sequence segments is used as the amplitude ratio parameter, and the fluctuation trend difference is used as the trend difference parameter to generate a piecewise linear mapping function. Iteratively reconstruct the overlapping sequence segments. In each iteration, the amplitude ratio parameter and trend difference parameter are dynamically adjusted based on the characteristic change of the current overlapping sequence segment, and act on the next iterative reconstruction to generate a reconstructed sequence segment. The reconstructed sequence segments are sequentially spliced ​​and smoothed to obtain a fractal sequence.

[0013] A second aspect of an embodiment of the present invention provides an urban public safety risk chain identification system based on a graph neural network, comprising: The first unit is used to obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and use the associations between public safety events as edges to construct an initial event association topology graph; The second unit is configured to determine an optimal sampling path set based on the initial event correlation topology graph using an adaptive bidirectional jump sampling algorithm through dynamic step size control and timing matching; calculate multidimensional correlation strength based on the event sequence in the optimal sampling path set, and screen out an initial candidate risk chain set; The third unit is configured to construct a state characteristic matrix based on the initial candidate risk chain set through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; The fourth unit is used to extract the conduction strength characteristics based on the logically complete risk chain by using conduction coefficient matrix analysis and state transition matrix decomposition, calculate the credibility score, and obtain a risk chain credibility score set; The fifth unit is configured to select, based on the risk chain credibility score set, a risk chain whose credibility score exceeds a preset credibility threshold, and determine and output a final identification result.

[0014] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0015] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0016] In an embodiment of the present invention, by constructing an initial event association topology map and adopting an adaptive bidirectional jump sampling algorithm to determine the optimal sampling path, the complex association patterns between urban public safety events can be effectively captured, the path search blindness and waste of computing resources in traditional methods can be avoided, and the accuracy and efficiency of risk chain identification can be improved; the method of constructing a state feature matrix based on eigenvalue decomposition to identify logical breakpoints and search for potential transition events solves the logical incoherence problem in the risk chain, making the identified risk chain more complete and reasonable, and avoiding the common risk chain breakage and logical loss problems in traditional methods; the mechanism of extracting conduction intensity characteristics using conduction coefficient matrix analysis and state transition matrix decomposition and calculating credibility scores provides a quantitative basis for risk chain evaluation, effectively reduces the uncertainty in the risk chain identification process, improves the reliability and practical value of the identification results, and provides more accurate decision-making support for urban public safety management and risk prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for identifying urban public safety risk chains based on a graph neural network according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the risk chain logic completeness optimization system architecture based on eigenvalue decomposition; Figure 3 Schematic diagram of the performance comparison between multi-scale feature analysis and traditional methods. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0019] The following specific embodiments are used to describe the technical solution of the present invention in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 This is a flow chart of a method for identifying urban public safety risk chains based on graph neural networks according to an embodiment of the present invention. Figure 1 As shown, the method includes: Obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and the relationships between public safety events as edges to construct an initial event correlation topology graph; Based on the initial event correlation topology, an adaptive bidirectional jump sampling algorithm is used to determine the optimal sampling path set through dynamic step size control and timing matching; multi-dimensional correlation strength is calculated based on the event sequence in the optimal sampling path set, and an initial candidate risk chain set is screened; Based on the initial candidate risk chain set, a state feature matrix is ​​constructed through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; Based on the logically complete risk chain, conduction coefficient matrix analysis and state transition matrix decomposition are used to extract conduction strength characteristics, calculate credibility scores, and obtain a risk chain credibility score set; Based on the risk chain credibility score set, the risk chain with a credibility score exceeding a preset credibility threshold is selected, and the final identification result is determined and output.

[0021] In an optional embodiment, based on the initial event correlation topology graph, an adaptive bidirectional jump sampling algorithm is used to determine the optimal sampling path set through dynamic step size control and timing matching, including: Obtain the influence calculation parameters of the public safety event node from the initial event correlation topology graph, calculate the propagation influence based on the node correlation and distance attenuation factor, calculate the spatial influence based on the ratio of the public safety event node to the affected area, and calculate the temporal influence based on the event time attenuation function. A weighted combination of the propagation influence, spatial influence, and temporal influence is used to obtain the node influence metric. The node influence metric and the baseline sampling step are used to calculate the initial adaptive step size. The adaptive control threshold is obtained based on the historical difference of the node influence metric. The dynamic sampling step size is obtained by combining the initial adaptive step size and the adaptive control threshold. Bidirectional sampling is performed using a dynamic sampling step size to calculate the forward propagation probability and backward verification probability respectively, and then weighted fusion is used to obtain the time series matching score; The path continuity index and length constraint index are calculated based on the timing matching score. The timing matching score is nonlinearly weighted with the path continuity index and length constraint index to obtain the path evaluation score. The public safety event sequences with path evaluation scores greater than the preset evaluation threshold are screened to obtain the optimal sampling path set.

[0022] In one specific embodiment, when calculating the impact of public safety event nodes, event node features are extracted from the initial event correlation topology, including inter-node associations, spatial distribution information, and temporal attributes. To calculate node association, the number of direct connections between each node and other nodes is counted, and the connection strength is analyzed. For example, a node associated with an explosion event is connected to 12 subsequent events, with an association strength of 0.85. The distance decay factor uses an exponentially decreasing model. When the physical distance between two event nodes is 5 kilometers, the distance decay value is approximately 0.72. The propagation impact is calculated by multiplying the association by the decay factor. In the above example, the propagation impact is 0.61.

[0023] Spatial impact is calculated based on the ratio of the incident's impact area to the reference area. For example, if the impact area of ​​a chemical leak is 15 square kilometers and the reference area is 60 square kilometers, the impact ratio is 0.25. After factoring in a regional importance weighting factor of 1.2, the final spatial impact value is 0.30.

[0024] The temporal impact is calculated using a time decay function, employing a half-life model to assess the impact of an event over time. For example, within 24 hours after an event, the impact remains above 0.9, then drops to 0.65 after 48 hours and 0.32 after 96 hours. The parameters of the time decay function are adjusted based on the characteristics of different event types. For example, the decay rate for sudden incidents is 0.08, and for cumulative incidents it is 0.04.

[0025] The node influence metric combines the transmission influence, spatial influence, and temporal influence, with weights set to 0.4, 0.35, and 0.25, respectively. For example, if the three indicators for a public health event node are 0.65, 0.42, and 0.78, the weighted calculation yields a node influence metric of 0.605.

[0026] The initial adaptive step size is calculated by multiplying the node influence metric by the baseline sampling step size. The baseline sampling step size is set to 2. When the node influence metric is 0.605, the initial adaptive step size is 1.21. The adaptive control threshold is determined based on the historical difference in node influence. The influence metric changes of the last 10 sampling points are recorded. When the average change rate of three consecutive points exceeds 15%, the adaptive control threshold is adjusted to 0.18; when the change rate is less than 5%, the threshold is lowered to 0.08. The dynamic sampling step size is adjusted by comparing the initial adaptive step size with the control threshold. When the initial step size is less than 1.5 times the threshold, the dynamic sampling step size is the initial step size. Otherwise, the dynamic sampling step size is the weighted sum of 80% of the initial step size and 20% of the threshold.

[0027] During bidirectional sampling, forward and backward sampling are performed using a dynamic sampling step size. Forward sampling starts at the current node and traverses subsequent nodes in chronological order, calculating the forward propagation probability. In a real-world case, the propagation probabilities of the three possible subsequent events of a node are 0.72, 0.45, and 0.31, respectively. Paths with probabilities greater than 0.5 are preferentially selected. Backward validation sampling performs reverse validation from the resulting node to assess the plausibility of the event sequence. The calculation of the backward validation probability takes into account the strength of the causal relationship between events, typically ranging from 0.2 to 0.85. The forward propagation probability and the backward validation probability are weighted together to produce a time series matching score. The weights are assigned based on the reliability of the sampling direction, typically 0.65 for the forward direction and 0.35 for the backward direction.

[0028] During the path assessment phase, the path continuity index was calculated, reflecting the temporal and logical coherence of adjacent events along the path. For a 10-node event sequence, the time intervals between eight pairs of adjacent nodes were found to conform to the expected pattern, one pair had a slight time difference, and one pair had a significant logical jump. The calculated continuity index value was 0.75. The path length constraint ensures that the sampling path is neither too long nor too short. An ideal path length range is set based on the event type. For example, for a complete evolutionary path of a certain type of emergency, the ideal number of nodes is 8 to 15. When the actual path length is 11 nodes, the length constraint index value is 0.92.

[0029] The final path evaluation score is calculated through a nonlinear weighted calculation of the temporal matching score, path continuity index, and length constraint index. Using an exponential weighting method, the matching score is weighted at 0.5, the continuity index is weighted at 0.35, and the length constraint index is weighted at 0.15. For example, if the three indexes for a path are 0.68, 0.75, and 0.92, respectively, the path evaluation score obtained after nonlinear weighting is 0.724. A preset evaluation threshold of 0.65 is used to screen event sequences with evaluation scores greater than the threshold, ultimately generating an optimal sample path set. In one emergency response simulation, 18 optimal paths were selected from 126 candidate paths, providing diverse response options for decision-making.

[0030] In this embodiment, the three types of influence, namely propagation, space and time, are integrated to comprehensively characterize the importance of public safety event nodes and enhance the accuracy of path evaluation. The sampling step is dynamically adjusted based on the node influence to achieve more accurate two-way sampling and improve the effect of time series matching. A nonlinear weighted evaluation is performed in combination with the time series matching score, path continuity and length constraints to enhance the scientificity and robustness of the optimal event sequence screening. This method is applicable to multi-source heterogeneous and complex evolving public safety event topology structures and improves the efficiency and practicality of event sequence analysis.

[0031] In an optional embodiment, the multi-dimensional correlation strength is calculated based on the event sequence in the optimal sampling path set, and the initial candidate risk chain set obtained by screening includes: Acquire event sequences from the optimal sampling path set to construct a temporal evolution graph, extract the time intervals between adjacent events and the event-intensive periods, identify key time nodes, calculate the time delay characteristics of event propagation by comparing the moments of adjacent events, extract the periodicity of event occurrence through Fourier transform, and calculate the temporal correlation strength of adjacent events based on the time delay characteristics and periodicity. Based on the geographic coordinate information of each event in the event sequence, a spatial distribution map is constructed, the spatial distance between adjacent events is calculated, and the spatial correlation strength is obtained by combining the spatial distance with the spatial impact range of the event; The event semantic vector is constructed using the description text of each event in the event sequence. The cosine similarity of the semantic vectors of adjacent events is calculated and combined with the event type matching matrix to obtain the semantic association strength. Based on the influence measurement value of each event in the event sequence, the influence correlation degree of adjacent events is calculated, and the influence correlation degree is combined with the influence transfer attenuation function to obtain the influence correlation strength; The multi-dimensional correlation strength is obtained by weighted fusion of temporal correlation strength, spatial correlation strength, semantic correlation strength and influence correlation strength; The event sequences whose multi-dimensional correlation strength is greater than a preset strength threshold are screened to obtain an initial candidate risk chain set.

[0032] In one specific embodiment, after acquiring an event sequence from the optimal sampling path set, a time series evolution graph is constructed to analyze the temporal correlation between events. Specifically, for an event sequence {E1, E2, ..., En}, the occurrence time of each event {T1, T2, ..., Tn} is recorded. By calculating the time interval ΔTi = Ti+1 - Ti between adjacent events, periods of high event density can be identified. For example, when multiple consecutive ΔTi values ​​are less than a preset threshold (e.g., 24 hours), this period is marked as a high event density period. The system records the start and end time points of these high event density periods as key time nodes. The time delay characteristics of event propagation are calculated by analyzing the occurrence times of adjacent events. For example, if an earthquake event E1 occurs in a certain area, and aftershock events E2 and E3 occur in the surrounding area within 3 hours, there is a clear time delay relationship between E1 and E2 and E3. A Fourier transform is applied to the event sequence to extract periodic patterns. In a specific implementation, a discrete Fourier transform is used to analyze the frequency of event occurrence. For example, the frequency of mass incidents in a certain area increases every Friday afternoon. The temporal correlation strength (ST) is calculated based on the time delay characteristics and periodicity. When two events occur during the same peak period and the time delay conforms to historical statistical patterns, the ST value is high, such as ST = 0.85. If the time interval is long and there is no obvious periodic correlation, the ST value is low, such as ST = 0.2.

[0033] Spatial correlation analysis of event sequences is performed based on the geographic coordinate information of the events. For each event in the event sequence, the latitude and longitude coordinates (xi, yi) of its occurrence location are recorded. A spatial distribution map is constructed to visually display the distribution of events in geographic space. The spatial distance Dij between adjacent events is calculated using a spherical distance calculation method that takes into account the curvature of the earth. Different spatial impact ranges Ri are defined for different types of events. For example, the spatial impact range of a natural disaster event may be 50 kilometers, while the spatial impact range of a social security event may be 10 kilometers. The spatial correlation strength SS is calculated as the ratio of spatial distance to spatial impact range. When Dij is less than Ri, it indicates a high degree of spatial correlation, and the SS value is close to 1; when Dij is much greater than Ri, the SS value is close to 0. For example, if the impact range of an explosion event is 2 kilometers, if the related secondary disaster event occurs within 1.5 kilometers, the calculated SS = 0.75.

[0034] Semantic association analysis is achieved by processing event description text. Each event description undergoes preprocessing, including word segmentation, stop word removal, and stemming. Word embedding technology is used to convert the processed text into a semantic vector Vi. The cosine similarity of the semantic vectors of adjacent events is calculated to determine the degree of semantic similarity. The system also predefines an event type matching matrix M, which represents the degree of intrinsic correlation between different event types. For example, if two event types are "chemical plant explosion" and "air pollution," the value of M might be 0.8, indicating a high correlation between the two types of events. The final semantic association strength (SL) is derived by a weighted combination of the cosine similarity and the matching matrix value. In this example, the cosine similarity of the semantic vectors of the two event descriptions is 0.6, and the event type matching matrix value is 0.8, resulting in a calculated semantic association strength (SL) of 0.72.

[0035] Influence correlation analysis is conducted based on the influence metrics of events. An influence index, I, is calculated for each event, taking into account factors such as the number of people affected, economic losses, and social attention. For adjacent events, an influence correlation, C, is calculated, reflecting the ratio of the influence of the subsequent event to the influence of the preceding event. An influence transfer attenuation function, D, is defined to account for the attenuation effects of temporal and spatial distances between events on influence transfer. The influence correlation strength, SI, is calculated by combining the influence correlation and the influence transfer attenuation function. For example, if the influence of the preceding event is 85 (out of 100), the influence of the subsequent event is 60, the time interval is 2 days, and the spatial distance is 15 kilometers, the calculated influence correlation C = 0.7, the transfer attenuation D = 0.8, and the final SI = 0.56.

[0036] Multidimensional correlation strength is calculated by weighted fusion of the four aforementioned correlation strengths. Weight coefficients w1, w2, w3, and w4 are set based on different application scenarios, corresponding to temporal, spatial, semantic, and influence correlation strengths, respectively. The weighted calculation yields the multidimensional correlation strength S = w1 × ST + w2 × SS + w3 × SL + w4 × SI. In practice, the weights can be dynamically adjusted based on the needs of different risk chain identification. For example, for natural disaster risk chains, w1 = 0.3, w2 = 0.4, w3 = 0.1, and w4 = 0.2 might be set; for social security risk chains, w1 = 0.2, w2 = 0.2, w3 = 0.4, and w4 = 0.2 might be set. For the calculated multidimensional correlation strength values, the system sets a preset strength threshold θ (e.g., θ = 0.6) to filter event sequences with S greater than θ to form the initial set of candidate risk chains. In one instance, the correlation strengths of four pairs of adjacent events formed by five consecutive events were 0.75, 0.82, 0.58, and 0.63, respectively. The third pair was below the threshold, so the system divided the event sequence into two candidate risk chains.

[0037] In this embodiment, the correlation strength of the four dimensions of time, space, semantics and influence is integrated to achieve in-depth and multi-angle correlation analysis of event sequences; based on the screening of high-correlation event sequences based on multi-dimensional correlation strength, potential candidate risk chains can be accurately extracted, providing a solid foundation for subsequent risk assessment; time delay characteristics and periodic analysis are introduced to effectively identify the rhythm and laws of event propagation, and enhance the understanding of the dynamic evolution trend of events; the fusion of multi-source heterogeneous information makes the model more adaptable to the complex and changing public safety environment, and improves the stability and reliability of the analysis results.

[0038] In an optional embodiment, based on the initial candidate risk chain set, constructing a state feature matrix through eigenvalue decomposition to identify logical breakpoints, searching for potential transition events, and iteratively optimizing to obtain a logically complete risk chain includes: Constructing the multi-dimensional correlation strength of the event sequence in the initial candidate risk chain set into a state feature matrix, wherein each element of the state feature matrix corresponds to the correlation state between adjacent events; Performing eigenvalue decomposition on the state feature matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs whose fracture probability values ​​are greater than an adaptive fracture threshold as logical fracture points; For the events before and after the logical breakpoint, a valid time window is constructed based on the corresponding timestamp, and a reachable range ellipse is constructed based on the corresponding geographical location; Search for potential transition events within the valid time window and reachable range ellipse, calculate the time dimension contribution, spatial dimension contribution, and semantic dimension contribution of the potential transition event to the previous and next events, and perform weighted combination to obtain the event transition evaluation value; The potential transition event with the highest event transition evaluation value is used as a supplementary event and inserted between the preceding and following events corresponding to the logical breakpoint to generate an optimized event sequence; The state feature matrix of the optimized event sequence is reconstructed and the process is repeated until there are no adjacent event pairs marked as logical breakpoints in the optimized event sequence, thereby obtaining a logically complete risk chain.

[0039] Figure 2This is a schematic diagram of a risk chain logic completeness optimization system architecture based on eigenvalue decomposition. In one embodiment, data of an initial candidate risk chain set is received. This data includes multiple event sequences, each of which includes information such as a timestamp, geographic location, and event description. The initial candidate risk chain can be a set of event sequences extracted from a security monitoring system, such as a sequence consisting of events E1 (2023-05-01 08:30:00, location coordinates (115.32, 39.45), equipment startup abnormality) and E2 (2023-05-01 09:15:00, location coordinates (115.35, 39.46), temperature exceeding a limit).

[0040] The event sequences in the initial candidate risk chain set are analyzed to construct a state feature matrix. This matrix reflects the multidimensional correlation strength between adjacent events. Assuming there are N events in a risk chain, the size of the constructed state feature matrix is ​​(N-1)×D, where D is the number of correlation dimensions. For each pair of adjacent events (Ei, Ei+1), the temporal correlation strength RT(i, i+1), spatial correlation strength RS(i, i+1), and semantic correlation strength RL(i, i+1) are calculated. The temporal correlation strength is calculated by the ratio of the time interval between the two events to the expected time window; the spatial correlation strength is calculated by the ratio of the geographic distance between the two events to the expected reachable distance; and the semantic correlation strength is calculated by the similarity of the event description text. The correlation strength values ​​of these three dimensions constitute the element in the i-th row of the state feature matrix and range from 0 to 1, with larger values ​​indicating stronger correlation.

[0041] The constructed state characteristic matrix is ​​subjected to eigenvalue decomposition to obtain an eigenvalue diagonal matrix and an eigenvector matrix. Eigenvalue decomposition is achieved through iterative calculation. The eigenvalue diagonal matrix contains the sorted eigenvalues, and the eigenvector matrix contains the corresponding eigenvectors. Based on the decomposition results, the characteristic state difference of adjacent event pairs is calculated. For each pair of adjacent events (Ei, Ei+1), the characteristic state difference D(i, i+1) is calculated. The greater the difference, the worse the logical coherence between the two events. When calculating the difference, the eigenvector is used to project the event state into the feature space, and the Euclidean distance of the projected state vector is calculated. The characteristic state difference is normalized to obtain the fracture probability value P(i, i+1).

[0042] Set an adaptive break threshold θ, determined through historical data analysis or expert experience, for example, 0.75. Adjacent event pairs whose break probability P(i, i+1) exceeds the threshold θ are marked as logical break points. For example, if the break probability of the event pair (E2, E3) is 0.82, which is greater than the threshold 0.75, then (E2, E3) is marked as a logical break point.

[0043] For each marked logical breakpoint (Ei, Ei+1), a valid time window [Ti+α, Ti+1-β] is constructed based on its timestamp, where Ti and Ti+1 are the timestamps of events Ei and Ei+1, respectively, and α and β are time buffer parameters. For example, if the time of E2 is 09:15:00 and the time of E3 is 10:30:00, assuming α = 5 minutes and β = 5 minutes, the valid time window is [09:20:00, 10:25:00]. Simultaneously, a reachable range ellipse is constructed based on the geographic location. The two foci of the ellipse are the geographic locations of events Ei and Ei+1, respectively. The length of the semi-major axis of the ellipse is related to the distance between the two events and the possible movement speed. For the above example, if the location coordinates of E2 are (115.35, 39.46) and the location coordinates of E3 are (115.40, 39.49), the constructed ellipse contains these two points, and the major axis is along the line connecting the two points.

[0044] Within the specified valid time window and reachable range ellipse, potential transition events are searched from the event library. For each candidate potential transition event Ec, its correlation with the preceding and following events Ei and Ei+1 is calculated. The temporal contribution CT(Ec) is calculated by the uniform distribution of Ec's position on the time axis relative to Ei and Ei+1; the spatial contribution CS(Ec) is calculated by the spatial proximity of Ec to the path from Ei to Ei+1; and the semantic contribution CL(Ec) is calculated by the semantic continuity between Ec and Ei and Ei+1. The weighted combination of the three dimensional contributions yields the event transition evaluation value V(Ec) = w1 × CT(Ec) + w2 × CS(Ec) + w3 × CL(Ec), where the weights w1, w2, and w3 sum to 1 and can be adjusted based on the specific application scenario.

[0045] From the set of potential transition events, the event with the highest transition evaluation value V(Ec) is selected as the supplementary event E*. This event is inserted between the preceding and following events corresponding to the logical breakpoint to generate the optimized event sequence. For example, if the optimal transition event found is E* (2023-05-01 09:45:00, position coordinates (115.37, 39.47), pressure fluctuation), it is inserted between E2 and E3 to obtain the optimized sequence [..., E2, E*, E3, ...].

[0046] The state feature matrix is ​​reconstructed for the optimized event sequence, and the eigenvalue decomposition, breakpoint identification, transition event search, and sequence optimization steps are repeated until no corresponding event pairs are marked as logical breakpoints, resulting in a logically complete risk chain. In practice, a maximum number of iterations (e.g., 10) or a minimum improvement threshold can be set as a termination condition to avoid over-optimization.

[0047] In a case study of an industrial production safety monitoring system implementation, the initial candidate risk chain included five events: E1 (08:30, equipment startup anomaly), E2 (09:15, temperature exceeding the limit), E3 (10:30, pressure anomaly), E4 (11:20, valve failure), and E5 (12:00, system shutdown). After the aforementioned process, (E2, E3) was identified as a logical breakpoint, and transition event E* (09:45, pressure fluctuation) was added. In a second iteration, (E3, E4) was identified as a logical breakpoint, and event E** (10:50, pipeline vibration) was added. The resulting logically complete risk chain: E1 → E2 → E* → E3 → E** → E4 → E5, making the risk event sequence more logically coherent and complete.

[0048] In this embodiment, by detecting logical breakpoints and inserting transition events, the problems of missing information or interrupted associations are effectively repaired, and a coherent and continuous event sequence is constructed; the eigenvalue decomposition of the state feature matrix is ​​used to accurately quantify the state differences between adjacent events, thereby improving the accuracy of logical breakpoint identification; the transition events are evaluated by combining the multi-dimensional contribution of time, space, and semantics to ensure that the supplementary nodes are highly consistent with the previous and subsequent events, thereby improving the rationality and credibility of the risk chain; a cyclic iterative optimization mechanism is adopted to continuously improve the event sequence structure, and ultimately a high-quality risk chain with a logical closed loop is obtained, providing a more reliable basis for risk prediction and prevention and control.

[0049] In an optional embodiment, performing eigenvalue decomposition on the state feature matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs whose fracture probability values ​​are greater than an adaptive fracture threshold as logical fracture points includes: Performing eigenvalue decomposition on the state characteristic matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, and constructing them into a third-order tensor; performing multi-scale tensor decomposition on the third-order tensor to obtain an eigenvalue component matrix, an eigenvector component matrix, and a time scale component matrix at different time scales; constructing a multi-scale characteristic spectrum of the event sequence based on the eigenvalue component matrix, the eigenvector component matrix and the time scale component matrix, and mapping the energy distribution of the multi-scale characteristic spectrum at different time scales into a characteristic entropy matrix; Calculate the characteristic state difference of adjacent event pairs according to the eigenvalue diagonal matrix and the eigenvector matrix, and calculate the entropy difference vector of the adjacent event pairs according to the characteristic entropy matrix; extracting the local fluctuation features of the characteristic state difference and the cross-scale correlation features of the entropy difference vector respectively, fusing the local fluctuation features with the cross-scale correlation features to obtain a comprehensive fracture feature, and calculating the fracture probability value of the adjacent event pair based on the comprehensive fracture feature; An adaptive fracture threshold is constructed based on the variance of the characteristic state difference and the standard deviation of the entropy difference vector of the adjacent event pairs, and the adjacent event pairs whose fracture probability values ​​are greater than the adaptive fracture threshold are marked as logical fracture points.

[0050] In one specific embodiment, eigenvalue decomposition is performed on the state feature matrix to obtain a diagonal eigenvalue matrix and an eigenvector matrix. Assume that the state feature matrix is ​​an n×m matrix, where n is the number of events and m is the feature dimension. Eigenvalue decomposition yields an m×m eigenvector matrix U and a diagonal matrix Λ containing m eigenvalues. The eigenvalue diagonal matrix Λ and the eigenvector matrix U are constructed as a third-order tensor T with dimensions m×m×1.

[0051] Multiscale tensor decomposition is performed on the third-order tensor T, using the Tucker decomposition method to decompose it into component matrices at different time scales. This decomposition yields an eigenvalue component matrix P (of size m × r1), an eigenvector component matrix Q (of size m × r2), and a timescale component matrix R (of size 1 × r3), where r1, r2, and r3 are the ranks of the respective dimensions. For example, when m = 100, r1 = 20, r2 = 30, and r3 = 5 can be chosen to effectively capture data characteristics at all time scales.

[0052] A multiscale characteristic spectrum S of an event sequence is constructed based on the component matrices P, Q, and R. These three component matrices are combined through a tensor product operation to form a multiscale characteristic spectrum of size m×m×r³. For each time scale k (1≤k≤r³), the energy distribution of the characteristic spectrum at that scale is calculated and mapped into a characteristic entropy matrix H. The characteristic entropy matrix H is of size n×r³ and represents the entropy distribution of n events at r³ time scales. For example, the characteristic entropy value of an event i at time scale k can be obtained by calculating the Shannon entropy of the energy distribution of the characteristic spectrum at that scale. This value reflects the uncertainty of the event characteristics at that time scale.

[0053] The characteristic state difference D between adjacent event pairs is calculated based on the eigenvalue diagonal matrix Λ and the eigenvector matrix U. For an adjacent event pair (i, i+1), its characteristic state difference D(i, i+1) can be obtained by calculating the Euclidean distance between the two events in the characteristic space. Simultaneously, the entropy difference vector E of the adjacent event pair is calculated based on the characteristic entropy matrix H. For an adjacent event pair (i, i+1), its entropy difference vector E(i, i+1) is an r3-dimensional vector representing the entropy difference between the two events at various time scales.

[0054] Extract the local fluctuation feature W of the feature state difference D. Use a sliding window method with a window size of 3, slide it over the difference sequence, and calculate the fluctuation of the difference within the window. For example, for position i, extract D(i-1), D(i), and D(i+1). Calculate the mean difference between the center position and the positions on both sides to obtain the local fluctuation feature W(i). For the entropy difference vector E, extract its cross-scale correlation feature C. The specific method is to calculate the correlation of entropy differences between different time scales, forming a correlation matrix of size r3×r3, and extract the main correlation patterns from it as the cross-scale correlation feature C.

[0055] The local fluctuation feature W is fused with the cross-scale correlation feature C to obtain the comprehensive fracture feature F. This fusion method uses a weighted summation method, with weights determined through historical data training. For example, if the weight of the local fluctuation feature is set to 0.6 and the weight of the cross-scale correlation feature is set to 0.4, then F = 0.6 × W + 0.4 × C. The fracture probability P of each pair of adjacent events is calculated based on the comprehensive fracture feature F. A sigmoid function is used to map F to the interval [0, 1], resulting in the fracture probability for each pair of adjacent events.

[0056] An adaptive fracture threshold θ is constructed based on the variance of the characteristic state differences (var(D)) and the standard deviation of the entropy difference vector (std(E)) between adjacent event pairs. The specific calculation method is θ = α × var(D) + β × std(E), where α and β are coefficients that balance the contributions of the two components and can be set to α = 0.7 and β = 0.3. This adaptive threshold can be dynamically adjusted based on the distribution characteristics of the data itself, making it more robust.

[0057] Finally, adjacent event pairs with a break probability P greater than the adaptive break threshold θ are marked as logical break points. For example, in a sequence containing 1000 events, the break probabilities of adjacent event pairs are calculated. Suppose the break probabilities of pairs 150 and 632 are 0.78 and 0.85, respectively, and the adaptive threshold θ is 0.75. These two event pairs are marked as logical break points, meaning that events 151 and 633 are the starting points of a new logical segment.

[0058] Traditional event sequence analysis techniques typically operate on a single time scale, failing to effectively capture the complex cross-scale correlations between events. Common event segmentation methods rely primarily on fixed thresholds or manually preset rules. These methods often exhibit low adaptability and accuracy when faced with complex and changing urban public safety risk scenarios. For example, sliding window-based event segmentation algorithms struggle to handle event sequences of varying lengths; clustering-based methods are sensitive to parameter settings and have limited generalization capabilities; and traditional feature extraction techniques mostly focus solely on feature changes in a single dimension, ignoring the interactions between multidimensional features. The method of this embodiment captures feature evolution patterns at different time scales by performing eigenvalue decomposition on the event feature matrix and constructing a third-order tensor. By introducing feature spectrum analysis and feature entropy calculation, it simultaneously considers local fluctuations and cross-scale correlations, providing a more comprehensive picture of structural changes in event sequences. Compared to existing technologies, the method of this embodiment incorporates an adaptive break threshold mechanism that dynamically adjusts the threshold based on data distribution characteristics, effectively addressing the over-segmentation or under-segmentation issues caused by fixed thresholds. By integrating local fluctuations and cross-scale correlation information, this method can more accurately identify key turning points in the public safety risk chain. Experimental results demonstrate that this method improves the accuracy of identifying logical breakpoints in different types of urban public safety event sequences, increases processing speed, and demonstrates greater robustness in the face of noise interference and data loss. The application of this technology makes the identification of public safety risk chains more accurate and efficient, providing strong support for the timely identification of potential risk points and the development of prevention and response strategies.

[0059] Figure 3 The following figure shows a performance comparison between multi-scale feature analysis and traditional methods. The data in the figure clearly shows that the multi-scale feature analysis method performs best across all evaluation metrics. In terms of recognition accuracy, the multi-scale method achieved 91.5%, significantly higher than the 70.6% of the single-scale method and the 60.8% of the fixed-threshold method. In terms of noise immunity, the multi-scale method achieved 86.3%, nearly 30 percentage points higher than the single-scale method (58.5%) and over 40 percentage points higher than the fixed-threshold method (43.2%), demonstrating exceptional noise resistance.

[0060] Notably, in terms of processing speed, the fixed threshold method (72.5%) outperformed the single-scale method (56.3%), but still lagged behind the multi-scale method (78.4%). This demonstrates that multi-scale feature analysis improves performance without sacrificing processing efficiency. Finally, in the comprehensive evaluation metric F1-Score, the multi-scale method also leads with a high score of 89.2%, further confirming its advantage in balancing precision and recall.

[0061] Overall, the chart strongly demonstrates the comprehensive advantages of the multi-scale feature analysis method in the task of identifying logical breakpoints in the risk chain, especially its excellent performance in processing complex data and resisting noise interference.

[0062] In an optional embodiment, based on the logically complete risk chain, the conduction coefficient matrix analysis and state transition matrix decomposition are used to extract the conduction strength characteristics, calculate the credibility score, and obtain the risk chain credibility score set including: For a logically complete risk chain, a transmission coefficient matrix is ​​constructed based on the fluctuation amplitude, propagation distance, and impact range of adjacent events. The risk transmission intensity sequence is calculated through the temporal evolution of the transmission coefficient matrix. Constructing a state transfer matrix for the risk conduction intensity sequence, performing eigenvalue decomposition on the state transfer matrix, calculating and determining eigenvalues, determining events for which the product of the eigenvalue and the impact range exceeds a preset conduction threshold as diffusion conduction nodes, and calculating the energy diffusion index of the diffusion conduction nodes; Dividing the risk transmission intensity sequence into multiple time scales, calculating the fluctuation characteristic sequence at each time scale, performing segmented iterative reconstruction on the fluctuation characteristic sequence to obtain a fractal sequence, and calculating the complexity index of risk transmission through the self-similarity of the fractal sequence; Based on the risk transmission intensity sequence and the fluctuation characteristic sequence, a three-dimensional phase space including transmission intensity, fluctuation amplitude and time delay is constructed, the phase space trajectory is extracted, and the Lyapunov exponent of the phase space trajectory at different time scales is calculated; A three-dimensional feature vector is constructed based on the energy diffusion index, complexity index and Lyapunov index. The conduction stability score of the logically complete risk chain is calculated using a pre-trained adaptive weight network. The conduction stability score is transformed into a probability distribution to obtain a risk chain credibility score set.

[0063] In one specific implementation, when conducting risk transmission analysis on a logically complete risk chain, a transmission coefficient matrix is ​​constructed based on the fluctuation amplitude, propagation distance, and impact range of adjacent events. The fluctuation amplitude is calculated by the rate of change of key indicators before and after the event. For example, if the crowd density in a public place suddenly increases from 2.3 people / square meter to 4.8 people / square meter, the fluctuation amplitude is 108.7%. The propagation distance is a weighted combination of the geographic distance and time interval between events. For example, if the physical distance between two security incidents is 1.2 kilometers and the time interval is 0.5 hours, the normalized propagation distance is 0.65. The impact range is comprehensively assessed based on the size of the geographic area covered by the event and the number of people affected. For example, if the impact range of a public security incident is a circular area with a radius of 2.3 kilometers and an impact population of approximately 25,000, the normalized impact range index is 0.78. These three parameters are combined to form a transmission coefficient matrix with a dimension of n×n, where n is the number of events in the risk chain. For a risk chain containing 12 events, a 12×12 transmission coefficient matrix is ​​generated, where each element in the matrix represents the strength of the transmission relationship between a pair of events.

[0064] The risk transmission strength sequence is calculated by analyzing the temporal evolution of the transmission coefficient matrix. The specific method is to divide the transmission coefficient matrix into multiple time windows in chronological order, with a window size of 3 and a sliding step size of 1. The transmission coefficients within each time window are weighted and accumulated to obtain the transmission strength value for that window. For example, for a risk chain of 12 events, a sequence of 10 transmission strength values ​​can be obtained. This sequence is normalized to the interval [0, 1] to facilitate subsequent analysis.

[0065] A state transition matrix is ​​constructed for the risk transmission intensity sequence, discretizing the transmission intensity values ​​into multiple state levels. For example, the interval [0, 1] is equally divided into five states: [0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), and [0.8, 1]. The state transition frequencies between adjacent transmission intensity values ​​are counted to construct a 5×5 state transition matrix. This matrix is ​​subjected to eigenvalue decomposition to calculate the eigenvalue set. The eigenvalue is multiplied by the corresponding event's impact range. If the product exceeds the preset transmission threshold of 0.5, the event is identified as a diffusion transmission node. Assume that in a risk chain of 12 events, three events are identified as diffusion transmission nodes. The products of their eigenvalues ​​and impact ranges are 0.67, 0.58, and 0.72, respectively. For each diffusion transmission node, an energy diffusion index is calculated, which reflects the rate of outward diffusion of event energy. This index is calculated by multiplying the rate of change in transmission intensity with the rate of change in impact range within a certain time window before and after the node, and then multiplying by a decay factor. For example, for a diffusion conduction node with a product of eigenvalue and influence range of 0.72, its energy diffusion index is calculated to be 0.63.

[0066] The risk transmission intensity series is divided into multiple time scales and decomposed using the wavelet transform method. At each time scale, the volatility characteristic series is calculated by calculating the local volatility and volatility frequency. The risk chain of 12 events can be divided into three time scales: short-term (1-2 hours), medium-term (6-12 hours), and long-term (over 24 hours). The volatility characteristic series at each time scale is reconstructed piecewise and iteratively to obtain a fractal series. The reconstruction process uses the sliding window R / S analysis method, with the window size starting at 4 and increasing by 2 each time until it reaches half the series length. The Hurst exponent is obtained by fitting the rescaled range values ​​at different window sizes. A risk transmission complexity index is calculated based on the self-similarity of the fractal series. This index reflects the complexity and uncertainty of the risk transmission process. For example, the complexity indexes for the short-term, medium-term, and long-term time scales are 0.72, 0.58, and 0.41, respectively, indicating that short-term risk transmission is more complex.

[0067] Based on the risk transmission intensity sequence and the volatility characteristic sequence, a three-dimensional phase space was constructed, containing transmission intensity, volatility amplitude, and time delay. The transmission intensity ranged from [0, 1], the volatility amplitude ranged from [-1, 1], and the time delay ranged from [0, 24] hours. Phase space trajectories were extracted from this three-dimensional space, with the number of trajectory points equal to the number of events. The Lyapunov exponent of the phase space trajectory was calculated at different time scales. This exponent reflects the system's sensitivity to initial conditions. The calculation method is to select adjacent pairs of points in the phase space, track their trajectory development, and measure the growth rate of the distance between the trajectories. For the risk chain of the 12 events mentioned above, the Lyapunov exponents were 0.24, 0.16, and 0.09 at the short, medium, and long time scales, respectively, indicating that the risk transmission system is more sensitive to initial conditions in the short term.

[0068] A three-dimensional feature vector is constructed based on the energy diffusion index, complexity index, and Lyapunov exponent. For the above risk chain, the feature vector is [0.63, 0.72, 0.24]. A pre-trained adaptive weight network is used to calculate the conduction stability score of the logically complete risk chain. The adaptive weight network consists of a three-layer neural network, with the three-dimensional feature vector as input, eight neurons in the hidden layer, and a single stability score as output. The network is trained on 500 historical risk chain cases to learn the mapping relationship between features and stability. When the feature vector [0.63, 0.72, 0.24] is input, the network outputs a conduction stability score of 0.37. The conduction stability score is transformed through a probability distribution to a credibility score in the interval [0, 1]. This transformation uses a sigmoid function, resulting in a credibility score of 0.68. This process is repeated for each event segment in the entire risk chain, resulting in a set of risk chain credibility scores, such as [0.68, 0.42, 0.75, 0.83].

[0069] Traditional urban public safety risk chain analysis methods primarily rely on statistical models and rule matching, making it difficult to effectively capture the nonlinear dynamic characteristics of the risk transmission process. Common analysis techniques such as Markov chain models and linear regression methods can describe the temporal correlation of risk events, but they have limitations when dealing with multi-scale transmission characteristics. The method of this embodiment introduces a multi-scale analysis framework, treating the risk transmission process as a complex system with fractal properties. By constructing a three-dimensional phase space and calculating the Lyapunov exponent, it more comprehensively captures the dynamic characteristics of risk transmission. Compared with traditional methods, this method addresses the inadequate characterization of scale dependence and nonlinear transmission characteristics in risk transmission assessment. By integrating the energy diffusion index, complexity index, and Lyapunov exponent, a more comprehensive representation of risk transmission characteristics is constructed. Experimental results show that this method improves the accuracy of risk chain identification in complex urban environments compared to traditional methods, enhances the accuracy of transmission path prediction, and significantly enhances robustness to data noise and missing information. The application of this technology provides a more reliable risk chain identification tool for urban public safety management, helping to identify potential risk points in advance and implement targeted preventive measures.

[0070] In an optional embodiment, performing segmented iterative reconstruction on the fluctuation characteristic sequence to obtain a fractal sequence includes: Obtaining a fluctuation characteristic sequence, processing the fluctuation characteristic sequence using an overlapping segmentation method, determining the number of overlapping points of adjacent segments by a preset segment length and overlapping rate, and generating a plurality of overlapping sequence segments; Extracting fluctuation characteristics of each of the overlapping sequence segments, the fluctuation characteristics including fluctuation amplitude and fluctuation trend, and pairing adjacent overlapping sequence segments in pairs according to the fluctuation characteristics; For each pair of adjacent overlapping sequence segments, a segmented mapping rule is constructed based on the fluctuation characteristics. The fluctuation amplitude ratio of the adjacent overlapping sequence segments is used as the amplitude ratio parameter, and the fluctuation trend difference is used as the trend difference parameter to generate a piecewise linear mapping function. Iteratively reconstruct the overlapping sequence segments. In each iteration, the amplitude ratio parameter and trend difference parameter are dynamically adjusted based on the characteristic change of the current overlapping sequence segment, and act on the next iterative reconstruction to generate a reconstructed sequence segment. The reconstructed sequence segments are sequentially spliced ​​and smoothed to obtain a fractal sequence.

[0071] In one specific embodiment, obtaining a fluctuation signature sequence is a key step in analyzing urban public safety risk chains. This sequence reflects the fluctuating characteristics of risk events across time and space. The fluctuation signature sequence can be obtained by processing raw risk indicator data. For example, by normalizing the safety risk index for a region over 30 consecutive days, a fluctuation signature sequence with a value range of [0, 1] is obtained. The fluctuation signature sequence is processed using an overlapping segmentation method. The number of overlapping points between adjacent segments is determined by a preset segment length and overlap ratio, generating multiple overlapping sequence segments. In practical applications, a segment length of 10 and an overlap ratio of 50% can be set, resulting in a total of 5 overlapping points between adjacent segments. Taking a 30-point fluctuation signature sequence as an example, five overlapping sequence segments can be generated: segment 1 contains points 1 to 10, segment 2 contains points 6 to 15, segment 3 contains points 11 to 20, segment 4 contains points 16 to 25, and segment 5 contains points 21 to 30. This overlapping segmentation method maintains sequence continuity and avoids information fragmentation caused by segmentation.

[0072] Fluctuation characteristics are extracted for each overlapping sequence segment, including amplitude and trend. The amplitude is calculated as the difference between the maximum and minimum values ​​within the segment, reflecting the severity of the fluctuation. The trend is calculated by dividing the difference between the first and last points of the segment by the number of points, indicating the overall upward or downward trend. For example, for the first segment, assuming its values ​​are [0.2, 0.3, 0.25, 0.4, 0.35, 0.45, 0.5, 0.48, 0.52, 0.55], the amplitude is 0.55-0.2=0.35, and the trend is (0.55-0.2) / 10=0.035, indicating an overall upward trend. Based on the fluctuation characteristics, adjacent pairs of overlapping sequence segments are paired, forming four pairs: segment 1 with segment 2, segment 2 with segment 3, segment 3 with segment 4, and segment 4 with segment 5.

[0073] For each pair of adjacent overlapping sequence segments, a segmented mapping rule is constructed based on the fluctuation characteristics. The fluctuation amplitude ratio of adjacent overlapping sequence segments is used as the amplitude ratio parameter, and the fluctuation trend difference is used as the trend difference parameter to generate a piecewise linear mapping function. Assuming that the fluctuation amplitude of the first segment is 0.35 and the fluctuation amplitude of the second segment is 0.42, the amplitude ratio parameter is 0.42 / 0.35=1.2; if the fluctuation trend of the first segment is 0.035 and the fluctuation trend of the second segment is 0.02, the trend difference parameter is 0.02-0.035=-0.015. The piecewise linear mapping function uses these two parameters to map the feature space of the first segment to the second segment, achieving consistent feature conversion. In the specific mapping process, each point in the first segment is transformed by multiplying it by the amplitude ratio parameter and adding the product of the trend difference parameter and the point position. For example, the value of the 7th point in the 1st segment is 0.5, and the mapped value is 0.5×1.2+(-0.015)×7=0.6-0.105=0.495.

[0074] Overlapping sequence segments are iteratively reconstructed. During each iteration, the amplitude ratio and trend difference parameters are dynamically adjusted based on the characteristic variation of the current overlapping sequence segment. These parameters are then applied to the next iterative reconstruction to generate a reconstructed sequence segment. The number of iterations is typically set to 3 to 5, which is sufficient to achieve good reconstruction results. In the first iteration, the initially calculated amplitude ratio and trend difference parameters are used. Starting from the second iteration, the parameters are fine-tuned based on the difference between the previous iteration result and the target segment. For example, if the fluctuation amplitude of the previous iteration result is 10% smaller than that of the target segment, the amplitude ratio parameter is increased by 0.1; if the fluctuation trend is 0.005 greater than that of the target segment, the trend difference parameter is decreased by 0.005. Through this dynamic adjustment mechanism, the iterative process can gradually approach the target feature.

[0075] The reconstructed sequence segments are sequentially spliced ​​and smoothed to produce a fractal sequence. During the splicing process, a weighted average is used to calculate the final value of overlapping data points, with the weights being related to the point's position in the respective sequence segment. For example, for five overlapping points, the first point has a weight of 0.8 in the first segment and 0.2 in the second segment; the last point has a weight of 0.2 in the first segment and 0.8 in the second segment. Smoothing is performed using a moving average method with a window size of 3 to ensure a smooth transition at the splicing point. The resulting fractal sequence retains the key characteristics of the original fluctuation characteristic sequence while exhibiting improved self-similarity and structural stability.

[0076] For example, the 30-day safety risk fluctuation sequence for a commercial area is [0.25, 0.28, 0.30, 0.35, 0.32, 0.40, 0.45, 0.42, 0.38, 0.48, 0.50, 0.47, 0.52, 0.55, 0.51, 0.58, 0.60, 0.57, 0.55, 0.65, 0.68, 0.64, 0.70, 0.72, 0.68, 0.75, 0.78, 0.76, 0.80, 0.82]. The segment length is set to 10, and the overlap rate is 50%, generating five overlapping sequence segments. The fluctuation amplitude of the first segment is calculated to be 0.23, and the fluctuation trend is 0.023; the fluctuation amplitude of the second segment is 0.20, and the fluctuation trend is 0.020. The amplitude ratio parameter is 0.87, and the trend difference parameter is -0.003. Through three iterative reconstructions, the resulting fractal sequence can more accurately reflect the inherent laws of risk fluctuations, providing a reliable basis for subsequent risk chain identification.

[0077] Traditional methods for analyzing fluctuation characteristics primarily rely on Fourier transforms or wavelet analysis. While these methods can effectively extract frequency-domain features, they struggle to capture the fractal properties and self-similar structures found in nonlinear systems. Common time series analysis techniques, such as the autoregressive integrated moving average (ARIMA) model and long short-term memory (LSTM) networks, often exhibit low prediction accuracy and adaptability when dealing with risk fluctuation series with long-range dependencies. The method of this embodiment introduces an overlapping segmented iterative reconstruction method, which, by constructing dynamic mapping rules, more accurately captures the fractal properties of risk fluctuation series. Compared to traditional methods, the method of this embodiment addresses the scale mixing problem and the difficulty of extracting nonlinear features in risk series analysis. By introducing a dynamic adjustment mechanism for the amplitude ratio parameter and the trend difference parameter, adaptive extraction of multi-scale features is achieved. Experimental results demonstrate that this method improves the prediction accuracy of urban public safety risk fluctuation series compared to traditional methods, enhances the efficiency of identifying fluctuation patterns, and significantly enhances robustness to noise and outliers. The application of this technology provides a more accurate analytical tool for early identification and early warning of urban public safety risk chains, helps to improve the foresight and accuracy of risk management, and achieve timely intervention and effective control of potential risks.

[0078] The urban public safety risk chain identification system based on graph neural network in an embodiment of the present invention includes: The first unit is used to obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and use the associations between public safety events as edges to construct an initial event association topology graph; The second unit is configured to determine an optimal sampling path set based on the initial event correlation topology graph using an adaptive bidirectional jump sampling algorithm through dynamic step size control and timing matching; calculate multidimensional correlation strength based on the event sequence in the optimal sampling path set, and screen out an initial candidate risk chain set; The third unit is configured to construct a state characteristic matrix based on the initial candidate risk chain set through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; The fourth unit is used to extract the conduction strength characteristics based on the logically complete risk chain by using conduction coefficient matrix analysis and state transition matrix decomposition, calculate the credibility score, and obtain a risk chain credibility score set; The fifth unit is configured to select, based on the risk chain credibility score set, a risk chain whose credibility score exceeds a preset credibility threshold, and determine and output a final identification result.

[0079] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0080] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0081] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying urban public safety risk chains based on graph neural networks, characterized by: include: Obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and the relationships between public safety events as edges to construct an initial event correlation topology graph; Based on the initial event correlation topology, an adaptive bidirectional jump sampling algorithm is used to determine the optimal sampling path set through dynamic step size control and timing matching; multi-dimensional correlation strength is calculated based on the event sequence in the optimal sampling path set, and an initial candidate risk chain set is screened; Based on the initial candidate risk chain set, a state feature matrix is ​​constructed through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; Based on the logically complete risk chain, conduction coefficient matrix analysis and state transition matrix decomposition are used to extract conduction strength characteristics, calculate credibility scores, and obtain a risk chain credibility score set; Based on the risk chain credibility score set, the risk chain with a credibility score exceeding a preset credibility threshold is selected, and the final identification result is determined and output.

2. The method according to claim 1, characterized in that Based on the initial event correlation topology, an adaptive bidirectional jump sampling algorithm is used to determine the optimal sampling path set through dynamic step size control and timing matching, including: Obtain the influence calculation parameters of the public safety event node from the initial event correlation topology graph, calculate the propagation influence based on the node correlation and distance attenuation factor, calculate the spatial influence based on the ratio of the public safety event node to the affected area, and calculate the temporal influence based on the event time attenuation function. A weighted combination of the propagation influence, spatial influence, and temporal influence is used to obtain the node influence metric. The node influence metric and the baseline sampling step are used to calculate the initial adaptive step size. The adaptive control threshold is obtained based on the historical difference of the node influence metric. The dynamic sampling step size is obtained by combining the initial adaptive step size and the adaptive control threshold. Bidirectional sampling is performed using a dynamic sampling step size to calculate the forward propagation probability and backward verification probability respectively, and then weighted fusion is used to obtain the time series matching score; The path continuity index and length constraint index are calculated based on the timing matching score. The timing matching score is nonlinearly weighted with the path continuity index and length constraint index to obtain the path evaluation score. The public safety event sequences with path evaluation scores greater than the preset evaluation threshold are screened to obtain the optimal sampling path set.

3. The method according to claim 1, characterized in that Based on the event sequence in the optimal sampling path set, the multi-dimensional correlation strength is calculated and the initial candidate risk chain set is screened, including: Acquire event sequences from the optimal sampling path set to construct a temporal evolution graph, extract the time intervals between adjacent events and the event-intensive periods, identify key time nodes, calculate the time delay characteristics of event propagation by comparing the moments of adjacent events, extract the periodicity of event occurrence through Fourier transform, and calculate the temporal correlation strength of adjacent events based on the time delay characteristics and periodicity. Based on the geographic coordinate information of each event in the event sequence, a spatial distribution map is constructed, the spatial distance between adjacent events is calculated, and the spatial correlation strength is obtained by combining the spatial distance with the spatial impact range of the event; The event semantic vector is constructed using the description text of each event in the event sequence. The cosine similarity of the semantic vectors of adjacent events is calculated and combined with the event type matching matrix to obtain the semantic association strength. Based on the influence measurement value of each event in the event sequence, the influence correlation degree of adjacent events is calculated, and the influence correlation degree is combined with the influence transfer attenuation function to obtain the influence correlation strength; The multi-dimensional correlation strength is obtained by weighted fusion of temporal correlation strength, spatial correlation strength, semantic correlation strength and influence correlation strength; The event sequences whose multi-dimensional correlation strength is greater than a preset strength threshold are screened to obtain an initial candidate risk chain set.

4. The method according to claim 1, wherein Based on the initial candidate risk chain set, a state feature matrix is ​​constructed through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain including: Constructing the multi-dimensional correlation strength of the event sequence in the initial candidate risk chain set into a state feature matrix, wherein each element of the state feature matrix corresponds to the correlation state between adjacent events; Performing eigenvalue decomposition on the state feature matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs whose fracture probability values ​​are greater than an adaptive fracture threshold as logical fracture points; For the events before and after the logical breakpoint, a valid time window is constructed based on the corresponding timestamp, and a reachable range ellipse is constructed based on the corresponding geographical location; Search for potential transition events within the valid time window and reachable range ellipse, calculate the time dimension contribution, spatial dimension contribution, and semantic dimension contribution of the potential transition event to the previous and next events, and perform weighted combination to obtain the event transition evaluation value; The potential transition event with the highest event transition evaluation value is used as a supplementary event and inserted between the preceding and following events corresponding to the logical breakpoint to generate an optimized event sequence; The state feature matrix of the optimized event sequence is reconstructed and the process is repeated until there are no adjacent event pairs marked as logical breakpoints in the optimized event sequence, thereby obtaining a logically complete risk chain.

5. The method according to claim 4, characterized in that Performing eigenvalue decomposition on the state characteristic matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, calculating the characteristic state difference of adjacent event pairs to determine a fracture probability value, and marking adjacent event pairs whose fracture probability values ​​are greater than an adaptive fracture threshold as logical fracture points includes: Performing eigenvalue decomposition on the state characteristic matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix, and constructing them into a third-order tensor; performing multi-scale tensor decomposition on the third-order tensor to obtain an eigenvalue component matrix, an eigenvector component matrix, and a time scale component matrix at different time scales; constructing a multi-scale characteristic spectrum of the event sequence based on the eigenvalue component matrix, the eigenvector component matrix and the time scale component matrix, and mapping the energy distribution of the multi-scale characteristic spectrum at different time scales into a characteristic entropy matrix; Calculate the characteristic state difference of adjacent event pairs according to the eigenvalue diagonal matrix and the eigenvector matrix, and calculate the entropy difference vector of the adjacent event pairs according to the characteristic entropy matrix; extracting the local fluctuation features of the characteristic state difference and the cross-scale correlation features of the entropy difference vector respectively, fusing the local fluctuation features with the cross-scale correlation features to obtain a comprehensive fracture feature, and calculating the fracture probability value of the adjacent event pair based on the comprehensive fracture feature; An adaptive fracture threshold is constructed based on the variance of the characteristic state difference and the standard deviation of the entropy difference vector of the adjacent event pairs, and the adjacent event pairs whose fracture probability values ​​are greater than the adaptive fracture threshold are marked as logical fracture points.

6. The method according to claim 1, wherein Based on the logically complete risk chain, the conduction coefficient matrix analysis and state transition matrix decomposition are used to extract the conduction strength characteristics, calculate the credibility score, and obtain the risk chain credibility score set including: For a logically complete risk chain, a transmission coefficient matrix is ​​constructed based on the fluctuation amplitude, propagation distance, and impact range of adjacent events. The risk transmission intensity sequence is calculated through the temporal evolution of the transmission coefficient matrix. Constructing a state transfer matrix for the risk conduction intensity sequence, performing eigenvalue decomposition on the state transfer matrix, calculating and determining eigenvalues, determining events for which the product of the eigenvalue and the impact range exceeds a preset conduction threshold as diffusion conduction nodes, and calculating the energy diffusion index of the diffusion conduction nodes; Dividing the risk transmission intensity sequence into multiple time scales, calculating the fluctuation characteristic sequence at each time scale, performing segmented iterative reconstruction on the fluctuation characteristic sequence to obtain a fractal sequence, and calculating the complexity index of risk transmission through the self-similarity of the fractal sequence; Based on the risk transmission intensity sequence and the fluctuation characteristic sequence, a three-dimensional phase space including transmission intensity, fluctuation amplitude and time delay is constructed, the phase space trajectory is extracted, and the Lyapunov exponent of the phase space trajectory at different time scales is calculated; A three-dimensional feature vector is constructed based on the energy diffusion index, complexity index and Lyapunov index. The conduction stability score of the logically complete risk chain is calculated using a pre-trained adaptive weight network. The conduction stability score is transformed into a probability distribution to obtain a risk chain credibility score set.

7. The method according to claim 6, characterized in that The fractal sequence obtained by segmented iterative reconstruction of the fluctuation characteristic sequence includes: Obtaining a fluctuation characteristic sequence, processing the fluctuation characteristic sequence using an overlapping segmentation method, determining the number of overlapping points of adjacent segments by a preset segment length and overlapping rate, and generating a plurality of overlapping sequence segments; Extracting fluctuation characteristics of each of the overlapping sequence segments, the fluctuation characteristics including fluctuation amplitude and fluctuation trend, and pairing adjacent overlapping sequence segments in pairs according to the fluctuation characteristics; For each pair of adjacent overlapping sequence segments, a segmented mapping rule is constructed based on the fluctuation characteristics. The fluctuation amplitude ratio of the adjacent overlapping sequence segments is used as the amplitude ratio parameter, and the fluctuation trend difference is used as the trend difference parameter to generate a piecewise linear mapping function. Iteratively reconstruct the overlapping sequence segments. In each iteration, the amplitude ratio parameter and trend difference parameter are dynamically adjusted based on the characteristic change of the current overlapping sequence segment, and act on the next iterative reconstruction to generate a reconstructed sequence segment. The reconstructed sequence segments are sequentially spliced ​​and smoothed to obtain a fractal sequence.

8. An urban public safety risk chain identification system based on graph neural network, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to obtain historical data on urban public safety events, use different types of public safety events as graph nodes, and use the associations between public safety events as edges to construct an initial event association topology graph; The second unit is configured to determine an optimal sampling path set based on the initial event correlation topology graph using an adaptive bidirectional jump sampling algorithm through dynamic step size control and timing matching; calculate multidimensional correlation strength based on the event sequence in the optimal sampling path set, and screen out an initial candidate risk chain set; The third unit is configured to construct a state characteristic matrix based on the initial candidate risk chain set through eigenvalue decomposition to identify logical breakpoints, search for potential transition events, and iteratively optimize to obtain a logically complete risk chain; The fourth unit is used to extract the conduction strength characteristics based on the logically complete risk chain by using conduction coefficient matrix analysis and state transition matrix decomposition, calculate the credibility score, and obtain a risk chain credibility score set; The fifth unit is configured to select, based on the risk chain credibility score set, a risk chain whose credibility score exceeds a preset credibility threshold, and determine and output a final identification result.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images