Urban domain social risk event evolution analysis and intelligent prediction method based on affair graph

Abstract

The invention discloses a municipal social risk event evolution analysis and intelligent prediction method based on a affair graph, relates to the technical field of artificial intelligence, and solves the problem of lack of hierarchical analysis of municipal social risk event development. The technical problem that it is difficult to more accurately analyze and predict different levels of development of different urban social risk events due to establishment of a social dispute atlas for structuring the levels of development of the urban social risk events is also lacked. Through the development level of the city social risk events, the time sequence and importance of the city social risk events can be clearly displayed, so that the information is easier to understand and analyze. By linking the dispute subject node and the content node which are associated with each other, common features in similar city social risk events can be conveniently identified, and data support is provided for subsequent prediction; by establishing the associated edge and the unassociated edge, the relationship among the nodes can be clearly displayed.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and specifically is a method for analyzing the evolution and intelligent prediction of urban social risk events based on a cause-and-effect graph. Background Art

[0002] Urban social risk events are one of the core concepts of human society, and people's social activities are often driven by urban social risk events. The evolutionary laws and patterns of urban social risk events that occur one after another in time and space are a kind of very valuable knowledge. At present, the core objects of knowledge bases such as knowledge graphs and semantic networks are not urban social risk events, and there is a lack of characterization of the important human knowledge of logic. In order to make up for this deficiency, the logic graph came into being, which can reveal the evolutionary laws and development logic of urban social risk events, and characterize and record human behavior activities. The present invention aims at major cases of social contradictions and disputes, establishes a multi-dimensional social contradiction and dispute logic graph, and supports the functions of automatic collection of social contradiction and dispute information, analysis of evolutionary laws, intelligent judgment of homologous disputes, and intelligent guidance.

[0003] The urban social risk event evolution analysis method based on event logic graph is a method that uses event logic graph to model and analyze the complex temporal evolution relationship between urban social risk events. Event Logi c Graph (ELG) is an event logic knowledge base that describes the evolution laws and patterns between urban social risk events.

[0004] Most of the evolution analysis methods of urban social risk events directly use historical urban social risk event data to conduct predictive analysis of urban social risk events caused by social disputes. It is difficult to analyze and predict different levels of development of urban social risk events through structured social dispute cause maps at different levels of development of urban social risk events, which affects the accuracy of the evolution analysis of urban social risk events. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method for analyzing the evolution and intelligent prediction of urban social risk events based on a cause-and-effect graph, which is used to solve the technical problems of lack of hierarchical analysis of the development of urban social risk events and lack of establishment of a structured social dispute cause-and-effect graph for each level of the development of urban social risk events, making it difficult to conduct more accurate analysis and prediction of different levels of the development of different urban social risk events.

[0006] To solve the above problems, the first aspect of the present invention provides a method for analyzing the evolution of urban social risk events and intelligent prediction based on a reasoning graph, comprising the following steps:

[0007] Collect information data on social disputes and urban social risk events, and clean the collected urban social risk event information data through natural language processing technology;

[0008] Extract the dispute group attribute information, dispute category information and dispute content information of the social risk events in the city of social disputes, and classify the social risk events in the city with the same dispute group attribute information and dispute category information into the same category of social risk events in the city. The dispute content information includes: dispute time point, dispute area, focus of the social risk event in the city and the popularity of the social risk event in the city.

[0009] According to the time point of the dispute, the city-wide social risk event development level is set up. In the city-wide social risk event development level, a dispute subject node is established, and the dispute area is marked on the node. The interrelated dispute subject nodes are linked to the dispute content nodes, including: the focus of the city-wide social risk event and the heat of the city-wide social risk event;

[0010] In the classification of social risk events in each city, the similarity of each dispute subject node and dispute content node is evaluated, and associated edges are established between high-similarity nodes, and unassociated edges are established between low-similarity nodes. The attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map are constructed;

[0011] The unfinished social dispute municipal social risk events are collected and assigned to the corresponding municipal social risk event categories. Based on the constructed social dispute cause-and-effect map, the unfinished social dispute municipal social risk events are analyzed for evolution and dispute prediction.

[0012] Optionally, in an example of the above aspect, a city-wide social risk event development level is set according to the time point of the dispute, a dispute subject node is established in the city-wide social risk event development level, and the dispute area is marked on the node, and the mutually related dispute subject nodes are linked to the dispute content node, including the following steps:

[0013] The development level of the city-wide social risk events is set up according to the time point of the dispute, including: basic facts layer, main city-wide social risk event layer, impact diffusion layer and mediation intervention layer;

[0014] In each level, independent nodes are created for the dispute subject and the mediation subject as subject nodes. The subject node attributes include: dispute subject node and mediation subject node;

[0015] By dividing the dispute area of ​​the dispute subject node into administrative regions, the dispute area is marked on the dispute subject node;

[0016] Count the number of disputes involving the same type of dispute subject nodes of the subject node, and use the Redd it algorithm to count the popularity of social risk events in the city;

[0017] Calculate the subject influence based on the number of dispute participations of the same type of dispute subject nodes of the subject node and the heat of the urban social risk events, and mark the subject influence on the subject node;

[0018] Establish dispute content nodes corresponding to the levels, and store the key points of the urban social risk events and the heat data of the urban social risk events in the dispute content nodes;

[0019] Extract the subject information keywords in the key points of the urban social risk events, calculate the cosine similarity between the subject information keywords and the subject nodes, and link the subject nodes with the cosine similarity greater than the preset threshold to the corresponding dispute content nodes.

[0020] In this embodiment, the basic fact layer: records objective information such as the time, place, and involved subjects of the dispute;

[0021] The main urban social risk event layer: starting from the start of the key behavior urban social risk event of the dispute escalation and ending at the end of the key behavior urban social risk event of the dispute escalation, mark the key behaviors that lead to the dispute escalation, such as the outbreak of conflicts and the application for legal proceedings;

[0022] The influence diffusion layer: starting from the time point when the mediation subject node starts to intervene and ending at the time point when the mediation subject node ends to intervene, mark the diffusion influence of the dispute on related personnel, regions, or social public opinions;

[0023] The mediation intervention layer: starting from the time point when the official takes mediation measures and legal rulings and ending at the time point when the mediation measures and legal rulings end, record the mediation measures, legal rulings, and implementation results.

[0024] There may be intersections in time between each layer.

[0025] The key points of the urban social risk events and the heat data of the urban social risk events are stored in the dispute content nodes. The key points of the urban social risk events stored in the dispute content nodes include: dispute demands (economic compensation, restoration of rights and interests); focus of disputes (ambiguity of contract terms, attribution of responsibilities); evidentiary materials.

[0026] Optionally, in an example of the above aspect, calculate the subject influence based on the number of dispute participations of the same type of dispute subject nodes of the subject node and the heat of the urban social risk events, and calculate through the following formula:

[0027]

[0028] where d in is the subject influence, and C inFor the same - type dispute subject nodes of the main body node, the number of participations in the urban - rural social risk events of the same dispute category, C0 is the number of dispute participations of the same - type dispute subject nodes of the main body node, R in For the same - type dispute subject nodes of the main body node, the average heat value of the urban - rural social risk events of the same dispute category, R0 is the average heat value of the urban - rural social risk events of the same dispute category.

[0029] Optionally, in an example of the above - mentioned aspect, in each classification of urban - rural social risk events, to evaluate the similarity between each dispute subject node and dispute content node, the following steps are included:

[0030] Extract features from the dispute subject nodes, including: regional features, influence features, and social dispute relationship features;

[0031] Among them, regional features: Collect the standard administrative region division of the dispute subject node and map it to the map;

[0032] Influence features: Collect the marked subject influence value of the dispute subject node;

[0033] Social relationship features: Collect the number of times that the dispute subject nodes of the urban - rural social risk events participate in disputes together in the same classification of urban - rural social risk events;

[0034] Analyze the regional similarity, subject influence similarity, and social relationship feature similarity, and calculate the similarity between dispute subject nodes according to the regional similarity, subject influence similarity, and social relationship feature similarity;

[0035] Extract features from the dispute content nodes, including: key features of urban - rural social risk events, heat features of urban - rural social risk events, and time - decay features;

[0036] Among them, key features of urban - rural social risk events: Extract the key semantic information of the disputed focus and demands through TF - IDF or BERT embedding;

[0037] Heat features of urban - rural social risk events: Based on the social media discussion volume, news report frequency, and like - forwarding times, statistically calculate the heat value of urban - rural social risk events through the Reddit algorithm;

[0038] Time - decay features: In each level, divide the urban - rural social risk events into several time intervals on average. By calculating the difference between the heat values of the urban - rural social risk events at the start and end points of the time interval, form an array of time - decay features {A1, A2, …, An} with the heat differences of each time interval;

[0039] Analyze the similarity of key features, heat features, and time decay features of urban social risk events. Calculate the similarity between dispute content nodes based on the similarity of key features, heat features, and time decay features of urban social risk events.

[0040] Optionally, in an example of the above aspect, analyze the regional similarity, subject influence similarity, and social relationship feature similarity. Calculate the similarity between dispute subject nodes based on the regional similarity, subject influence similarity, and social relationship feature similarity, including the following steps:

[0041] Calculate the distance between dispute subject nodes and count the average distance between dispute subject nodes in the same classification of urban social risk events. Calculate the regional similarity Dlo = distance between dispute subject nodes / average distance between dispute subject nodes in the same classification of urban social risk events;

[0042] Calculate the difference between the subject influence values marked by dispute subject nodes and count the average value of the differences in subject influence between dispute subject nodes in the same classification of urban social risk events. Calculate the subject influence similarity Din = difference between subject influence values / average value of the differences in subject influence between dispute subject nodes in the same classification of urban social risk events;

[0043] Count the number of times a single dispute subject node participates in disputes in the same classification of urban social risk events. Calculate the social relationship feature similarity Dso = number of times of jointly participating in disputes in the same classification of urban social risk events / sum of the number of times each dispute subject node participates in disputes in the same classification of urban social risk events;

[0044] Based on the regional similarity, subject influence similarity, and social relationship feature similarity, calculate the similarity between dispute subject nodes through the following formula:

[0045] Spa = w1 * Dlo + w2 * Din + w3 * Dso

[0046] Among them, Spa is the similarity between dispute subject nodes, and w1, w2, and w3 are the weights corresponding to the regional similarity, subject influence similarity, and social relationship feature similarity;

[0047] Optionally, in an example of the above aspect, analyze the similarity of key features, heat features, and time decay features of urban social risk events. Calculate the similarity between dispute content nodes based on the similarity of key features, heat features, and time decay features of urban social risk events, including the following steps:

[0048] By calculating the cosine similarity between the key semantic information extracted between the dispute content nodes and counting the average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification, the key feature similarity of the city-wide social risk event Dke is calculated = cosine similarity between the key semantic information extracted between the dispute content nodes / average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification;

[0049] By calculating the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes, and counting the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification, the similarity of the key characteristics of the city-wide social risk events Dpo is calculated = the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes / the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification;

[0050] Count the time decay feature array between dispute content nodes, the difference between the heat differences of corresponding time intervals, and the average heat value of dispute content nodes, and calculate the time decay feature similarity Dt i = (the difference between the heat differences of corresponding time intervals between dispute content nodes / the average heat value of dispute content nodes) / n;

[0051] Based on the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events, and the similarity of time decay features, the similarity between dispute content nodes is calculated using the following formula:

[0052] Sco=r1*Dke+r2*Dpo+r3*Dti

[0053] Among them, Sco is the similarity between dispute content nodes, r1, r2 and r3 are the weights corresponding to the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events and the similarity of time attenuation features.

[0054] Optionally, in an example of the above aspect, establishing associated edges between high-similarity nodes, establishing unassociated edges between low-similarity nodes, constructing attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map, including the following steps:

[0055] The similarities between the dispute subject nodes and the dispute content nodes are counted respectively. If the similarity between the dispute subject nodes is greater than the first threshold of the dispute subject node similarity, the dispute subject nodes are judged to be high-similarity nodes. If the similarity between the dispute subject nodes is less than the second threshold of the dispute subject node similarity, the dispute subject nodes are judged to be low-similarity nodes.

[0056] If the similarity between the dispute content nodes is greater than the first threshold of the dispute content node similarity, the dispute content nodes are judged to be high-similarity nodes; if the similarity between the dispute content nodes is less than the second threshold of the dispute content node similarity, the dispute content nodes are judged to be low-similarity nodes;

[0057] Establish associated edges between high-similarity nodes, and unassociated edges between low-similarity nodes, and construct a social dispute cause-and-effect graph corresponding to the classification of urban social risk events.

[0058] Optionally, in an example of the above aspect, unfinished social dispute city-level social risk events are grouped and assigned to corresponding city-level social risk event categories, and based on the constructed social dispute reasoning map, evolution analysis and dispute prediction are performed on the unfinished social dispute city-level social risk events, including the following steps:

[0059] Collect the unfinished social dispute city-level social risk events, assign them to the corresponding city-level social risk event classification, and obtain the social dispute event map corresponding to the city-level social risk event classification;

[0060] An incomplete map of unfinished urban social risk events will be constructed for unfinished urban social risk events involving social disputes;

[0061] Analyze the similarity between dispute subject nodes and dispute content nodes in the hierarchical graph of unfinished social dispute city-wide social risk events and the corresponding hierarchical social dispute graph, and select nodes with similarity greater than a threshold as high-similarity nodes with corresponding nodes of unfinished social dispute city-wide social risk events;

[0062] Conduct centrality analysis on the selected high-similarity nodes, and select the main reference dispute subject node and the main reference dispute content node from the high-similarity nodes;

[0063] The main reference dispute subject node and the main reference dispute content node are fused with other corresponding high-similarity nodes to conduct evolutionary analysis and dispute prediction on unfinished social dispute urban social risk events.

[0064] Optionally, in an example of the above aspect, centrality analysis is performed on the selected highly similar nodes, and the main reference dispute subject node and the main reference dispute content node are selected from the highly similar nodes, including the following steps:

[0065] Count the number of edges connected to each of the selected highly similar nodes, as well as the number of unconnected edges;

[0066] Perform centrality analysis on the selected highly similar nodes through the following formula:

[0067]

[0068] Where Q is the centrality analysis value of the corresponding highly similar node, C0 is the number of edges connected to the corresponding highly similar node, C1 is the number of unconnected edges connected to the corresponding highly similar node, and S is the similarity of the corresponding highly similar node to the uncompleted social dispute urban social risk event;

[0069] Select the dispute subject node and the dispute content node with the largest centrality analysis value among the highly similar nodes as the main reference dispute subject node and the main reference dispute content node.

[0070] Optionally, in an example of the above aspect, information fusion is performed on the main reference dispute subject node and the main reference dispute content node with other corresponding highly similar nodes, and evolutionary analysis and dispute prediction are performed on the uncompleted social dispute urban social risk events, including the following steps:

[0071] Extract the content in the main reference dispute subject node and the main reference dispute content node, and extract the content of the nodes at the subsequent levels of the main reference dispute subject node and the main reference dispute content node, and use the extracted content as the main reference content;

[0072] Extract the content of other corresponding highly similar nodes, and extract the content of the dispute content nodes at the subsequent levels of the corresponding nodes. After summarization through large language processing technology, it is used as the auxiliary reference content.

[0073] Compared with the prior art, the beneficial effects of the present invention are:

[0074] The present invention facilitates clearly demonstrating the chronological order and importance of urban social risk events through the development hierarchy of urban social risk events, making the information easier to understand and analyze. By linking the interrelated dispute subject nodes and content nodes, the interactive relationships between different subjects and their impacts on the development of urban social risk events can be deeply analyzed. It is convenient to identify the common features in similar urban social risk events, providing data support for subsequent prediction. By annotating various information such as the dispute area, key points, and popularity of urban social risk events in the nodes, more comprehensive background information is provided, which helps to understand the complexity of urban social risk events.

[0075] The present invention facilitates clearly demonstrating the relationships between each node by establishing associated edges and unassociated edges, helping users quickly understand the connections between different dispute subjects and contents. The graph structure makes the information hierarchical, facilitating analysis and mining; by the associated edges between high-similarity nodes and unassociated edges between low-similarity nodes, it is convenient to help identify more similar and representative dispute groups, thereby conducting effective clustering analysis. It can identify the common features of different types of disputes, providing data support for subsequent research. By analyzing the attribute information and category information of different dispute groups and monitoring high-similarity nodes, it is convenient to accurately predict the subsequent development of urban social risk events. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0077] Figure 1 It is a schematic diagram of the principle of the present invention;

[0078] Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0080] Please refer to Figure 1 - Figure 2 , the first aspect embodiment of the present invention provides an evolution analysis and intelligent prediction method for urban social risk events based on an event logic graph, including the following steps:

[0081] Collect information data on social disputes and urban social risk events, and clean the collected urban social risk event information data through natural language processing technology;

[0082] Extract the dispute group attribute information, dispute category information and dispute content information of the social risk events in the city of social disputes, and classify the social risk events in the city with the same dispute group attribute information and dispute category information into the same category of social risk events in the city. The dispute content information includes: dispute time point, dispute area, focus of the social risk event in the city and the popularity of the social risk event in the city.

[0083] According to the time point of the dispute, the city-wide social risk event development level is set up. In the city-wide social risk event development level, a dispute subject node is established, and the dispute area is marked on the node. The interrelated dispute subject nodes are linked to the dispute content nodes, including: the focus of the city-wide social risk event and the heat of the city-wide social risk event;

[0084] In the classification of social risk events in each city, the similarity of each dispute subject node and dispute content node is evaluated, and associated edges are established between high-similarity nodes, and unassociated edges are established between low-similarity nodes. The attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map are constructed;

[0085] The unfinished social dispute municipal social risk events are collected and assigned to the corresponding municipal social risk event categories. Based on the constructed social dispute cause-and-effect map, the unfinished social dispute municipal social risk events are analyzed for evolution and dispute prediction.

[0086] Specifically, in this embodiment, the dispute group attribute information, dispute category information and dispute content information of the social dispute city social risk event are extracted, including:

[0087] Dispute group attribute information: Identify the participating groups in urban social risk events, such as farmers, workers, enterprises, government departments, etc.; extract group characteristics, such as age, gender, occupation and region.

[0088] Dispute category information: Based on the description of the city-wide social risk event, determine what type of dispute it is, such as land disputes, labor disputes, contract disputes, environmental pollution disputes, etc. The judgment of the dispute category should be based on the core controversial points of the city-wide social risk event.

[0089] Dispute content information, including:

[0090] Dispute time point: Extract the specific time or time period when the social risk event occurs in the city.

[0091] Dispute area: Determine the geographical location where the urban social risk event occurs, such as cities, villages, specific areas, and specific location information.

[0092] Key points of urban social risk events: Summarize the main controversial points or core contents of urban social risk events.

[0093] The intensity of social risk events in a city: It can be assessed by the number of media reports, the intensity of discussion on social media, etc.

[0094] Based on the extracted dispute group attribute information and dispute category information, a preliminary classification of urban social risk events is carried out.

[0095] Municipal social risk events with the same dispute group attribute information and dispute category information are classified into the same category.

[0096] Set the city-wide social risk event development level according to the time point of the dispute, establish the dispute subject node in the city-wide social risk event development level, mark the dispute area on the node, and link the interrelated dispute subject nodes to the dispute content node;

[0097] Through the development level of urban social risk events, it is easy to clearly display the time sequence and importance of urban social risk events, making the information easier to understand and analyze. By linking the interrelated dispute subject nodes and content nodes, we can deeply analyze the interactive relationship between different subjects and their impact on the development of urban social risk events. It is easy to identify the common characteristics of similar urban social risk events and provide data support for subsequent predictions. By marking the dispute area, the focus and popularity of urban social risk events in the nodes, more comprehensive background information is provided, which helps to understand the complexity of urban social risk events.

[0098] In the classification of social risk events in each city, the similarity of each dispute subject node and dispute content node is evaluated, and associated edges are established between high-similarity nodes, and unassociated edges are established between low-similarity nodes. The attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map are constructed;

[0099] By establishing associated edges and unassociated edges, the relationship between nodes can be clearly displayed, helping users to quickly understand the connection between different dispute subjects and contents. The graph structure makes the information hierarchy clear and easy to analyze and mine. The associated edges between high-similarity nodes and the unassociated edges between low-similarity nodes help identify more similar and representative dispute groups, thereby conducting effective cluster analysis. It can identify the common characteristics of different types of disputes and provide data support for subsequent research. By analyzing the attribute information and category information of different dispute groups and monitoring high-similarity nodes, it is easy to accurately predict the subsequent development of urban social risk events.

[0100] As new data is added, the graph can be dynamically updated to maintain its accuracy and timeliness; new nodes, new attributes or new categories can be added as needed to adapt to the ever-changing social environment. This will help to better predict the development trajectory of urban social risk events and provide reference for similar cases in the future.

[0101] The unfinished social dispute municipal social risk events are collected and assigned to the corresponding municipal social risk event categories. Based on the constructed social dispute cause-and-effect map, the unfinished social dispute municipal social risk events are analyzed for evolution and dispute prediction.

[0102] By analyzing the evolution of unfinished urban social risk events, it is easy to identify the development trajectories of different urban social risk events. Using the data in the social dispute map to predict the development of future time is easy to improve the accuracy of the prediction of the future development of unfinished urban social risk events. By analyzing the information of different dispute groups, new insights can be obtained to promote theoretical research and practical improvements.

[0103] In one embodiment of the present invention, a city-wide social risk event development level is set according to the time point of the dispute, a dispute subject node is established at the city-wide social risk event development level, and the dispute area is marked at the node, and the mutually related dispute subject nodes are linked to the dispute content node, including the following steps:

[0104] The development level of the city-wide social risk events is set up according to the time point of the dispute, including: basic facts layer, main city-wide social risk event layer, impact diffusion layer and mediation intervention layer;

[0105] In each level, independent nodes are created for the dispute subject and the mediation subject as subject nodes. The subject node attributes include: dispute subject node and mediation subject node;

[0106] By dividing the dispute area of ​​the dispute subject node into administrative regions, the dispute area is marked on the dispute subject node;

[0107] Count the number of disputes involving the same type of dispute subject nodes of the subject node, and use the Redd it algorithm to count the popularity of social risk events in the city;

[0108] Calculate the subject influence based on the number of disputes involving the subject nodes of the same type of disputes and the heat of urban social risk events, and mark the subject influence on the subject node;

[0109] Establish dispute content nodes corresponding to the levels, and store the key points and popularity data of urban social risk events in the dispute content nodes;

[0110] Extract and store the subject information keywords in the key points of the city's social risk events, calculate the cosine similarity between the subject information keywords and the subject nodes, and link the subject nodes with cosine similarity greater than the preset threshold to the corresponding dispute content nodes.

[0111] In this embodiment, the basic fact layer: records objective information such as the time, location, and parties involved in the dispute;

[0112] Main city-level social risk event layer: The starting point is the beginning of the city-level social risk event for key behaviors that escalate disputes, and the ending point is the end of the city-level social risk event for key behaviors that escalate disputes. The key behaviors that lead to the escalation of disputes are marked, such as the outbreak of conflicts and the application for legal proceedings.

[0113] Impact diffusion layer: The time point when the mediation subject node begins to intervene is the starting point, and the time point when the mediation subject node ends its intervention is the end point, marking the diffusion impact of the dispute on related personnel, regions or social public opinion;

[0114] Mediation intervention layer: The starting point is the time when the official mediation measures and legal decisions are taken, and the end point is the time when the mediation measures and legal decisions are ended. The mediation measures, legal decisions and execution results are recorded.

[0115] There may be overlap between the various levels in time.

[0116] The dispute content node stores the key points of urban social risk events and the heat data of urban social risk events. The dispute content node stores the key points of urban social risk events, including: dispute demands (economic compensation, restoration of rights and interests); focus of dispute (ambiguous contract terms, attribution of liability); and evidence materials.

[0117] In one embodiment of the present invention, the subject influence is calculated based on the number of disputes participated by the subject nodes of the same type of disputes and the popularity of the city-wide social risk events, and is calculated by the following formula:

[0118]

[0119] Among them, d in is the main influence, C in is the number of times the subject node of the same type of dispute as the subject node participates in the same type of urban social risk events, C0 is the number of times the subject node of the same type of dispute as the subject node participates in the dispute, R in R0 is the average heat value of the social risk events of the same type of dispute subject nodes as the subject node, and R1 is the average heat value of the social risk events of the same dispute category.

[0120] In one embodiment of the present invention, in each classification of urban social risk events, to evaluate the similarity between each dispute subject node and dispute content node, the following steps are included:

[0121] Extract features from the dispute subject nodes, including: regional features, influence features, and social dispute relationship features;

[0122] Among them, regional features: collect the standard administrative region division of the dispute subject node and map it to the map;

[0123] Influence features: collect the subject influence value marked on the dispute subject node;

[0124] Social relationship features: collect the dispute subject nodes of urban social risk events and count the number of times they jointly participate in disputes in the same classification of urban social risk events;

[0125] Analyze the regional similarity, subject influence similarity, and social relationship feature similarity, and calculate the similarity between dispute subject nodes based on the regional similarity, subject influence similarity, and social relationship feature similarity;

[0126] Extract features from the dispute content nodes, including: key features of urban social risk events, popularity features of urban social risk events, and time decay features;

[0127] Among them, key features of urban social risk events: extract the key semantic information of the dispute focus and demands through TF-IDF or BERT embedding;

[0128] Popularity features of urban social risk events: based on the social media discussion volume, news report frequency, and like and forward times, use the Reddit algorithm to statistically calculate the popularity value of urban social risk events;

[0129] Time decay features: in each level, evenly divide the urban social risk events into several time intervals, and by calculating the difference in the popularity values of the urban social risk events at the start and end points of the time intervals, form an array of time decay features {A1, A2,..., An} from the popularity differences of each time interval;

[0130] Analyze the similarity of key features of urban social risk events, the similarity of popularity features of urban social risk events, and the similarity of time decay features, and calculate the similarity between dispute content nodes based on the similarity of key features of urban social risk events, the similarity of popularity features of urban social risk events, and the similarity of time decay features.

[0131] In one embodiment of the present invention, the regional similarity, the subject influence similarity, and the social relationship feature similarity are analyzed. According to the regional similarity, the subject influence similarity, and the social relationship feature similarity, the similarity between dispute subject nodes is calculated, including the following steps:

[0132] By calculating the distance between dispute subject nodes and counting the average distance between dispute subject nodes in the classification of social risk events in the same city domain, the regional similarity Dlo = the distance between dispute subject nodes / the average distance between dispute subject nodes in the classification of social risk events in the same city domain;

[0133] By calculating the difference between the subject influence values marked by dispute subject nodes and counting the average value of the subject influence differences between dispute subject nodes in the classification of social risk events in the same city domain, the subject influence similarity Din = the difference between subject influence values / the average value of the subject influence differences between dispute subject nodes in the classification of social risk events in the same city domain;

[0134] By counting the number of times a single dispute subject node participates in disputes in the classification of social risk events in the same city domain, the social relationship feature similarity Dso = the number of times of jointly participating in disputes in the classification of social risk events in the same city domain / the sum of the number of times each dispute subject node participates in disputes in the classification of social risk events in the same city domain;

[0135] According to the regional similarity, the subject influence similarity, and the social relationship feature similarity, the similarity between dispute subject nodes is calculated through the following formula:

[0136] Spa = w1 * Dlo + w2 * Din + w3 * Dso

[0137] Wherein, Spa is the similarity between dispute subject nodes, and w1, w2, and w3 are the weights corresponding to the regional similarity, the subject influence similarity, and the social relationship feature similarity; in this embodiment, w1, w2, and w3 are 0.2, 0.4, and 0.4 respectively.

[0138] In one embodiment of the present invention, the key feature similarity of city domain social risk events, the popularity feature similarity of city domain social risk events, and the time decay feature similarity are analyzed. According to the key feature similarity of city domain social risk events, the popularity feature similarity of city domain social risk events, and the time decay feature similarity, the similarity between dispute content nodes is calculated, including the following steps:

[0139] By calculating the cosine similarity between the key semantic information extracted between the dispute content nodes and counting the average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification, the key feature similarity of the city-wide social risk event Dke is calculated = cosine similarity between the key semantic information extracted between the dispute content nodes / average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification;

[0140] By calculating the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes, and counting the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification, the similarity of the key characteristics of the city-wide social risk events Dpo is calculated = the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes / the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification;

[0141] Count the time decay feature array between dispute content nodes, the difference between the heat differences of corresponding time intervals, and the average heat value of dispute content nodes, and calculate the time decay feature similarity Dt i = (the difference between the heat differences of corresponding time intervals between dispute content nodes / the average heat value of dispute content nodes) / n;

[0142] Based on the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events, and the similarity of time decay features, the similarity between dispute content nodes is calculated using the following formula:

[0143] Sco=r1*Dke+r2*Dpo+r3*Dti

[0144] Among them, Sco is the similarity between dispute content nodes, r1, r2 and r3 are the weights corresponding to the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events and the similarity of time attenuation features.

[0145] In one embodiment of the present invention, an associated edge is established between high-similarity nodes, an unassociated edge is established between low-similarity nodes, and attribute information of different dispute groups and dispute category information are constructed, and a corresponding social dispute reasoning map includes the following steps:

[0146] The similarities between the dispute subject nodes and the dispute content nodes are counted respectively. If the similarity between the dispute subject nodes is greater than the first threshold of the dispute subject node similarity, the dispute subject nodes are judged to be high-similarity nodes. If the similarity between the dispute subject nodes is less than the second threshold of the dispute subject node similarity, the dispute subject nodes are judged to be low-similarity nodes.

[0147] In this embodiment, based on the statistics of a large number of similar social dispute urban social risk events, the dispute subject node similarities of the corresponding nodes are averaged and the first threshold of the dispute subject node similarity is set to 0.85. In addition, based on the statistics of a large number of completely unrelated social dispute urban social risk events, the dispute subject node similarities of the corresponding nodes are averaged and the second threshold of the dispute subject node similarity is set to 0.2.

[0148] If the similarity between the dispute content nodes is greater than the first threshold of the dispute content node similarity, the dispute content nodes are judged to be high-similarity nodes; if the similarity between the dispute content nodes is less than the second threshold of the dispute content node similarity, the dispute content nodes are judged to be low-similarity nodes;

[0149] In this embodiment, based on the statistics of a large number of similar social dispute urban social risk events, the dispute subject node similarities of the corresponding nodes are averaged and the first threshold of the dispute subject node similarity is set to 0.85. In addition, based on the statistics of a large number of completely unrelated social dispute urban social risk events, the dispute subject node similarities of the corresponding nodes are averaged and the second threshold of the dispute subject node similarity is set to 0.2.

[0150] Establish associated edges between high-similarity nodes, and unassociated edges between low-similarity nodes, and construct a social dispute cause-and-effect graph corresponding to the classification of urban social risk events.

[0151] In one embodiment of the present invention, unfinished social dispute city-level social risk events are collected and assigned to corresponding city-level social risk event categories, and evolution analysis and dispute prediction are performed on unfinished social dispute city-level social risk events based on the constructed social dispute reasoning map, including the following steps:

[0152] Collect the unfinished social dispute city-level social risk events, assign them to the corresponding city-level social risk event classification, and obtain the social dispute event map corresponding to the city-level social risk event classification;

[0153] An incomplete map of unfinished urban social risk events will be constructed for unfinished urban social risk events involving social disputes;

[0154] Analyze the similarity between dispute subject nodes and dispute content nodes in the hierarchical graph of unfinished social dispute city-wide social risk events and the corresponding hierarchical social dispute graph, and select nodes with similarity greater than a threshold as high-similarity nodes with corresponding nodes of unfinished social dispute city-wide social risk events;

[0155] Perform centrality analysis on the selected highly similar nodes, and select the main reference dispute subject node and the main reference dispute content node from the highly similar nodes;

[0156] For the main reference dispute subject node and the main reference dispute content node, perform information fusion with other corresponding highly similar nodes, and conduct evolutionary analysis and dispute prediction on the unfinished social dispute urban social risk events.

[0157] In one embodiment of the present invention, performing centrality analysis on the selected highly similar nodes and selecting the main reference dispute subject node and the main reference dispute content node from the highly similar nodes includes the following steps:

[0158] Count the number of edges connected to each of the selected highly similar nodes and the number of unconnected edges;

[0159] Perform centrality analysis on the selected highly similar nodes through the following formula:

[0160]

[0161] where Q is the centrality analysis value of the corresponding highly similar node, C0 is the number of edges connected to the corresponding highly similar node, C1 is the number of unconnected edges connected to the corresponding highly similar node, and S is the similarity of the corresponding highly similar node to the unfinished social dispute urban social risk event;

[0162] Select the dispute subject node and the dispute content node with the largest centrality analysis value from the highly similar nodes as the main reference dispute subject node and the main reference dispute content node.

[0163] In one embodiment of the present invention, for the main reference dispute subject node and the main reference dispute content node, performing information fusion with other corresponding highly similar nodes and conducting evolutionary analysis and dispute prediction on the unfinished social dispute urban social risk events includes the following steps:

[0164] Extract the content in the main reference dispute subject node and the main reference dispute content node, and extract the content of the nodes at the subsequent levels of the main reference dispute subject node and the main reference dispute content node, and use the extracted content as the main reference content;

[0165] Extract the content of other corresponding highly similar nodes, and extract the content of the dispute content nodes at the subsequent levels of the corresponding nodes, and summarize it through large language processing technology as the auxiliary reference content.

[0166] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for analyzing the evolution of urban social risk events and intelligent prediction based on the event graph, characterized in that: The following steps are involved: Collect information data on social disputes and urban social risk events, and clean the collected urban social risk event information data through natural language processing technology; Extract the dispute group attribute information, dispute category information and dispute content information of the social risk events in the city of social disputes, and classify the social risk events in the city with the same dispute group attribute information and dispute category information into the same category of social risk events in the city. The dispute content information includes: dispute time point, dispute area, focus of the social risk event in the city and the popularity of the social risk event in the city. According to the time point of the dispute, the city-wide social risk event development level is set up. In the city-wide social risk event development level, a dispute subject node is established, and the dispute area is marked on the node. The interrelated dispute subject nodes are linked to the dispute content nodes, including: the focus of the city-wide social risk event and the heat of the city-wide social risk event; In the classification of social risk events in each city, the similarity of each dispute subject node and dispute content node is evaluated, and associated edges are established between high-similarity nodes, and unassociated edges are established between low-similarity nodes. The attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map are constructed; The unfinished social dispute municipal social risk events are collected and assigned to the corresponding municipal social risk event categories. Based on the constructed social dispute cause-and-effect map, the unfinished social dispute municipal social risk events are analyzed for evolution and dispute prediction.

2. According to claim 1, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event map is characterized in that: According to the time point of the dispute, the city-wide social risk event development level is set, and in the city-wide social risk event development level, a dispute subject node is established, and the dispute area is marked on the node, and the interrelated dispute subject nodes are linked to the dispute content node, including the following steps: The development level of the city-wide social risk events is set up according to the time point of the dispute, including: basic facts layer, main city-wide social risk event layer, impact diffusion layer and mediation intervention layer; In each level, independent nodes are created for the dispute subject and the mediation subject as subject nodes. The subject node attributes include: dispute subject node and mediation subject node; By dividing the dispute area of ​​the dispute subject node into administrative regions, the dispute area is marked on the dispute subject node; Count the number of disputes involving the same type of dispute subject nodes of the subject node, and use the Reddit algorithm to count the popularity of social risk events in the city; Calculate the subject influence based on the number of disputes involving the subject nodes of the same type of disputes and the heat of urban social risk events, and mark the subject influence on the subject node; Establish dispute content nodes corresponding to the levels, and store the key points and popularity data of urban social risk events in the dispute content nodes; Extract and store the main information keywords in the key points of urban social risk events, calculate the cosine similarity between the main information keywords and the main nodes, and link the main nodes with cosine similarity greater than the preset threshold to the corresponding dispute content nodes.

3. According to claim 2, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event map is characterized in that: The subject influence is calculated based on the number of disputes involving the subject nodes of the same type of disputes and the heat of social risk events in the city, using the following formula: Among them, d in is the main influence, C in is the number of times the subject node of the same type of dispute as the subject node participates in the same type of urban social risk events, C0 is the number of times the subject node of the same type of dispute as the subject node participates in the dispute, R in R0 is the average heat value of the social risk events of the same type of dispute subject nodes as the subject node, and R1 is the average heat value of the social risk events of the same dispute category.

4. According to claim 1, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event graph is characterized in that: In the classification of social risk events in each city, the similarity of each dispute subject node and dispute content node is evaluated, including the following steps: Extract features from the dispute subject nodes, including: regional features, influence features, and social dispute relationship features; Among them, regional characteristics: collect the administrative division of dispute subject node standards and map them to the map; Influence characteristics: collect the influence values ​​of the subject marked by the dispute subject node; Social relationship characteristics: collect the dispute subject nodes of the city-level social risk events and the number of times they jointly participate in disputes in the same city-level social risk event classification; Analyze the regional similarity, subject influence similarity and social relationship feature similarity, and calculate the similarity between the dispute subject nodes based on the regional similarity, subject influence similarity and social relationship feature similarity; Extract features from dispute content nodes, including: key features of urban social risk events, urban social risk event heat features, and time decay features; Among them, the key features of urban social risk events: extracting key semantic information of dispute focus and appeal through TF-IDF or BERT embedding; Characteristics of the popularity of urban social risk events: Based on the amount of social media discussion, news reporting frequency, and the number of likes and reposts, the popularity of urban social risk events is calculated using the Reddit algorithm; Time decay feature: According to each level, the city-wide social risk events are divided into several time intervals. By calculating the difference in the heat value of the city-wide social risk events at the starting point and the end point of the time interval, the heat difference of each time interval is combined into a time decay feature array {A1, A2, ..., An}; Analyze the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events and the similarity of time decay features, and calculate the similarity between dispute content nodes based on the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events and the similarity of time decay features.

5. According to claim 4, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event map is characterized in that: Analyze the regional similarity, subject influence similarity and social relationship feature similarity, and calculate the similarity between the dispute subject nodes based on the regional similarity, subject influence similarity and social relationship feature similarity, including the following steps: By calculating the distance between the dispute subject nodes and counting the average distance between the dispute subject nodes in the same city-wide social risk event classification, the regional similarity Dlo is calculated as the distance between the dispute subject nodes / the average distance between the dispute subject nodes in the same city-wide social risk event classification; By calculating the difference between the subject influence values ​​marked on the dispute subject nodes and counting the average of the subject influence difference between the dispute subject nodes in the same city-wide social risk event classification, the subject influence similarity Din is calculated = the difference between the subject influence values ​​ / the average of the subject influence difference between the dispute subject nodes in the same city-wide social risk event classification; By counting the number of times a single dispute subject node participates in disputes in the same city-wide social risk event classification, the social relationship feature similarity Dso is calculated = the number of times they jointly participate in disputes in the same city-wide social risk event classification / the sum of the number of times each dispute subject node participates in disputes in the same city-wide social risk event classification; Based on the regional similarity, subject influence similarity, and social relationship feature similarity, the similarity between the dispute subject nodes is calculated using the following formula: Spa=w1*Dlo+w2*Din+w3*Dso Among them, Spa is the similarity between the dispute subject nodes, w1, w2 and w3 are the weights corresponding to the regional similarity, subject influence similarity and social relationship characteristics similarity.

6. According to claim 4, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event map is characterized in that: Analyze the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events, and the similarity of time decay features, and calculate the similarity between dispute content nodes based on the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events, and the similarity of time decay features, including the following steps: By calculating the cosine similarity between the key semantic information extracted between the dispute content nodes and counting the average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification, the key feature similarity of the city-wide social risk event Dke is calculated = cosine similarity between the key semantic information extracted between the dispute content nodes / average cosine similarity between the key semantic information between the dispute content nodes in the same city-wide social risk event classification; By calculating the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes, and counting the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification, the similarity of the key characteristics of the city-wide social risk events Dpo is calculated = the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes / the average value of the difference between the heat characteristics of the city-wide social risk events between the dispute content nodes in the same city-wide social risk event classification; Count the time decay feature array between dispute content nodes, the difference between the heat differences of corresponding time intervals, and the average heat value of dispute content nodes, and calculate the time decay feature similarity Dti = (the difference between the heat differences of corresponding time intervals between dispute content nodes / the average heat value of dispute content nodes) / n; Based on the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events, and the similarity of time decay features, the similarity between dispute content nodes is calculated using the following formula: Sco=r1*Dke+r2*Dpo+r3*Dti Among them, Sco is the similarity between dispute content nodes, r1, r2 and r3 are the weights corresponding to the similarity of key features of urban social risk events, the similarity of heat features of urban social risk events and the similarity of time attenuation features.

7. According to claim 1, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event map is characterized in that: Establishing associated edges between high-similarity nodes and unassociated edges between low-similarity nodes, constructing attribute information of different dispute groups and dispute category information, and the corresponding social dispute reasoning map, including the following steps: The similarities between the dispute subject nodes and the dispute content nodes are counted respectively. If the similarity between the dispute subject nodes is greater than the first threshold of the dispute subject node similarity, the dispute subject nodes are judged to be high-similarity nodes. If the similarity between the dispute subject nodes is less than the second threshold of the dispute subject node similarity, the dispute subject nodes are judged to be low-similarity nodes. If the similarity between the dispute content nodes is greater than the first threshold of the dispute content node similarity, the dispute content nodes are judged to be high-similarity nodes; if the similarity between the dispute content nodes is less than the second threshold of the dispute content node similarity, the dispute content nodes are judged to be low-similarity nodes; Establish associated edges between high-similarity nodes, and unassociated edges between low-similarity nodes, and construct a social dispute cause-and-effect graph corresponding to the classification of urban social risk events.

8. According to claim 1, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event graph is characterized in that: The unfinished social dispute city-level social risk events are collected and assigned to the corresponding city-level social risk event classifications. Based on the constructed social dispute reasoning map, the unfinished social dispute city-level social risk events are subjected to evolutionary analysis and dispute prediction, including the following steps: Collect the unfinished social dispute city-level social risk events, assign them to the corresponding city-level social risk event classification, and obtain the social dispute event map corresponding to the city-level social risk event classification; An incomplete map of unfinished urban social risk events will be constructed for unfinished urban social risk events involving social disputes; Analyze the similarity between dispute subject nodes and dispute content nodes in the hierarchical graph of unfinished social dispute city-wide social risk events and the corresponding hierarchical social dispute graph, and select nodes with similarity greater than a threshold as high-similarity nodes with corresponding nodes of unfinished social dispute city-wide social risk events; Conduct centrality analysis on the selected high-similarity nodes, and select the main reference dispute subject node and the main reference dispute content node from the high-similarity nodes; The main reference dispute subject node and the main reference dispute content node are fused with other corresponding high-similarity nodes to conduct evolutionary analysis and dispute prediction on unfinished social dispute urban social risk events.

9. The method for analyzing the evolution of urban social risk events and intelligent prediction based on the event graph according to claim 8 is characterized in that: Performing centrality analysis on the selected high-similarity nodes and selecting the main reference dispute subject node and the main reference dispute content node from the high-similarity nodes includes the following steps: Count the number of edges connected to each high-similarity node that has been screened out, as well as the number of unrelated edges; The following formula is used to perform centrality analysis on the selected high-similarity nodes: Among them, Q is the centrality analysis value of the corresponding high-similarity node, C0 is the number of edges connected to the corresponding high-similarity node, C1 is the number of unrelated edges connected to the corresponding high-similarity node, and S is the similarity of the corresponding high-similarity node to the unfinished social dispute urban social risk events; The dispute subject node and dispute content node with the largest centrality analysis value are selected from the high similarity nodes as the main reference dispute subject node and the main reference dispute content node.

10. According to claim 8, a method for analyzing the evolution of urban social risk events and intelligent prediction based on the event graph is characterized in that: The main reference dispute subject node and the main reference dispute content node are fused with other corresponding high-similarity nodes to perform evolution analysis and dispute prediction on unfinished social dispute urban social risk events, including the following steps: Extracting the content in the main reference dispute subject node and the main reference dispute content node, and extracting the content of the nodes at the subsequent levels of the main reference dispute subject node and the main reference dispute content node, and using the extracted content as the main reference content; The contents of other corresponding high-similarity nodes are extracted, and the contents of dispute content nodes at subsequent levels of the corresponding nodes are extracted, summarized through big language processing technology, and used as auxiliary reference content.

Images