Regional safety evaluation method, device and equipment based on public opinion popularity diffusion trend analysis, medium and product

Abstract

The invention discloses a region safety evaluation method, device and equipment based on public opinion popularity diffusion trend analysis, a medium and a product, and relates to the field of artificial intelligence. The method comprises the steps that public opinion information on an internet platform is acquired, data cleaning, feature extraction and formatting processing are carried out, and formatted information features are extracted; performing classification and scoring based on the formatted information features of each piece of information to obtain a safety evaluation score of an index to which each piece of information belongs; clustering each piece of information on the basis of formatted information features, sorting the information according to a time sequence, and identifying a key event set; calculating an event trend index for each key event in the key event set; calculating the update quantity of the influence of each key event on the regional safety index based on the event trend index, and calculating the updated regional safety index; regional safety evaluation is carried out based on the regional safety indexes before and after updating and the event trend index, and the objectivity, timeliness and accuracy of regional safety evaluation can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment, medium and product for regional security evaluation based on analysis of the trend of public opinion diffusion. Background Art

[0002] Traditional regional safety assessment methods are limited to field survey data and have the following problems: on the one hand, the input information is single-angle, which cannot guarantee the objective records of field survey personnel; on the other hand, there is a lack of judgment on the impact of social public opinion trends on regional safety. The data collected by traditional surveys have a short time series, and it is difficult to incorporate the development trend of public opinion into decision-making considerations; on the other hand, data collection and processing are inefficient, and most of them are processed manually, which has timeliness problems, that is, decisions cannot keep up with changes. Summary of the invention

[0003] The purpose of this application is to provide a regional safety evaluation method, device, equipment, medium and product based on public opinion heat diffusion trend analysis to improve the objectivity, timeliness and accuracy of regional safety evaluation.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In the first aspect, the present application provides a regional security evaluation method based on public opinion heat diffusion trend analysis, including:

[0006] Obtaining public opinion information on the Internet platform; the public opinion information includes text information, video information, audio information and image information;

[0007] Perform data cleaning, feature extraction and formatting processing on public opinion information to extract formatted information features of each piece of information; the formatted information features of each piece of information include number, region, event, person, keyword, time and subject;

[0008] Classifying each piece of information into various indicators of a regional security index based on the formatted information characteristics of each piece of information; the various indicators of the regional security index include crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety and information security;

[0009] Score the formatted information features of each piece of information to obtain the security evaluation score of the indicator to which each piece of information belongs;

[0010] Cluster each piece of information based on its formatted information features and sort it in chronological order to identify a set of key events;

[0011] Calculating an event trend index for each key event in the key event set;

[0012] Based on the event trend index of each key event, calculate the updated amount of impact of each key event on the regional safety index and calculate the updated regional safety index;

[0013] Regional safety evaluation is conducted based on the regional safety index before and after the update and the event trend index of each key event.

[0014] Optionally, the data cleaning, feature extraction and formatting of the public opinion information to extract the formatted information features of each piece of information specifically includes:

[0015] Perform data cleaning on each piece of public opinion information to remove useless data and burr data, and then use feature engineering to extract text features, image features or audio features of each piece of information;

[0016] Based on the text features, image features or audio features of each piece of information, the region, event, person, keyword, time and theme contained in each piece of information are identified and expressed as a seven-tuple (i, a, e, p, k, t, s) as the formatted information feature of each piece of information; wherein i, a, e, p, k, t, s respectively represent the number, region, event, person, keyword, time and theme corresponding to each piece of information.

[0017] Optionally, the classification based on the formatted information features of each piece of information, and classifying each piece of information into various indicators of the regional safety index, specifically includes:

[0018] Based on the formatted information features of each piece of information, the classifier built using Transformer obtains the probability that each piece of information belongs to each indicator in the regional safety index, and determines the indicator corresponding to the maximum probability as the indicator to which the information belongs.

[0019] Optionally, scoring the formatted information features of each piece of information to obtain a security evaluation score of an indicator to which each piece of information belongs specifically includes:

[0020] Build an SVR scorer or use BERT to score the formatted information features of each message. The score value is in the range of [-5,5] to obtain the security evaluation score of the indicator to which each message belongs.

[0021] Optionally, calculating the event trend index for each key event in the key event set specifically includes:

[0022] For each key event in the key event set, use the formula Calculate its sentiment score S t ; where N t is the number of new public opinion information related to the key event within time t; si is the security evaluation score of the indicator to which the i-th information belongs within time t;

[0023] Using formula Calculate the event propagation rate R for each key event t ; Where T is the time window;

[0024] Using formula Calculate the public opinion volatility index V t ;

[0025] Using formula Calculate event entropy H t ; Where k1 is the first proportionality coefficient; P(E i ) is the different safety evaluation scores s within time t i The probability distribution of

[0026] Using formula Calculate the event temperature T t ; Where k2 is the second proportionality coefficient;

[0027] Using the STI formula t =κ*S t +(1-κ)*STI t-1 Calculate the statistical trend index STI at time t t ; where κ is the smoothing coefficient; STI t-1 is the statistical trend index at the previous time point t-1;

[0028] According to the sentiment score S of each key event t , event propagation rate R t , Public Opinion Fluctuation Index V t 、Event entropy H t , event temperature T t and the Statistical Trend Index STI t , using the formula ETI t =α*S t +β*R t +γ*V t +δ*H t +ε*T t +ζ*STI t Calculate the event trend index ETI of the key event t ;where α, β, γ, δ, ε, ζ are S respectively t , R t 、V t , H t 、T t and STI t The weight coefficient of .

[0029] Optionally, the step of calculating the update amount of the impact of each key event on the regional safety index based on the event trend index of each key event and calculating the updated regional safety index specifically includes:

[0030] Based on the diffusion-reaction equation summarized in physics, the development of public opinion is simulated, and the public opinion concentration C(x,t) of a key event n at time t is calculated;

[0031] Using formula Calculate the update amount of the impact of key event n on the regional safety index

[0032] Using formula Calculate the updated regional safety index in This is the regional safety index before the update.

[0033] In the second aspect, the present application provides a regional security evaluation device based on public opinion heat diffusion trend analysis, including:

[0034] A public opinion information acquisition module is used to acquire public opinion information on the Internet platform; the public opinion information includes text information, video information, audio information and image information;

[0035] The public opinion information processing module is used to perform data cleaning, feature extraction and formatting processing on the public opinion information, and extract the formatted information features of each piece of information; the formatted information features of each piece of information include number, region, event, person, keyword, time and theme;

[0036] A public opinion information classification module is used to classify each piece of information based on the formatted information characteristics, and classify each piece of information into various indicators of the regional security index; the various indicators of the regional security index include crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety and information security;

[0037] The public opinion information scoring module is used to score the formatted information features of each piece of information and obtain the security evaluation score of the indicator to which each piece of information belongs;

[0038] The public opinion information clustering module is used to cluster each piece of information based on its formatted information features and sort it in chronological order to identify a set of key events;

[0039] An event trend index calculation module, used to calculate the event trend index for each key event in the key event set;

[0040] A regional safety index updating module, used to calculate the update amount of the impact of each key event on the regional safety index based on the event trend index of each key event and calculate the updated regional safety index;

[0041] The regional safety evaluation module is used to perform regional safety evaluation based on the regional safety index before and after the update and the event trend index of each key event.

[0042] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the regional safety assessment method based on the analysis of the trend of public opinion heat diffusion.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the regional security assessment method based on the analysis of the trend of public opinion heat diffusion.

[0044] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the regional security assessment method based on the analysis of the trend of public opinion heat diffusion.

[0045] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0046] This application provides a regional security evaluation method, device, equipment, medium and product based on the analysis of the diffusion trend of public opinion heat, which combines the multimodal public opinion information on the Internet platform and realizes the comprehensive evaluation of regional security through deep learning technology. By introducing the event trend index, the key factors such as sentiment score, event propagation rate, public opinion fluctuation index, event entropy, event temperature and statistical trend index are effectively integrated to reflect the development trend of events in real time. At the same time, the diffusion-reaction equation is used to simulate the propagation of public opinion information, which provides a scientific basis for the update of the regional security index. The method of this application comprehensively evaluates regional security based on indicators such as crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety and information security, so that decision makers can adjust policies in time according to public opinion dynamics to deal with potential security risks. The results show that the method proposed in this application has broad application prospects in the fields of regional security assessment, public opinion monitoring and intelligent decision-making, and provides strong support for the sustainable development of society. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 This is a flowchart of a regional security evaluation method based on public opinion heat diffusion trend analysis for this application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] In response to the defects in the prior art, this application proposes a regional safety evaluation method, device, equipment, medium and product based on the analysis of the trend of public opinion diffusion, so as to improve the objectivity, timeliness and accuracy of regional safety evaluation.

[0051] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0052] In an exemplary embodiment, Figure 1 As shown, a regional security evaluation method based on public opinion heat diffusion trend analysis is provided, including the following steps 1 to 8.

[0053] Step 1: Obtain public opinion information on the Internet platform; the public opinion information includes text information, video information, audio information and image information.

[0054] The Internet platform can be a platform for netizens to express their opinions, such as regional forums, social media, question-and-answer platforms, vertical communities in various industries, video platforms, and online evaluation platforms. The acquisition of public opinion information needs to be carried out in compliance with the relevant data protection laws and policies of the country where it is located, and with the authorization of the corresponding device owner. Specifically, it is possible to crawl or apply to the official API to obtain panel data (text information, audio information, video information, image information, etc.) of a certain time series from mainstream social media, regional forums, video platforms, and online evaluation platforms, so as to obtain public opinion information on various Internet platforms, also known as public opinion data.

[0055] Step 2: Perform data cleaning, feature extraction and formatting on the public opinion information to extract the formatted information features of each piece of information.

[0056] The acquired public opinion data is cleaned, feature extracted and formatted, useless data and burr data are cleaned, and then the key attribute features of the data (including region, event, person, keyword, time, theme, etc.) are extracted using feature engineering. Among them, text information (or data) can use TF-IDF vector or NLP (natural language processing) models such as BERT to extract text features, image information can use CNN (convolutional neural network) models such as ResNet or VGG to extract image features, audio information can use audio processing methods such as Mel-Spectrogram or MFCC, combined with RNN (recurrent neural network) or CNN to extract audio features, and video data is divided into frame-by-frame images and audio and processed using a two-stream model to extract its image and audio features. Furthermore, deep learning technology is used to extract key attribute features such as the theme, time, place (region), person, keyword and specific event involved in each piece of information from text features, image features or audio features, and finally the extracted key attribute feature data is formatted into the following computer-recognizable form as the formatted information feature of each piece of information.

[0057] Let B be a set of data structures, where each element b can be represented as a seven-tuple (i, a, e, p, k, t, s), where i represents the number and can be an element of the integer set Z. a represents the area and can be an element of the area set A. e represents the event and can be an element of the event set E. p represents the person and can be an element of the person set P. k represents the keyword and can be an element of the keyword set K. t represents the time and can be an element of the time set T, usually expressed in date and time format. s represents the subject and can be an element of the subject set S. Therefore, the data structure B of the formatted information characteristics of each piece of information can be represented as:

[0058] B={(i,a,e,p,k,t,s)∣i∈Z,a∈A,e∈E,p∈P,k∈K,t∈T,s∈S} (1)

[0059] Here, Z, A, E, P, K, T, and S are the collections of number, region, event, person, keyword, time, and subject, respectively. No matter what the media format of each piece of information is, the key attribute features of each piece of information are obtained through feature extraction and formatted into a unified format for subsequent processing.

[0060] Step 3: Classify each piece of information into various indicators of the regional safety index based on the formatted information characteristics of each piece of information.

[0061] The various indicators of the Regional Security Index (RSI) include crime rate (CR), public security (PS), health security (HS), economic security (ES), ecological security (EC), political security (POL), social security (SS), traffic safety (TS) and information security (IS). The calculation method of the Regional Security Index RSI is as follows:

[0062] RSI=w_1*CR+w_2*PS+w_3*HS+w_4*ES+w_5*EC+w_6*POL+w_7*SS+w_8*TS+w_9*IS(2)

[0063] Where RSI is the value of the regional security index. CR, PS, HS, ES, EC, POL, SS, TS, IS are the values ​​of the nine indicators of crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety, and information security. W_m (m = 1, 2, 3, 4, 5, 6, 7, 8, 9) represents the corresponding weight coefficients. Substituting the initial values ​​of CR, PS, HS, ES, EC, POL, SS, TS, IS and the corresponding weight coefficients of a region into formula (2), the initial regional security index of the region can be calculated. The RSI index of a region is used to assess the safety of the region.

[0064] Based on the formatted information features of each piece of information, the classifier built with Transformer is used to obtain the probability that each piece of information belongs to each indicator in the regional safety index, and the indicator corresponding to the maximum probability is determined as the indicator to which the information belongs, that is, each piece of information is mapped to an independent variable (indicator) in the RSI.

[0065] Step 4: Score the formatted information features of each piece of information to obtain the security evaluation score of the indicator to which each piece of information belongs.

[0066] Construct an SVR scorer or use BERT to score the formatted information features (people, keywords, events, places, time, and topics) of each piece of information. The score is in the range of [-5,5]. Normalization can be performed on the basic score to obtain the safety evaluation score of the indicator to which each piece of information belongs. The safety evaluation score represents the degree of safety evaluation reflected by the information. Negative values ​​represent negative impacts, positive values ​​represent positive impacts, and the larger the absolute value, the greater the degree of impact.

[0067] Step 5: Cluster each piece of information based on its formatted information features and sort it in chronological order to identify a set of key events.

[0068] Each piece of information is clustered according to the formatted information features of each piece of information, and sorted in chronological order to facilitate the subsequent regional safety evaluation. Specifically, the keywords and events of each piece of information extracted previously can be clustered to obtain the events contained in all the information obtained, that is, each piece of information is based on what event the netizens expressed their opinions, etc. Then, according to the clustering results, each type of event is sorted by time, so that the development trend of the event can be obtained. The clustering set involving a large number of events or involving certain hot events in the clustering results is identified as a key event set. Subsequently, the trend index of each key event (that is, the development index of each category) is calculated for the key event set.

[0069] Step 6: Calculate the event trend index for each key event in the key event set.

[0070] The calculation of the Event Trend Index (ETI) of this application involves the following influencing factors.

[0071] Sentiment Score t : Reflects changes in public sentiment related to the event.

[0072] Event propagation rate (Spread Rate) R t : Measures the speed at which events spread across the Internet.

[0073] Volatility Index V t : Reflects the fluctuations in public opinion caused by the incident.

[0074] Event Entropy t : Measuring the complexity and uncertainty of the public opinion status of an event.

[0075] Event temperature (Temperature) T t : Reflects the “intensity” of public opinion on an incident.

[0076] Statistical Trend Index STI t : A trend change index calculated based on time series data, reflecting the changing trend of public opinion on an event.

[0077] Specifically, for each key event n in the key event set, its sentiment score S is calculated using the following formula: t :

[0078]

[0079] Where N t is the number of new public opinion information related to the key event within time t; s i is the security evaluation score of the indicator to which the i-th information belongs within time t.

[0080] The event propagation rate R of each key event is calculated using the following formula: t :

[0081]

[0082] Where T is the time window, such as one hour or one day.

[0083] The public opinion volatility index V is calculated using the following formula: t :

[0084]

[0085] Public Opinion Volatility Index V t is the safety evaluation score s within time t i The standard deviation of the event measures the fluctuation of public opinion sentiment caused by the event.

[0086] The event entropy H is calculated using the following formula: t :

[0087]

[0088] Where k1 is the first proportionality coefficient; P(E i ) is the different safety evaluation scores s within time t i (such as the probability distribution of positive (0,5], negative [-5,0), neutral 0).

[0089] The event temperature T is calculated using the following formula: t :

[0090]

[0091] Among them, k2 is the second proportional coefficient, which represents the relationship between the intensity of emotion and the activity of public opinion.

[0092] The statistical trend index STI at time t is calculated using the following formula: t :

[0093] STI t =κ*S t +(1-κ)*STI t-1 (8)

[0094] Where κ is the smoothing coefficient, 0<κ<1; STI t-1 It is the statistical trend index at the previous time point t-1.

[0095] According to the sentiment score S of each key event n t , event propagation rate R t , Public Opinion Fluctuation Index V t 、Event entropy H t , event temperature T t and the Statistical Trend Index STI t , the event trend index ETI of the key event n is calculated using the following formula: t :

[0096] ETI t =α*S t +β*R t +γ*V t +δ*H t +ε*T t +ζ*STI t (9)

[0097] Among them, α, β, γ, δ, ε and ζ are S respectively. t , R t 、V t , H t 、T t and STI t The weight coefficient indicates the influence of each factor on ETI.

[0098] The definition and calculation of the event trend index (ETI) of this application adopts multiple factors, including basic indicators obtained from statistics and indicator offsets obtained based on the temperature diffusion model in thermodynamics, which comprehensively reflects the public opinion dynamics of the event and its impact on regional security. Specifically, the method of this application combines the dynamic changes of public opinion data with the temperature concept in the thermodynamic model, and uses the event temperature T t The method combines statistics and thermodynamics models to quantify the heat and diffusion speed of public opinion about an event, thus comprehensively reflecting the public opinion dynamics of the event and its impact on regional security. This innovative method, which combines statistics and thermodynamics models, can more comprehensively describe the evolution of public opinion and its impact on regional security.

[0099] Specifically, if the event trend index ETI t >0, indicating that the public opinion of the event is developing positively, public sentiment is tending to be positive, or the speed of information dissemination is slowing down. t <0, indicating that the public opinion of the incident is developing in a negative direction, public sentiment is tending to be negative, or the speed of information dissemination is accelerating, which may trigger a crisis and requires attention.

[0100] Step 7: Based on the event trend index of each key event, calculate the updated amount of the impact of each key event on the regional safety index and calculate the updated regional safety index.

[0101] The score of each new piece of information is updated to the regional safety index. The diffusion-reaction equation summarized in physics is used as a reference to simulate the development (diffusion) of public opinion. The regional safety index update process is introduced as follows.

[0102] Let C(x,t) be the intensity of public opinion information at position x and time t, also known as the concentration or heat of public opinion; D be the diffusion coefficient, describing the speed of information propagation; R be the reaction rate, indicating the attenuation or reaction rate of information during propagation. Then:

[0103]

[0104] For the public opinion concentration (i.e., public opinion information intensity) C(x,t), its initial condition is defined as:

[0105] C(x,0)=S current *w (11)

[0106] Among them, S current is the score of the current event, i.e. the score s in step 4 i The average value calculated based on key events; w is the weight factor related to the influence of public opinion, usually determined by historical data.

[0107] The boundary conditions are:

[0108] C(0,t)=C(L,t)=0 (12)

[0109] Where L is the length of the region.

[0110] For the reaction rate, define:

[0111] R=R0*f network *f activity *f content (13)

[0112] Among them, R0 is the basic propagation speed, which is the initial speed calculated based on the network and time; f networkis the network structure factor, which reflects the impact of network connectivity on the propagation speed, and its value is between 0 and 1; activity is the user activity factor, which reflects the user's activity in the social network and has a value between 0 and 1; content It is the information content factor, which reflects the influence of information type on the propagation speed, and its value is between 0 and 1. An upper limit needs to be set for the reaction rate R, and the part exceeding it is regarded as the boundary value.

[0113] By solving the diffusion-reaction equation (10), we can obtain the public opinion concentration C(x, t) of a key event n at time t, and calculate the update amount:

[0114]

[0115] That is, it represents the update amount of the impact of key event n on the regional safety index.

[0116] Then we get the update formula:

[0117]

[0118] in The regional safety index after the update (current cycle or this year) is expressed as ∑ represents the sum of the updates of all key events n. The update here is to add the impact of various events on regional security in the current cycle (this year) to In the above data, we can obtain the regional safety evaluation of the current cycle (this year).

[0119] This application uses the diffusion-reaction equation (model) to simulate the diffusion process of public opinion dissemination, and dynamically updates the regional security index by solving the equation. The innovation of this model is that by combining the diffusion of the heat of public opinion information with the reaction process, a real-time dynamic simulation of the spread and impact of public opinion is achieved. Specifically, the diffusion-reaction equation takes into account the propagation characteristics of public opinion information in space and time (such as the diffusion coefficient D and the reaction rate R), and updates the regional security index (RSI) by calculating the solution of the equation. This method can accurately simulate the heat diffusion trend of public opinion and make regional security evaluation more timely and accurate.

[0120] Step 8: Conduct regional safety evaluation based on the regional safety index before and after the update and the event trend index of each key event.

[0121] First, the regional security index of the current cycle and the previous cycle is calculated. By comparing, we can get the direction of regional security changes in the current cycle. If regional security improves, we can maintain the current decision, or make corresponding decisions based on the events by checking the impact of each event in the current cycle on RSI; if regional security deteriorates, we can find out the approximate problem by checking the impact of each event in the current cycle on RSI, and further deal with the events that have a negative impact on regional security. For example, in year N, it is found that the overall regional security has deteriorated compared with last year. By comparing the change values ​​of each indicator this year and last year, it is found that the change values ​​of the PS (public security) part are large. Then check the events extracted by clustering in the PS (public security) part, and make corresponding decisions based on the scores in the events (emotional scores, ETI, etc.). For example, if a building in the area collapses, causing certain casualties, then corresponding processing can be carried out in combination with the data of field investigation, such as comforting nearby residents and placing the wounded.

[0122] This application combines multimodal public opinion information, including data from social media, forums, and video platforms, and uses deep learning technology to achieve a comprehensive assessment of regional security. The event trend index (ETI) is introduced to effectively integrate key indicators such as sentiment score, transmission rate, and volatility index to reflect the development trend of events in real time. At the same time, the diffusion-reaction equation is used to simulate the spread of public opinion information to provide a scientific basis for the update of the regional security index (RSI). The method of this application comprehensively evaluates factors such as crime rate, public security, and health safety, enabling decision makers to adjust policies in a timely manner according to public opinion dynamics and respond to potential security risks. The results show that the proposed method has broad application prospects in the fields of regional security assessment, public opinion monitoring, and intelligent decision-making, and provides strong support for the sustainable development of society.

[0123] A specific embodiment of the method of the present application is provided below, which includes the following process.

[0124] Suppose you want to evaluate the regional security situation in City F. Recently, a night robbery incident occurred in the city that attracted widespread public attention, and the government took related measures to strengthen public security management. This application will use public opinion data obtained from social media, regional forums, question-and-answer platforms, etc., and dynamically update the Regional Security Index (RSI) of City F based on the Event Trend Index (ETI) and the diffusion-reaction equation.

[0125] S1: Calculate the initial (before update) regional safety index

[0126] The weights of various indicators in the Regional Safety Index (RSI) of City F are as follows:

[0127] Crime rate (CR): 0.3;

[0128] Public security situation (PS): 0.4;

[0129] Health and safety (HS): 0.1;

[0130] Economic security (ES): 0.05;

[0131] Ecological safety (EC): 0.02;

[0132] Political security (POL): 0.01;

[0133] Social Security (SS): 0.01;

[0134] Traffic safety (TS): 0.005;

[0135] Information security (IS): 0.005.

[0136] The initial values ​​of each indicator are as follows:

[0137] Crime rate (CR): 0.8;

[0138] Public security situation (PS): 0.7;

[0139] Health safety (HS): 0.9;

[0140] Economic security (ES): 0.85;

[0141] Ecological safety (EC): 0.95;

[0142] Political security (POL): 0.9;

[0143] Social Security (SS): 0.9;

[0144] Traffic safety (TS): 0.95;

[0145] Information Security (IS): 0.9.

[0146] Substituting the above values ​​into formula (2), we get the initial calculation formula of RSI:

[0147]

[0148] S2: Obtain public opinion information on the Internet platform.

[0149] Relevant public opinion information of City F in the past week was obtained from the Internet platform, including:

[0150] Text message on a social media:

[0151] "There was a robbery in downtown F last night. It was really disturbing!"

[0152] "Public security has been strengthened. Everyone is talking about this and thinks it will get better."

[0153] Text message on a regional forum:

[0154] "I encountered a robbery while walking at night. I advise everyone to be careful."

[0155] Image information: photos of the scene uploaded by netizens, such as photos of police patrolling the scene of the incident.

[0156] Video information: Related video from local news reports showing the robbery and subsequent police actions.

[0157] S3: Data cleaning, feature extraction and formatting.

[0158] Data cleaning is performed to remove advertisements, spam, and irrelevant comments, and to ensure that the information in the data is meaningful, such as retaining only descriptions of real events.

[0159] For text information, key attribute features such as region, event, person, keyword, time, and topic are extracted and scored using the BERT model. For example, the safety evaluation score of the first comment is -0.7 (negative), and the safety evaluation score of the second comment is 0.4 (positive).

[0160] For image information, the ResNet model is used to extract image features and identify patrol police and the scene of the incident in the image. For video information, a two-stream model is used to extract image features frame by frame and analyze the event background in combination with audio features.

[0161] After extraction, the features of public opinion information are formatted into the data structure shown in formula (1): B = {(i, a, e, p, k, t, s) | i∈Z, a∈A, e∈E, p∈P, k∈K, t∈T, s∈S}. The formatted information features of each piece of information are represented as a seven-tuple, including event number, region, event, person, keyword, time and topic.

[0162] S4: Information classification and scoring.

[0163] The classifier built using Transformer is used to classify each piece of information into various indicators of RSI. Next, BERT is used to score the formatted information features of each piece of information. Assume that the following results are obtained:

[0164] Information 1 (negative): classified as public order situation (PS), score -4;

[0165] Information 2 (positive): classified as a public safety situation (PS), with a score of 2;

[0166] Information 3......

[0167] According to the safety evaluation score, the information is mapped to each independent variable of RSI.

[0168] S5: Clustering and sorting.

[0169] Cluster the extracted information. Suppose the following key events are identified:

[0170] Key event 1: The nighttime robbery incident caused a large fluctuation in public opinion and a strong emotional reaction from the public.

[0171] Key Event 2: The measures taken by the F city government to strengthen public security management received a lot of positive comments.

[0172] Then, with these two events as the center, similar incidents are clustered together and then sorted in chronological order to facilitate subsequent regional safety assessment.

[0173] S6: Calculate the event trend index.

[0174] Assume that the weight coefficients are: α = 0.3, β = 0.25, γ = 0.2, δ = 0.1, ε = 0.1, ζ = 0.05. For key event 1 and key event 2, calculate the event trend index ETI respectively t .

[0175] Among them, the sentiment score S t : The sentiment score of key event 1 is -0.7. The sentiment score of key event 2 is 0.4.

[0176] Event propagation rate R t :

[0177] Set the time window to one day, and the number of new public opinion information about key event 1 is N t is 200 (social media, forum comments, etc.), so the spread rate of event 1 is After normalization, it is 0.2.

[0178] The number of new public opinion information about key event 2 N t is 80, so the propagation rate of event 2 is After normalization, it is 0.08.

[0179] Public Opinion Volatility Index V t The standard deviation is calculated based on the sentiment score. Assume that the standard deviation of key event 1 is 0.4 and the standard deviation of key event 2 is 0.2.

[0180] Event entropy H t :

[0181] For critical event 1, assume that its safety evaluation score is s iThe distribution is positive, negative, and neutral with proportions of 0.1, 0.8, and 0.1 respectively. Let k1 = -1, so the event entropy is:

[0182] H t =-∑P(E i )logP(E i )=-[0.1log(0.1)+0.8log(0.8)+0.1log(0.1)]≈0.32

[0183] For critical event 2, assume that its safety evaluation score is s i The distribution is positive, negative, and neutral with proportions of 0.7, 0.1, and 0.2 respectively, so the event entropy is:

[0184] H t =-∑P(E i )logP(E i )=-[0.7log(0.7)+0.1log(0.1)+0.2log(0.2)]≈0.48

[0185] Event temperature T t : The calculated event temperature of key event 1 is 0.7, and the event temperature of key event 2 is 0.4.

[0186] Calculate ETI:

[0187] For key event 1 (recorded as "robbery"), we have:

[0188] ETI 抢劫 =0.3×(-0.7)+0.25×0.2+0.2×0.4+0.1×0.32+0.1×0.7+0.05×1=0.072

[0189] For key event 2 (referred to as “Public Security”), we have:

[0190] ETI 治安 =0.3×0.4+0.25×0.08+0.2×0.2+0.1×0.48+0.1×0.4+0.05×1=0.318

[0191] S7: RSI update process.

[0192] The diffusion-reaction equation (10) is used to simulate the impact of public opinion, assuming that the diffusion coefficient D = 0.5 and the reaction rate R = 0.3 (the limit rate is set to 1000, f network 、f activity 、f content Here, it is considered as 1), and the initial public opinion concentration C(0,0) = -0.7. Solve to get the impact of the robbery incident C 抢劫=-0.5; Impact of security measures C 治安 =+0.2.

[0193] Then the impact update of the robbery event is:

[0194]

[0195] Impact of security measures Update volume:

[0196]

[0197] Calculate S1 to get the result Substituting the update amount of key events 1 and 2 into formula (15), we have:

[0198]

[0199] S8: Conduct regional safety assessment.

[0200] After considering the impact of the two key events, the Regional Security Index (RSI) of City F was updated from the initial value of 0.69875 to 0.72635. This result shows that although the robbery incident had a negative impact on the Regional Security Index (decreased the RSI by 0.036), the positive impact of the public security measures (increased the RSI by 0.0636) was more significant. Therefore, the overall regional security index has improved, indicating that the public security measures have effectively enhanced the public's sense of security, offset the negative impact of the robbery incident, and ultimately slightly improved the RSI.

[0201] In an exemplary embodiment, the present application also proposes a regional security evaluation device based on the analysis of the trend of the diffusion of public opinion heat, including the following functional modules. A public opinion information acquisition module, used to acquire public opinion information on the Internet platform; the public opinion information includes text information, video information, audio information and image information; a public opinion information processing module, used to perform data cleaning, feature extraction and formatting processing on the public opinion information, and extract the formatted information features of each piece of information; the formatted information features of each piece of information include number, region, event, person, keyword, time and subject; a public opinion information classification module, used to classify based on the formatted information features of each piece of information, and classify each piece of information into various indicators of the regional security index; the various indicators of the regional security index include crime rate, public security situation, health security, economic security, ecological security, political security, social security, traffic safety and information security All; public opinion information scoring module, used to score the formatted information features of each piece of information, and obtain the safety evaluation score of the indicator to which each piece of information belongs; public opinion information clustering module, used to cluster each piece of information based on the formatted information features of each piece of information and sort them in chronological order, and identify a set of key events; event trend index calculation module, used to calculate the event trend index for each key event in the key event set; regional safety index update module, used to calculate the update amount of the impact of each key event on the regional safety index based on the event trend index of each key event and calculate the updated regional safety index; regional safety evaluation module, used to perform regional safety evaluation based on the regional safety index before and after the update and the event trend index of each key event.

[0202] In an exemplary embodiment, the present application also provides a computer device, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface, and a communication interface. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the regional safety evaluation method based on the analysis of the trend of the diffusion of public opinion heat is implemented.

[0203] In an exemplary embodiment, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the regional safety assessment method based on public opinion heat diffusion trend analysis.

[0204] In an exemplary embodiment, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the regional security assessment method based on public opinion heat diffusion trend analysis.

[0205] It can be understood by a person of ordinary skill in the art that all or part of the processes in the above-mentioned embodiment method can be completed by hardware related to computer program instructions, and the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can include the process of the embodiment of the above-mentioned method when executed. Among them, any reference to memory or other media in each embodiment provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0206] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0207] This application obtains public opinion data from platforms where netizens express their opinions, such as regional forums, social media, question-and-answer platforms, vertical communities in various industries, video platforms, and online evaluation platforms. After cleaning the data, deep learning technology is used to extract data features from the data, including people, keywords, events, locations, time, topics, etc. The processed data is then evaluated by region, mode, and event to obtain the regional security score and the impact of each event. This can help users have a basic understanding of regional security, and optimize regional security and support decision-making in other areas, including but not limited to economic and ecological aspects. It can be used in scenarios such as regional security assessment, regional development decision-making, and AI-mediated decision-making.

[0208] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0209] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A regional safety evaluation method based on public opinion heat diffusion trend analysis, characterized in that: include: Obtain public opinion information on Internet platforms; The public opinion information includes text information, video information, audio information and image information; Perform data cleaning, feature extraction and formatting processing on public opinion information to extract formatted information features of each piece of information; the formatted information features of each piece of information include number, region, event, person, keyword, time and subject; Classifying each piece of information into various indicators of a regional security index based on the formatted information characteristics of each piece of information; the various indicators of the regional security index include crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety and information security; Score the formatted information features of each piece of information to obtain the security evaluation score of the indicator to which each piece of information belongs; Cluster each piece of information based on its formatted information features and sort it in chronological order to identify a set of key events; Calculating an event trend index for each key event in the key event set; Based on the event trend index of each key event, calculate the updated amount of impact of each key event on the regional safety index and calculate the updated regional safety index; Regional safety evaluation is conducted based on the regional safety index before and after the update and the event trend index of each key event.

2. The regional safety evaluation method based on public opinion heat diffusion trend analysis according to claim 1 is characterized in that: The data cleaning, feature extraction and formatting of the public opinion information are performed to extract the formatted information features of each piece of information, specifically including: Perform data cleaning on each piece of public opinion information to remove useless data and burr data, and then use feature engineering to extract text features, image features or audio features of each piece of information; Based on the text features, image features or audio features of each piece of information, the region, event, person, keyword, time and theme contained in each piece of information are identified and expressed as a seven-tuple (i, a, e, p, k, t, s) as the formatted information feature of each piece of information; wherein i, a, e, p, k, t, s respectively represent the number, region, event, person, keyword, time and theme corresponding to each piece of information.

3. According to claim 1, the regional safety evaluation method based on public opinion heat diffusion trend analysis is characterized in that: The classification is performed based on the formatted information features of each piece of information, and each piece of information is classified into various indicators of the regional safety index, specifically including: Based on the formatted information features of each piece of information, the classifier built using Transformer obtains the probability that each piece of information belongs to each indicator in the regional safety index, and determines the indicator corresponding to the maximum probability as the indicator to which the information belongs.

4. The regional safety evaluation method based on public opinion heat diffusion trend analysis according to claim 1 is characterized in that: Scoring the formatted information features of each piece of information to obtain the security evaluation score of the indicator to which each piece of information belongs specifically includes: Build an SVR scorer or use BERT to score the formatted information features of each message. The score value is in the range of [-5,5] to obtain the security evaluation score of the indicator to which each message belongs.

5. According to claim 1, the regional safety evaluation method based on public opinion heat diffusion trend analysis is characterized in that: The calculating of the event trend index for each key event in the key event set specifically includes: For each key event in the key event set, use the formula Calculate its sentiment score S t ; where N t is the number of new public opinion information related to the key event within time t; s i is the security evaluation score of the indicator to which the i-th information belongs within time t; Using formula Calculate the event propagation rate R for each key event t ; Where T is the time window; Using formula Calculate the public opinion volatility index V t ; Using formula Calculate event entropy H t ; Where k1 is the first proportionality coefficient; P(E i ) is the different safety evaluation scores s within time t i The probability distribution of Using formula Calculate the event temperature T t ; Where k2 is the second proportionality coefficient; Using the STI formula t =κ*S t +(1-κ)*STI t-1 Calculate the statistical trend index STI at time t t ; where κ is the smoothing coefficient; STI t-1 is the statistical trend index at the previous time point t-1; According to the sentiment score S of each key event t , event propagation rate R t , Public Opinion Fluctuation Index V t 、Event entropy H t , event temperature T t and the Statistical Trend Index STI t , using the formula ETI t =α*S t +β*R t +γ*V t +δ*H t +ε*T t +ζ*STI t Calculate the event trend index ETI of the key event t ;where α, β, γ, δ, ε, ζ are S respectively t , R t 、V t , H t , T t and STI t The weight coefficient of .

6. The regional safety evaluation method based on public opinion heat diffusion trend analysis according to claim 5 is characterized in that: The method of calculating the updated amount of the impact of each key event on the regional safety index based on the event trend index of each key event and calculating the updated regional safety index specifically includes: Based on the diffusion-reaction equation summarized in physics, the development of public opinion is simulated, and the public opinion concentration C(x,t) of a key event n at time t is calculated; Using formula Calculate the update amount of the impact of key event n on the regional safety index Using formula Calculate the updated regional safety index in This is the regional safety index before the update.

7. A regional safety evaluation device based on public opinion heat diffusion trend analysis, characterized in that: include: The public opinion information acquisition module is used to obtain public opinion information on the Internet platform; The public opinion information includes text information, video information, audio information and image information; The public opinion information processing module is used to perform data cleaning, feature extraction and formatting processing on the public opinion information, and extract the formatted information features of each piece of information; the formatted information features of each piece of information include number, region, event, person, keyword, time and theme; A public opinion information classification module is used to classify each piece of information based on the formatted information characteristics, and classify each piece of information into various indicators of the regional security index; the various indicators of the regional security index include crime rate, public security, health security, economic security, ecological security, political security, social security, traffic safety and information security; The public opinion information scoring module is used to score the formatted information features of each piece of information and obtain the security evaluation score of the indicator to which each piece of information belongs; The public opinion information clustering module is used to cluster each piece of information based on its formatted information features and sort it in chronological order to identify a set of key events; An event trend index calculation module, used to calculate the event trend index for each key event in the key event set; A regional safety index updating module, used to calculate the update amount of the impact of each key event on the regional safety index based on the event trend index of each key event and calculate the updated regional safety index; The regional safety evaluation module is used to perform regional safety evaluation based on the regional safety index before and after the update and the event trend index of each key event.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the regional safety assessment method based on public opinion heat diffusion trend analysis as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for evaluating regional safety based on analysis of the trend of diffusion of public opinion heat is implemented as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for evaluating regional safety based on analysis of the trend of diffusion of public opinion heat is implemented as described in any one of claims 1 to 6.

Images