A social network emergency perception method based on dynamic perception maximization

Abstract

The application discloses a social network emergency perception method based on dynamic perception maximization, which comprises the following steps: 1, searching for historical emergencies and the data of publishers, publishing time and the like related to the emergencies in social media; 2, defining the time of emergency occurrence, the time of user participation in the emergency and the ability index of user perception of the emergency; 3, using the dynamic perception ability distribution sequence of the user, training different sets of perceptrons at different time; 4, acquiring the information propagated in the selected set of perceptrons at different time, detecting the hot time by using the moving average method, and if the information amount related to a hot spot is greater than or equal to a threshold value, the hot spot is considered to develop into an emergency. The application regards the user as a perceiver, selects different sets of perceptrons according to the historical behavior data of the user at different time, monitors different sets of perceptrons at different time, and monitors the information propagated in the social media in real time, so as to achieve the purpose of predicting network emergencies.

Description

Technical Field

[0001] This invention relates to the field of data mining, specifically to a method for perceiving sudden events in social networks based on dynamic perception maximization. Background Technology

[0002] In recent years, social media platforms, represented by Weibo, have gradually become an important channel for people to obtain information. However, due to the diverse sources of information, the suddenness of events, and the wide dissemination of information within social networks, the real-time and accurate perception of emergencies has become a significant social issue.

[0003] From a methodological perspective, current static perceptron selection strategies for detecting sudden trending events have significant limitations. They consider users' perception capabilities over a long period, resulting in a set of perceptrons that is far from meeting the demands of rapidly changing social networks for event prediction. This approach ignores the dynamic randomness of user behavior and the repetitive nature of perception strategies, leading to incomplete information coverage, untimely perception, and low perception efficiency. Summary of the Invention

[0004] This invention addresses the shortcomings of existing technologies by proposing a social network emergency perception method based on dynamic perception maximization. This method aims to monitor information disseminated on social media in real time, thereby achieving effective prediction and dynamic perception of network emergencies and meeting the requirements for timeliness, accuracy, and reliability in emergency prediction.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The present invention provides a method for detecting sudden events in social networks based on dynamic perception maximization, characterized by the following steps:

[0007] Step 1: Obtain the set of historical sudden events E = {e1, e2, ..., e} k ,…,e K}, where e k Let k represent the k-th historical emergency, k∈[1,K], where K represents the number of events in the set of historical emergencies; retrieve all users who have published information related to historical emergencies in the set E, and form a candidate user set U={u1,u2,…,u…} i ,…,u I}, where u i Let i represent the i-th user, i∈[1,I], where I represents the number of users in the candidate user set;

[0008] Step 2, constructing the evaluation index of user perception ability:

[0009] Step 2.1, defining the time sequence T = {t1, t2, …, tn, …, tN}, tn represents the nth time period, n ∈ [1, N], N represents the number of time periods; n ,…,t N}, tn represents the nth time period, n ∈ [1, N], N represents the number of time periods; n

[0010] Step 2.2, obtaining all users who published the kth historical emergency event e k related information and their publishing time from the candidate user set U, and selecting the earliest publishing time as the occurrence time of the kth historical emergency event e k ;

[0011] If the ith user u i published the kth historical emergency event e n related information in the nth time period t k , define the difference between the time when the ith user u i published the kth historical emergency event e n related information in the nth time period t k and the occurrence time of the kth historical emergency event e k as the time when the ith user u i participated in the kth historical emergency event e k , denoted as

[0012] If the ith user u i has not published the kth historical emergency event e k related information, record the publishing time of the ith user u i to the kth historical emergency event e k related information as T max ;

[0013] The perception time of the ith user u i to the kth historical emergency event e n in the nth time period t k is denoted as

[0014] Step 2.3, the number of historical emergency events related to the information published by the ith user u i in the nth time period t n is denoted as The evaluation index of the perception ability of the ith user u i to the historical emergency event set E in the nth time period t n is denoted as ​

[0015] Step 3, constructing a distribution sequence of user dynamic perception ability:

[0016] Step 3.1, recording all users in the candidate user set U who published the related information of the historical emergency events in the historical emergency event set E in a historical time M, the publishing time, the historical emergency events published by the user in the historical time M and the number of the historical emergency events, the perception time of the user to the historical emergency events in the historical time M, thereby obtaining the information perception sequence of the i-th user u i wherein, represents the evaluation index of the perception ability of the i-th user u i to the historical emergency event set E in the n-th time period t n of the m-th historical period;

[0017] Step 3.2, performing singular spectrum analysis on the information perception sequence R i of the i-th user u i , and obtaining the denoised information perception sequence of the i-th user u i , denoted as S wherein, represents the evaluation index of the denoised perception ability of the i-th user u i to the historical emergency event set E in the n-th time period t n of the m-th historical period;

[0018] Step 3.3, taking the time t1~t N as the horizontal axis, taking the evaluation index value of the perception ability as the vertical axis, performing fitting on the denoised information perception sequence S i of the i-th user u i by using the least square method, and obtaining the dynamic information perception ability distribution function of the i-th user u i , wherein the evaluation index value corresponding to the n-th time period t n on the dynamic information perception ability distribution function is denoted as then the dynamic information perception ability distribution sequence of the i-th user u i in the historical time M is denoted as

[0019] Step 4, training N sets of perceiver set:

[0020] Step 4.1, selecting all users satisfying the dynamic social perception maximization model shown in formula (1) and formula (2) from the candidate user set U, and constituting N sets of perceiver set:

[0021]

[0022] ​

[0023] In equations (1) and (2), A n Indicates the nth time period t n The selected set of users that satisfy equations (1) and (2), y(A n ) represents the nth time period t n The set of perceptrons A n The sum of perception capabilities for a set of historical emergencies E, where f is the cost function and f(u) represents the selection of the i-th user u. i Place into the perceptron set A n The cost value, where b represents the upper limit of the cost that can be tolerated;

[0024] Step 4.2, Solving the dynamic social perception maximization model:

[0025] Step 4.2.1: Let the i-th user u be a descriptor. i A list of perceived historical emergencies N(u) i )={e i,1 ,...,e i,a}, where e i,a Represents the i-th user u i The perceived a-th historical event, where 'a' represents the i-th user u. i The total number of perceived historical emergencies, a∈[1,K];

[0026] Step 4.2.2: Calculate the i-th user u using equation (3). i and the j-th user u j Similarity S(u) of perceived historical emergency event list i ,u j ):

[0027]

[0028] In equation (3), N(u) j ) represents the j-th user u j A list of perceived historical emergencies;

[0029] Step 4.2.3: Define δ as the set threshold parameter, and δ is between 0 and 1; if S(u i ,u j If )≥δ, then it means that the i-th user u i and the j-th user u j If there is high similarity between them, then it means that the i-th user u... i and the j-th user u j There is no high similarity between them, thus obtaining all the similarities with the i-th user u iA set of similar users B(u) consisting of users with high similarity i );

[0030] Step 4.2.4: Based on the dynamic information perception ability distribution sequence {Y1,...,Y} of all users in the candidate user set U, i ,...,Y I}, for the nth time period t n Corresponding evaluation index value Sort in descending order and select the user who ranks first. Place in the nth time period t n The set of perceptrons A n middle;

[0031] Step 4.2.5: Based on the user ranked first The corresponding set of similar users Remove users from the candidate user set U. and all that belong to For users, obtain the set of candidate users U' after deletion;

[0032] Step 4.2.6: Based on the dynamic information perception ability distribution sequence of all users in the deleted candidate user set U', for the nth time period t n The corresponding evaluation index values ​​are sorted in descending order, and the users ranked first and second are respectively denoted as... Allow users In the nth time period t of the current dynamic information perception capability distribution sequence n The corresponding evaluation index value is

[0033] Step 4.2.7: Recalculate users The distribution sequence of dynamic information perception capability over historical time M, and the user's... In the nth time period t n Corresponding evaluation index value

[0034] Step 4.2.4, if Then the user Place in the nth time period t n Perceptor A n Collection, and based on user The corresponding set of similar users Remove users from the candidate user set U' and all that belong to For users, after obtaining the set of candidate users after deletion, assign it to U' and return to step 4.2.6;

[0035] like The candidate user set U' after deletion is taken as the candidate user set U, and the step 4.2.4 is sequentially executed until formula (2) is not satisfied, thereby obtaining the nth time period t n the perceiver set A n of the nth time period t

[0036] Step 5, monitoring the emergency event;

[0037] Step 5.1, defining a hotspot set H = {h1, h2,..., h d ,..., h D}, wherein h d represents the Dth hotspot, d ∈ [1, D], and D represents the number of hotspots in the hotspot set;

[0038] Obtaining the information quantity x n of all users in the perceiver set A n of the nth time period t d publishing information related to the dth hotspot h n , thereby obtaining the information quantity sequence X = {x1,..., x d ,..., x n} of the N groups of perceivers about the dth hotspot h N ;

[0039] Step 5.2, judging whether formula (4) is established, if yes, the xth hotspot h x develops into an emergency event, otherwise, the xth hotspot h x does not develop into an emergency event:

[0040]

[0041] In formula (4), α represents a significance level, is a standard deviation estimated according to the information observation sequence, and is obtained by formula (6), represents the mean of the information observation sequence X, and is obtained by formula (5), β is a threshold value set, and z 1-α / 2 is a normal threshold value corresponding to the significance level α;

[0042]

[0043]

[0044] In formula (5)-formula (6), x j represents the jth information quantity in the partial information observation sequence {x1,..., x n}, j ∈ [1, n].

[0045] The electronic device comprises a memory and a processor, and is characterized in that the memory is used for storing a program supporting the processor to execute the social network emergency event perception method, and the processor is configured to execute the program stored in the memory.

[0046] The computer readable storage medium stores a computer program, and the computer program is configured to execute the steps of the social network emergency event perception method when executed by a processor.

[0047] Compared with the prior art, the application considers the historical behavior data of the user in a period of time when selecting the set of perceivers, and can obtain a better set of perceivers compared with the existing static perception method, thereby improving the prediction accuracy of the emergency event.

[0048] 1. The application uses the historical behavior data of the user to select different sets of perceivers at different times, thereby making up for the limitations of the existing method of selecting perceivers, avoiding the problem that some perceivers have low perception ability at some times, and being more effective in predicting the emergency event.

[0049] 2. The application integrates the similarity between users into the solution of the CELF algorithm when solving the dynamic social perception maximization model, thereby avoiding the redundancy of calculation and the poor effect of the selected perceivers caused by the repeated perception of the same historical emergency event by different users, and improving the efficiency of the solution and the quality of the selected perceivers. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The figure is a flowchart of the method of the application. DETAILED DESCRIPTION

[0051] In this embodiment, the social network emergency event perception method based on dynamic perception maximization is combined with microblogging, and is constructed according to the historical behavior data of the user, thereby describing the perception ability of the user to the emergency event at different times. The set of perceivers at different times is selected according to the perception ability of the user to the emergency event at different times. Specifically, as shown in the figure, the method is performed according to the following steps: Figure 1

[0052] Step 1, in the microblogging, a set of historical emergency events E = {e1, e2, …, en} is obtained, wherein en represents the n-th historical emergency event, and n is an integer greater than or equal to 1. k K k ​​​represents the kth historical event, k ∈ [1, K], K represents the number of events in the historical event set E; all users who have published information related to the historical event in the historical event set E are obtained, and a candidate user set U = {u1, u2, …, u i ,…,u I} is formed, where u i represents the ith user, i ∈ [1, I], I represents the number of users in the candidate user set; the historical event is defined in the form of a keyword;

[0053] Step 2, construct an evaluation index of user perception ability:

[0054] Step 2.1, define a time period sequence T = {t1, t2, …, t n ,…,t N}, t n represents the nth time period, n ∈ [1, N], N represents the number of time periods; the time period sequence must be able to reflect the periodicity of user behavior, for example, a day can be regarded as a time period sequence.

[0055] Step 2.2, obtain all users who have published information related to the kth historical event e k and their publishing time from the candidate user set U, and select the earliest publishing time as the occurrence time of the kth historical event e k .

[0056] If the ith user u i publishes information related to the kth historical event e n in the nth time period t k , then the difference between the time when the ith user u i publishes information related to the kth historical event e n in the nth time period t k and the occurrence time of the kth historical event e k is defined as the time when the ith user u i participates in the kth historical event e k , denoted as If the ith user u i publishes information related to the kth historical event e k multiple times, then the first time when the ith user u i publishes information related to the kth historical event e n in the nth time period t k is taken as the research object;

[0057] If the ith user u i has not published the kth historical event e kIf the relevant information is obtained, then record the i-th user u. i For the k-th historical emergency e k The relevant information was released at time T. max ; where T max It is a large number representing the i-th user u i No historical emergency event e has been released. k The penalty value imposed on the relevant information;

[0058] The i-th user u i In the nth time period t n For the k-th historical emergency e k The perception time is denoted as

[0059] Step 2.3: Connect with the i-th user u i In the nth time period t n The number of historical emergencies related to the information released is recorded as follows: Then the i-th user u i In the nth time period t n The evaluation index for the ability to perceive the set of historical emergencies E is denoted as .

[0060] Step 3: Construct the distribution sequence of user dynamic perception capabilities:

[0061] Step 3.1: Search on Weibo and record all users in the candidate user set U who have posted information related to historical emergencies in the historical emergency event set E within a historical time period M, along with their posting times, the historical emergencies posted by each user within the historical time period M and their number, and the user's perception time of the historical emergencies within the historical time period M. This will yield the i-th user u. i Information perception sequence in, Represents the i-th user u i In the nth time period t of the mth historical cycle n Evaluation indicators of the ability to perceive a set of historical emergencies E;

[0062] Step 3.2, for the i-th user u i Information perception sequence R i Perform singular spectrum analysis to obtain the denoised i-th user u i The information perception sequence, denoted as in, Represents the i-th user u i In the nth time period t of the mth historical cycle nEvaluation index of the perception capability of a set of historical emergencies E after denoising; singular spectrum analysis, removing subsequences with smaller singular values ​​and reconstructing the remaining subsequences with larger singular values;

[0063] Step 3.3, with t1 to t N Using time as the horizontal axis and the evaluation index of perception ability as the vertical axis, the least squares method is used to evaluate the denoised i-th user u. i Information perception sequence S i By performing a fitting operation, the i-th user u is obtained. i Let t be the distribution function of dynamic information perception capability, and let t be the nth time period on the distribution function of dynamic information perception capability. n The corresponding evaluation index value is Then the i-th user u i The distribution sequence of dynamic information perception capability over historical time M is denoted as... User u was obtained after singular spectrum analysis for noise reduction and least squares fitting. i The dynamic information perception ability distribution sequence over historical time M can truly reflect the user's information perception ability at different times.

[0064] Step 4: Train N sets of perceptrons:

[0065] Step 4.1: Select all users from the candidate user set U who satisfy the dynamic social perception maximization model shown in equations (1) and (2), and form N sets of perceptrons:

[0066]

[0067]

[0068] In equations (1) and (2), A n Indicates the nth time period t n The selected set of users that satisfy equations (1) and (2), y(A n ) represents the nth time period t n The set of perceptrons A n The sum of perception capabilities for a set of historical emergencies E, where f is the cost function and f(u) represents the selection of the i-th user u. i Place into the perceptron set A n The cost value, b represents the upper limit of the cost that can be tolerated; simply, we can consider f(u) as... i b ≡ 1, b = |A|, where |A| is the upper limit of the number of perceptrons in a set of perceptrons, that is, the upper limit of the number of perceptrons in a set of perceptrons is b;

[0069] Step 4.2, solving the dynamic social-aware maximization model (integrating the similarity between users into the solution of C ELF algorithm):

[0070] Step 4.2.1, recording the i-th user u i The list of historical events perceived by the i-th user u i i,1 ,...,e i,a}, where e i,a represents the a-th historical event perceived by the i-th user u i , and a represents the total number of historical events perceived by the i-th user u i , a∈[1,K];

[0071] Step 4.2.2, calculating the similarity S(u i ,u j ) between the list of historical events perceived by the i-th user u i and the j-th user u j using formula (3):

[0072]

[0073] In formula (3), N(u j ) represents the list of historical events perceived by the j-th user u j ;

[0074] Step 4.2.3, defining δ as a threshold parameter set, and δ is between 0 and 1; if S(u i ,u j )≥δ, it means that there is high similarity between the i-th user u i and the j-th user u j , otherwise, it means that there is no high similarity between the i-th user u i and the j-th user u j , thereby obtaining the similarity user set B(u i ) composed of all users with high similarity to the i-th user u i ;

[0075] Step 4.2.4, according to the dynamic information perception ability distribution sequence {Y1,...,Y i ,...,Y I} of all users in the candidate user set U, the evaluation index value of the n-th time period t n is sorted in descending order, and the user with the first ranking is put into the perceiver set A n of the n-th time period t n ; ​

[0076] Step 4.2.5, according to the first user in the order The corresponding similar user set Delete the user from the candidate user set U and all users belonging to , to get the deleted candidate user set U'; delete the user and all users belonging to from the candidate user set U, which can improve the probability of selecting high-quality sensors and computational efficiency;

[0077] Step 4.2.6, according to the dynamic information perception ability distribution sequence of all users in the deleted candidate user set U', the n th time period t n The corresponding evaluation index value is sorted in descending order, and the first and second users in the order are respectively Let the user The evaluation index value corresponding to the n th time period t n of the current dynamic information perception ability distribution sequence is

[0078] Step 4.2.7, recalculate the dynamic information perception ability distribution sequence of user in history time M, and get the evaluation index value corresponding to the n th time period t n

[0079] Step 4.2.4, if Put the user into the sensor set A n of the n th time period t n , and delete the user and all users belonging to from the candidate user set U' according to the similar user set corresponding to the user , get the deleted candidate user set U' after assignment, and return to step 4.2.6;

[0080] If , the deleted candidate user set U' is taken as the candidate user set U, and the step 4.2.4 is sequentially executed until the formula (2) is not satisfied, so as to get the sensor set A n of the n th time period t n , and further get N groups of sensor sets;

[0081] Step 5, monitor the emergency event;

[0082] Step 5.1, real-time acquisition of sensor set A n ​Information disseminated on Weibo is aggregated into different hot topics using a text similarity detection method. A hot topic set H = {h1, h2, ..., h...} is defined. d ,…,h D}, where h d Let d represent the Dth hotspot, where d∈[1,D] and D represents the number of hotspots in the hotspot set;

[0083] Get the nth time period t n The set of perceptrons A n All users in the d-th hot topic h d The amount of information in the relevant information is x n Thus, we obtain N sets of perceptrons regarding the d-th hotspot h. d The information sequence is X = {x1,...,x} n ,…,x N};

[0084] Step 5.2: Determine whether equation (4) holds true. If it does, then at the significance level α, the xth hotspot h can be considered as... x It will develop into a sudden event; otherwise, the xth hotspot h x It did not develop into a sudden incident:

[0085]

[0086] In equation (4), α refers to the significance level. The standard deviation is estimated based on the information observation sequence and obtained from equation (6). Let z represent the mean of the information observation sequence X, obtained from equation (5), where β is the set threshold, and z 1-α / 2 It is the normal threshold corresponding to the significance level α;

[0087]

[0088]

[0089] In equations (5)-(6), x j Represents a partially informative observation sequence {x1,...,x} n The j-th information in}, j∈[1,n].

[0090] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the aforementioned social network incident detection method. The processor is configured to execute the program stored in the memory.

[0091] In this embodiment, a computer readable storage medium stores a computer program, and the computer program is run by a processor to execute the steps of the social network burst perception method.

Claims

1. A method for detecting sudden events in social networks based on dynamic perception maximization, characterized in that, Includes the following steps: Step 1: Obtain a collection of historical emergencies ,in, Indicates the first k A sudden historical event, , K This indicates the number of events within the historical emergency event set; retrieve all historical emergency event sets that have been published. Users with information related to historical emergencies in China, forming a candidate user set. ,in, Indicates the first i One user, , I This represents the number of users within the candidate user set; Step 2: Construct evaluation metrics for user perception capabilities: Step 2.1: Define the time period sequence , This represents the nth time period. N represents the number of time periods; Step 2.2: From the candidate user set The first one published in the middle k A historical emergency All users of the relevant information and their posting times, selecting the earliest posting time as the [number]. k A historical emergency The time of occurrence; If the first i individual users In the nth time period Published the first k A historical emergency Relevant information, then define the first i individual users In the nth time period Published and the k A historical emergency The time and number of relevant information k A historical emergency The difference between the occurrence times is the first i individual users Participating in the k A historical emergency The time is recorded as ; If the first i individual users No first published k A historical emergency For relevant information, record the number. i individual users For the k A historical emergency The relevant information was released at the following time. ; The first i individual users In the nth time period For the k A historical emergency The perception time is denoted as ; Step 2.3, with the first i individual users In the nth time period The number of historical emergencies related to the information released is recorded as follows: Then the first i individual users In the nth time period Collection of historical emergencies The evaluation index for perception ability is denoted as . ; Step 3: Construct the distribution sequence of user dynamic perception capabilities: Step 3.1: Record all users in the candidate user set U who have published information related to historical emergencies in the historical emergency event set E within a historical time period M, along with their publication times, the historical emergencies published by each user within the historical time period M and their quantity, and the user's perception time of the historical emergencies within the historical time period M, thereby obtaining the first... i individual users Information perception sequence ,in, Indicates the first i individual users In the nth period of the mth historical cycle Collection of historical emergencies Evaluation indicators of perception ability; Step 3.2, for the first i individual users Information perception sequence Perform singular spectrum analysis to obtain the denoised first... i individual users The information perception sequence, denoted as ;in, Indicates the first i individual users In the nth period of the mth historical cycle Collection of historical emergencies Evaluation metrics for the perceived ability after noise reduction; Step 3.3, with Using time as the horizontal axis and the evaluation index of perception ability as the vertical axis, the least squares method is used to evaluate the denoised time... i individual users Information perception sequence By fitting, the first... i individual users Let the distribution function of dynamic information perception capability be denoted by the nth time period on the distribution function of dynamic information perception capability. The corresponding evaluation index value is Then the first i individual users The distribution sequence of dynamic information perception capability over historical time M is denoted as... ; Step 4: Train N sets of perceptrons: Step 4.1: From the candidate user set Select all users who satisfy the dynamic social perception maximization model shown in equations (1) and (2), and form N sets of perceivers: (1) (2) In equations (1) and (2), Indicates the nth time period The selected set of users that satisfy equations (1) and (2) Indicates the nth time period Perceptor set The sum of perception capabilities for a set E of historical emergencies, where f is the cost function. This indicates selecting the i-th user. Add to the perceptron set The cost value, where b represents the upper limit of the cost that can be tolerated; Step 4.2, Solving the dynamic social perception maximization model: Step 4.2.1: Record the i-th user List of perceived historical emergencies ,in, Represents the i-th user The a-th historical emergency perceived Represents the i-th user The total number of perceived historical emergencies ; Step 4.2.2: Calculate the i-th user using equation (3). and the j-th user Similarity of perceived historical emergency events : (3) In equation (3), Represents the j-th user A list of perceived historical emergencies; Step 4.2.3: Define δ as the set threshold parameter, and δ is between 0 and 1; if S Then it represents the i-th user. and the j-th user If there is high similarity between them, then it indicates that the i-th user... and the j-th user There is no high similarity between them, thus obtaining all the results related to the i-th user. A set of similar users that exhibit high similarity ; Step 4.2.4: Based on the candidate user set Distribution sequence of dynamic information perception capabilities of all users For the nth time period Corresponding evaluation index value Sort in descending order and select the user who ranks first. Place in the nth time period Perceptor set middle; Step 4.2.5: Based on the user ranked first The corresponding set of similar users From the candidate user set Delete user and all that belong to For users, obtain the set of candidate users after deletion. ; Step 4.2.6: Based on the deleted candidate user set The distribution sequence of dynamic information perception ability of all users in the nth time period. The corresponding evaluation index values ​​are sorted in descending order, and the users ranked first and second are respectively denoted as... , ; to make users In the nth time period of the current dynamic information perception capability distribution sequence The corresponding evaluation index value is ; Step 4.2.7: Recalculate users The distribution sequence of dynamic information perception capability over historical time M, and the user's... In the nth time period Corresponding evaluation index value ; Step 4.2.8, if Then the user Place in the nth time period Perceptor Collection, and based on user The corresponding set of similar users From the candidate user set Delete user and all that belong to After obtaining the set of candidate users after deletion, the user is assigned a value. Then return to step 4.2.6; like Then the set of candidate users after deletion will be... As a set of candidate users Then return to step 4.2.4 and execute sequentially until equation (2) is not satisfied, thus obtaining the nth time period. Perceptor set This leads to N sets of perceptrons; Step 5: Monitor for emergencies; Step 5.1: Define the hotspot set ,in, This represents the d-th hotspot. D represents the number of hotspots within the hotspot set; Get the nth time period Perceptor set All users posted in the dth hot topic The amount of information in the relevant information is Thus, we obtain N sets of perceptrons about the d-th hotspot. The information content sequence is ; Step 5.2: Determine whether equation (4) is true. If it is true, then the xth hotspot... It will develop into a sudden event; otherwise, the xth hotspot It did not develop into a sudden incident: (4) In equation (4), α refers to the significance level. The standard deviation is estimated based on the information observation sequence and obtained from equation (6). Represents information observation sequence The mean value is obtained from equation (5). For the set threshold, It is the normal threshold corresponding to the significance level α; (5) (6) In equations (5)-(6), Represents a partial information observation sequence The j-th piece of information in .

2. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the social network incident detection method of claim 1, wherein the processor is configured to execute the program stored in the memory.

3. A computer-readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to perform the steps of the social network outbreak perception method of claim 1.

Images