Intelligent early warning method and system for abnormal public opinion information

Abstract

The invention discloses an intelligent early warning method and system for abnormal public opinion information, and belongs to the technical field of electric data processing. The method comprises the following steps: constructing a network public opinion propagation core element anomaly perception model, and predicting a network public opinion propagation core element index value according to the core element anomaly perception model; obtaining actual measurement index values of network public opinion propagation core elements, further constructing an actual measurement index set H, and calculating index weights in the actual measurement index set to form a weight matrix; respectively judging the matching degree between the actual data of each index in the actual measurement index set H and each index in the predicted network public opinion propagation core element index value, if the matching degree is respectively greater than or equal to the threshold value of each index, determining that the actual data of each index is normal, otherwise, determining that the actual data of each index is abnormal, and constructing an abnormal actual measurement index set sequence; and calculating the network public opinion risk probability according to the abnormal index sequence. According to the method, the newborn risk can be monitored, the method does not depend on manual feature annotation, and the dynamically derived composite risk in public opinion propagation can be dealt with.

Description

Technical Field

[0001] The invention relates to an intelligent early warning method and system for abnormal public opinion information, belonging to the field of artificial intelligence technology. Background Art

[0002] Existing risk matching methods based on historically annotated data rely on building libraries of annotated risk features (such as keyword libraries and sentiment polarity rule libraries) and identifying known risk types (such as rumors and negative sentiment clusters) through pattern matching. Technical issues with these methods include: they can only identify annotated risks from historical cases and fail to monitor emerging risks (such as unannotated risk patterns in emergencies); and their reliance on manually annotated features makes it difficult to address the complex risks that arise dynamically during public opinion dissemination. Summary of the Invention

[0003] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide an intelligent early warning method and system for abnormal public opinion information, which can monitor new risks, does not rely on manual labeling features, and can respond to complex risks dynamically derived from the dissemination of public opinion.

[0004] It has improved the accuracy, timeliness and intelligence of public opinion risk identification, and provided reliable technical support for online public opinion management.

[0005] To achieve the above-mentioned purpose, the method for intelligent early warning of abnormal public opinion information includes the following steps: S1: Construct an abnormal perception model of the core elements of network public opinion dissemination, and predict the index values ​​of the core elements of network public opinion dissemination based on the core element abnormal perception model ; S2: Obtain the measured index values ​​of the core elements of online public opinion dissemination, filter the measured index values ​​to construct a measured index set H, and calculate the index weights in the measured index set to form a weight matrix W; S3: Determine the actual data of each indicator in the measured indicator set H and the predicted core element index value of network public opinion communication If the degree of fit of each indicator is greater than or equal to the threshold of each indicator, the actual data of each indicator is normal, otherwise it is abnormal, and a sequence of abnormal measured indicator sets is constructed. , K is a positive integer greater than or equal to 1; S4: According to the abnormal indicator sequence Calculate the probability of online public opinion risk: .

[0006] In order to achieve the purpose of the invention, the present invention also provides a system, which includes a storage medium and one or more processors, wherein the storage medium is used to store code that uses computer language to compile the above-mentioned abnormal public opinion information intelligent early warning method into a computer program, and the computer program can be called and executed by one or more processors.

[0007] Compared with the existing technology, the intelligent early warning method and system for abnormal public opinion information provided by the present invention realizes real-time detection and risk warning of public opinion data anomalies by constructing an abnormal perception model of the core elements of network public opinion dissemination, filters the monitoring data through the public opinion evolution model, and improves the intelligent perception ability of unlabeled public opinion risks. Therefore, it can monitor new risks without relying on manually labeled features, and can deal with complex risks dynamically derived in public opinion dissemination. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a flow chart of the intelligent early warning method for abnormal public opinion information provided by the present invention;

[0009] Figure 2 Schematic diagram of the abnormal monitoring range provided by the present invention. DETAILED DESCRIPTION

[0010] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0011] Figure 1 This is a flow chart of the intelligent early warning method for abnormal public opinion information provided by the present invention. Figure 1 As shown, the intelligent early warning method for abnormal public opinion information includes the following steps: S1: Construct an abnormal perception model of the core elements of network public opinion dissemination, and predict the index values ​​of the core elements of network public opinion dissemination based on the core element abnormal perception model ; S2: Obtain the measured index values ​​of the core elements of online public opinion dissemination, filter the measured index values ​​to construct a measured index set H, and calculate the index weights in the measured index set to form a weight matrix W; S3: Determine the actual data of each indicator in the measured indicator set H and the predicted core element index value of network public opinion communication If the degree of fit of each indicator is greater than or equal to the threshold of each indicator, the actual data of each indicator is normal, otherwise it is abnormal, and a sequence of abnormal measured indicator sets is constructed. , K is a positive integer greater than or equal to 1; S4: According to the abnormal indicator sequence Calculate the probability of online public opinion risk: .

[0012] Issue early warning based on the probability value of online public opinion risk.

[0013] In the present invention, the core element abnormal perception model includes: public opinion subject symbiosis model, public opinion information dissemination model and netizen emotion evolution model, which respectively predicts the number of ordinary netizens based on the public opinion subject symbiosis model, public opinion information dissemination model and netizen emotion evolution model. , public opinion vector and netizens' emotional vectors ,Right now ,in, , Respectively represent the predicted number of public opinion information releases, forwarding numbers, and comment data; , Represents the predicted number of public opinion information releases The number of neutral emotions, negative emotions, and positive emotions in .

[0014] In the intelligent early warning method for abnormal public opinion information provided by the present invention, obtaining the measured index values ​​of the core elements of network public opinion dissemination includes: S2-1: The number of ordinary netizens measured from public opinion monitoring data , public opinion vector and netizens' emotion vectors , Respectively represent the measured number of information releases, forwarding numbers, and comment data, They represent the number of neutral emotions, negative emotions, and positive emotions in the measured number of public opinion information releases M1; S2-2: Obtain the modeled data sequence by accumulating the information of the measured data a before time n ;Build a public opinion evolution model based on the modeling data sequence to predict the predicted value after N moments to further obtain the evolution trend, where a represents any of the following data: S, 、 ; S2-3: Modeling Data Sequences Train the Logistic regression model into a public opinion evolution model.

[0015] In the present invention, the Logistic regression model is: , Where, is the intrinsic growth rate, K is the information limit; is the amount of public opinion information about a social event at the initial moment, and n represents time.

[0016] The Logistic regression model training process includes: S2-3-1: Establish the loss function value according to the following formula:

[0017] Update K according to the following formula: , Where, is the adjustment coefficient, is the gradient of K; S2-3-2: Update r according to the following formula: , Where, is the adjustment coefficient, is the gradient of r; S2-3-3: Determine whether the loss function value L is the smallest. If so, output the optimal K and r respectively. and , and then execute step S2-4; if the loss function value L does not reach the minimum, the changed K and r are replaced by the values ​​before the change, and then repeat steps S2-3-1 to S2-3-3; S2-3-4: Predict the public opinion risk trend for a period of time in the future based on the following formula: .

[0018] The present invention trains the Logistic regression model into seven data public opinion evolution models through steps S2-3-1 to 2-3-4 for the following seven types of data. The seven types of data include S, 、 The seven data public opinion evolution models are: the public opinion evolution model of the number of ordinary netizens, the public opinion evolution model of the number of public opinion information released, the public opinion evolution model of public opinion information forwarding data, the public opinion evolution model of public opinion information comment data, the public opinion evolution model of the number of neutral emotional information, the public opinion evolution model of the number of negative emotional information and the public opinion evolution model of the number of positive emotional information.

[0019] The intelligent early warning method for abnormal public opinion information provided by the present invention also includes obtaining modeling data of the core element abnormal perception model, which includes: S3-1: Obtaining prediction data sets through public opinion evolution models ; Get the measured data set: ; S3-2: Define normal interval A, observation interval B, and abnormal interval C to determine whether the measured data is abnormal data, such as Figure 2As shown, the normal interval A is: ,in, and , , ,i=N,N+1,…,N+M, M is a positive integer greater than or equal to 1; The observation interval B is: , , , where is the observation threshold; the interval outside the observation interval is the abnormal interval C.

[0020] S3-3: Based on measured data And the interval to which the subsequent monitoring data belongs, determine the measured data Is it normal data, observation data or abnormal data? If it is normal data, it will be integrated into the modeling data of the core element abnormality perception model, the core element abnormality perception model parameters will be updated, and the data predicted by the core element abnormality perception model will be updated; if it is observation data or abnormal data, the measured data will be calculated. The deviation of the evaluation level is determined while continuing to use the original core element abnormal perception model to make predictions, that is, if If it falls into the normal interval A, it is classified as normal data and integrated into the modeling data of the core element abnormality perception model, the modeling data parameters of the core element abnormality perception model are updated, and the predicted future data is updated; if If it falls into observation interval B, it is regarded as observation data. If subsequent data continues to fall into observation area B, it means that this is the precursor data of public opinion anomaly. After calculating the skewness, the evaluation level is determined while continuing to use the original core element anomaly perception model to make predictions; if If it falls into the abnormal interval C, it will be regarded as abnormal data. After calculating the skewness, the abnormal assessment level is determined, and then the modeling data of the original core element abnormal perception model is used to carry out predictions.

[0021] In the present invention, the deviation is calculated by the following formula: , i=N,N+1,…,N+M.

[0022] The present invention is based on the deviation The absolute value of is used to divide the abnormal assessment level of online public opinion into four levels: mild, moderate, high, and severe, with the intervals being: 、 、 , and then determine plans at different assessment levels to provide theoretical support for public opinion governance decisions.

[0023] After processing from step 3-1 to step 3-3, the measured number of ordinary netizens S and the public opinion vector are obtained from the public opinion monitoring data. and netizens' emotion vectors After filtering the observation data and abnormal data, the modeling data of the abnormal perception model of the core elements of network public opinion dissemination is obtained as follows: 、 and .

[0024] In the present invention, the sequence Fitting the public opinion subject symbiosis model; , Where, is the predicted value of the number of ordinary Internet users, is the quantity growth rate, The maximum number of is the symbiosis coefficient.

[0025] Use sequence Fitting the public opinion information propagation model: , Where, is the growth rate of the number of public opinion information releases, 、 、 is the adjustment factor.

[0026] Use sequence Fitting the netizens’ sentiment evolution model: , Where, , are the increase rates of neutral emotional information, negative emotional information, and positive emotional information, respectively. are the influence coefficients of neutral emotional information, negative emotional information, and positive emotional information, respectively. The upper limits of the number of neutral emotions, negative emotions, and positive emotions respectively.

[0027] The present invention adopts the differential regression method to perform data fitting on the public opinion subject symbiosis model, public opinion information dissemination model and netizen emotion evolution model respectively, and the fitting degree coefficient is (i.e., coefficient of determination) as a key indicator for identifying anomalies.

[0028] The present invention also provides a system comprising a storage medium and one or more processors, wherein the storage medium is used to store code for compiling the above-mentioned abnormal public opinion information intelligent early warning method into a computer program using computer language, and the computer program can be called and executed by one or more processors.

[0029] The technical solution formed by a combination of one or more steps in the above-mentioned intelligent early warning method for abnormal public opinion information disclosed in the present invention also falls within the scope disclosed in the present invention.

[0030] The present invention realizes real-time detection of public opinion data anomalies and risk warning by constructing an anomaly perception model of the core elements of online public opinion dissemination, filters the monitoring data through the public opinion evolution model, and improves the intelligent perception ability of unlabeled public opinion risks. Therefore, it can monitor new risks without relying on manually labeled features, and can deal with complex risks dynamically derived in public opinion dissemination.

[0031] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the contents of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can better understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A method for intelligent early warning of abnormal public opinion information, comprising the following steps: S1: Construct an abnormal perception model of the core elements of network public opinion dissemination, and predict the index values ​​of the core elements of network public opinion dissemination based on the core element abnormal perception model ; S2: Obtain the measured index values ​​of the core elements of online public opinion dissemination, filter the measured index values ​​to construct a measured index set H, and calculate the index weights in the measured index set to form a weight matrix W; S3: Determine the actual data of each indicator in the measured indicator set H and the predicted core element index value of network public opinion communication If the degree of fit of each indicator is greater than or equal to the threshold of each indicator, the actual data of each indicator is normal, otherwise it is abnormal, and a sequence of abnormal measured indicator sets is constructed. , K is a positive integer greater than or equal to 1; S4: According to the abnormal indicator sequence Calculate the probability of online public opinion risk: 。 2. The intelligent early warning method for abnormal public opinion information according to claim 1 is characterized in that: The core element abnormal perception model includes: public opinion subject symbiosis model, public opinion information dissemination model and netizen emotion evolution model. The number of ordinary netizens is predicted based on the public opinion subject symbiosis model, public opinion information dissemination model and netizen emotion evolution model. , public opinion vector and netizens' emotion vectors ,Right now ,in, , Respectively represent the predicted number of public opinion information releases, forwarding numbers, and comment data; , Represents the predicted number of public opinion information releases The number of neutral emotions, negative emotions, and positive emotions in .

3. The intelligent early warning method for abnormal public opinion information according to claim 2 is characterized in that: Obtain the measured index values ​​of the core elements of online public opinion dissemination, and filter the measured index values ​​to construct the measured index set H, which includes: S2-1: Obtain the measured number of ordinary netizens S and public opinion vector from public opinion monitoring data and netizens' emotion vectors , Respectively represent the measured number of information releases, forwarding numbers, and comment data, They represent the number of neutral emotions, negative emotions, and positive emotions in the measured number of public opinion information releases M1; S2-2: Obtain the data sequence of the public opinion evolution model based on the information content of the measured data a at time n ; According to the data sequence of the public opinion evolution model, a public opinion evolution model is constructed to predict the predicted value after N moments to further obtain the evolution trend, where a represents any of the following data: S, 、 ; S2-3: Modeling Data Sequences through Public Opinion Evolution Model Train the Logistic regression model into a public opinion evolution model.

4. The intelligent early warning method for abnormal public opinion information according to claim 3 is characterized in that: The logistic regression model is: ， Where, is the intrinsic growth rate, K is the information limit; It is the amount of public opinion information about a social event at the initial moment.

5. The intelligent early warning method for abnormal public opinion information according to claim 4 is characterized in that: It also includes obtaining modeling data for the core element anomaly perception model, which includes: S3-1: Obtaining prediction data sets through public opinion evolution models ; Get the measured data set: ; S3-2: Define normal interval A, observation interval B, and abnormal interval C to determine whether the measured data is abnormal data; S3-3: Based on measured data And the interval to which the subsequent monitoring data belongs, determine the monitoring data Is it normal data, observation data or abnormal data? If it is normal data, it will be integrated into the modeling data of the core element abnormality perception model, the core element abnormality perception model parameters will be updated, and the data predicted by the core element abnormality perception model will be updated; if it is observation data or abnormal data, the measured data will be calculated. The deviation is determined, the evaluation level is determined, and the original core element anomaly perception model is continued to be used for prediction.

6. The intelligent early warning method for abnormal public opinion information according to claim 5 is characterized in that: The deviation is calculated by the following formula: , i=N,N+1,…,N+M.

7. The intelligent early warning method for abnormal public opinion information according to claim 2 is characterized in that: The symbiotic model of public opinion subjects is: ， Where, is the number of ordinary Internet users, is the quantity growth rate, The maximum number of is the symbiosis coefficient.

8. The intelligent early warning method for abnormal public opinion information according to claim 2 is characterized in that: Public opinion information dissemination model: ， Where, is the growth rate of the number of public opinion information releases, 、 、 is the adjustment factor.

9. The intelligent early warning method for abnormal public opinion information according to claim 2 is characterized in that: Internet users' emotional evolution model: ， Where, , are the increase rates of neutral emotional information, negative emotional information, and positive emotional information, respectively. are the influence coefficients of neutral emotional information, negative emotional information, and positive emotional information, respectively. The upper limits of the number of neutral emotions, negative emotions, and positive emotions respectively.

10. A system comprising a storage medium and one or more processors, wherein the storage medium is used to store code that uses computer language to compile the abnormal public opinion information intelligent early warning method described in any one of claims 1 to 9 into a computer program, and the computer program can be called and executed by one or more processors.

Images