CN111861122A - A Social Network Information Credibility Evaluation Method Based on Similarity of Propagation Attributes - Google Patents

Abstract

The invention provides a social network information credibility evaluation method based on propagation attribute similarity, which comprises the following steps: step A: extracting social network information propagation content and constructing a social network information propagation network; and B: extracting attributes of social network information transmission nodes, calculating topological attributes of social network information transmission networks, marking the reliability of historical information, and establishing a historical information transmission attribute database; and C: calculating the propagation attribute similarity of the target social network and the historical social network so as to evaluate the information credibility of the target social network; through the steps, the purpose of evaluating the credibility of the social network information based on the similarity of the propagation attributes can be achieved, the method is strong in integrity, high in objectivity and good in universality, and the problem that the credibility of the social network information is difficult to objectively measure and evaluate is solved.

Description

A Social Network Information Credibility Evaluation Method Based on Propagation Attribute Similarity

technical field

The invention proposes a method for evaluating the credibility of social network information based on the similarity of propagation attributes, which relates to the technical fields of network science, social network analysis and the like.

Background technique

In a narrow sense, social network information refers to information that can be traced to a certain period of time in a social network environment, that causes dissemination and has a certain impact on society. Hot news, hot topics, controversial events, etc. are all common social network information in social networks.

With the development of information technology and the emergence of new network media, the speed of social network information dissemination is getting faster and faster, the scale of dissemination is increasing, and the content of dissemination has an increasing influence with the increase of dissemination levels. For example, when faced with a major disaster, people can quickly and accurately transmit social network information related to the disaster through various social network platforms such as Weibo, form hot news, and disclose controversial events such as "the timeliness of disaster relief measures". It is better and faster to prompt the government and the people to take rescue measures; however, there are also false information and rumors in social network information, such as controversial events and information related to pseudoscience and religious superstition. It is widely spread on the Internet, which will cause panic among the masses and even affect social stability. Therefore, how to evaluate the credibility of a social network information has become a key research problem in all sectors of society.

Social network information with high credibility, its dissemination traceability is generally a socially certified information publishing agency or individual, such social network information can timely convey real events, and timely notify or convey valuable information to those in need from all walks of life information and knowledge. However, social network information that lacks credibility usually exists in the form of "misinformation", and its dissemination range is more scattered and uncertain than social network information with high credibility, and the dissemination groups are also difficult to distinguish. Therefore, It can confuse the eyes of the masses on social networks and spread untrustworthy information widely.

Due to the ubiquity of social network information lacking credibility and the resulting losses have a huge impact on people's lives and even social operations, scientific and reasonable methods are used to accurately and credible social network information. Degree evaluation and judgment are very necessary. In the process of evaluating the credibility of information, a variety of evaluation methods can be used, including credibility evaluation methods based on the attributes of information publishers, such as authority, influence, activity, and social relations, based on information content attributes such as information integrity The credibility evaluation method of the degree, language analysis, content position judgment, etc., as well as the credibility evaluation method based on the macro characteristics of the time and quantity of information release. And how to quickly evaluate the credibility of the information through the dissemination characteristics of the information is the focus of the present invention.

Most of the previous methods are credibility evaluation methods based on macro attributes such as publisher attributes, information content attributes, or information time and quantity, or require a lot of manpower and material resources to collect and judge the specific social network information attributes to be evaluated. Little research and utilization of information dissemination attributes are conducted, and the credibility of social network information is not evaluated with reference to historical information dissemination attributes. Therefore, although the above methods can quantitatively or qualitatively evaluate the credibility of specific social network information, they cannot accurately grasp the significant and fundamental differences in the dissemination characteristics of trusted and untrustworthy information in social networks, so that the generalization of the above methods The adaptability is poor, the general interpretability is not strong, and the labor and time cost is huge. However, existing research has shown that social network information with similar credibility has obvious similarities in propagation attributes, while social network information with different credibility has obvious differences in the dissemination structure attributes, and these differences are not caused by information. The specific content varies.

In order to evaluate the credibility of a target social network information, the present invention first extracts a certain amount of historical information and the dissemination attributes of the target social network information, including the disseminator attribute and the dissemination network structure, and takes the disseminator as a node to forward the information. The relationship is a directed edge to build an information dissemination network, calculate the dissemination attributes of the target social network information, and calculate the dissemination attribute similarity between the target information and the historical information by comparing with the existing information dissemination network database. Information credibility comparison, to evaluate the credibility of the target social network information.

This patent calculates and analyzes the credibility of social network information by introducing the propagation attribute features. Considering the historical credibility of social network information and the propagation attribute characteristics, the credibility of the target social network information can be reasonably evaluated, and it has better performance. versatility and innovation. Based on the above method basis and practical significance, a method for evaluating the credibility of social network information based on the similarity of propagation attributes is proposed.

SUMMARY OF THE INVENTION

(1) Purpose of the invention

The present invention is mainly used to solve the problem of information credibility evaluation under the background of social network platform dissemination of unverified mass information. Most of the existing methods are credibility evaluation methods based on macro attributes such as publisher attributes, information content attributes, or information time and quantity, or require a lot of manpower and material resources to collect and judge the specific social network information attributes being evaluated. Little research and utilization of information dissemination attributes are conducted, and the credibility of social network information is not evaluated with reference to historical information dissemination attributes. Therefore, in view of the deficiencies of the existing methods, this patent realizes the evaluation of the credibility of social network information based on the similarity of the propagation attributes from the perspective of the propagation path.

By using this method to realize the extraction of information dissemination node attributes and dissemination structure attributes in the social network environment, the information dissemination characteristics and historical information are reasonably calculated and compared to establish a credibility evaluation method of social network information, and then it can be Realize the credibility assessment of the target information, thus providing a solid foundation for the adjustment of the monitoring intensity of social network information, the rating of public opinion, and the necessary investigation and declaration of the target event.

(2) Technical solutions

In order to achieve the above object, the technical solution adopted by the method of the present invention is: a method for evaluating the credibility of social network information based on the similarity of propagation attributes.

The method for evaluating the credibility of social network information based on the similarity of propagation attributes described in the present invention is an idea of applying complex network modeling. The forwarding relationship of events, etc. and the disseminator establish a network as an edge and a node, and then build a historical social network information database, and then extract the target social network information to be evaluated (such as a specific hot news) and each historical social network information dissemination network Propagation attribute similarity, so as to measure and evaluate the credibility of this piece of social network information.

The steps of "a method for evaluating the credibility of social network information based on the similarity of propagation attributes" described in the present invention are as follows:

Step A: extracting social network information dissemination content, and constructing a social network information dissemination network;

Step B: extracting the attributes of social network information dissemination nodes, calculating the network topology attributes of social network information dissemination, marking the credibility of historical information, and establishing a historical information dissemination attribute database;

Step C: Calculate the dissemination attribute similarity between the target social network information and the historical social network information, so as to evaluate the information credibility of the target social network.

Through the above steps, the purpose of evaluating the credibility of social network information based on the similarity of dissemination attributes can be achieved. The problem.

Wherein, the method of "extracting social network information dissemination content and constructing a social network information dissemination network" described in step A is as follows: for a given number of historical social network information and target social network information with historical information credibility, Each of them selects an information publisher as the initial propagation node, and constructs the information propagation network layer by layer with the direct forwarding node of the initial propagation node and the direct forwarding directed relationship; the specific steps of the above-mentioned process are as follows:

式中：V₀表示初始传播节点V₀₀与对该初始节点的直接转发者节点V_0j的节点集合，E₀表示初始传播节点V₀₀与其直接转发者节点V_0j之间的转发关系连边集合；Step A1: According to the public content of the social network platform, select a publisher of a certain social network information Event (which may be a hot topic, hot news, controversial event, etc.) as the initial dissemination node V ₀₀ , and directly forward the initial node. As the lower-level node _{V 0j} , _the first-level information dissemination network G _{0 (} V ₀ , E ₀ ), where

In the formula: V ₀ represents the node set of the initial propagation node V ₀₀ and the direct forwarder node V _0j of the initial node, E ₀ represents the forwarding relationship between the initial propagation node V ₀₀ and its direct forwarder node V _0j The set of edges ;

Step A2: Set the threshold L of the number of layers of the information dissemination network, traverse each forwarding node of each layer, regard it as each initial node of the layer, repeat step A1, build the information dissemination network layer by layer until the dissemination level reaches the threshold, and obtain the social network. The overall dissemination network G(V,E) of the network information Event; in the formula: G(V,E) represents the overall dissemination network of the social network information Event, V represents the node set in the network, and E represents the node forwarding in the network relation;

Step A3: Repeating steps A1 and A2, establishing a respective information dissemination network G _history (V _history , E _history ) for a given amount of Num of historical social network information Event _history , and establishing an information dissemination network G _obj of the target social network information Event _obj (V _obj , E _obj ); in the formula: G _history (V _history , E _history ) represents the information dissemination network of each historical social network information Event _history , V _history represents the set of nodes in the network, and E _history represents the nodes in the network forwarding relationship, and G _obj (V _obj , E _obj ) represents the information dissemination network of the target social network information Event _obj , V _obj represents the set of nodes in the network, and E _obj represents the forwarding relationship of nodes in the network.

Wherein, the method of "extracting social network information dissemination node attributes, calculating social network information dissemination network topology attributes, marking the credibility of historical information, and establishing a historical information dissemination attribute database" described in step B is as follows: extracting the constructed social network information dissemination network Each dissemination node in the network information dissemination network and the dissemination topology structure attributes, then mark the credibility of each historical social network information, and establish a historical social network information dissemination attribute database; the specific steps of the above-mentioned process are as follows:

Step B1: Extract the unique identification information of each dissemination node V in the constructed social network information dissemination network G(V, E) as the dissemination node attribute F _vec of the information;

Step B2: Extract the network topology of the constructed social network information dissemination network G(V,E), and extract the network dissemination structure attribute F _struc , including but not limited to:

(1) The ratio of the number of initial forwarding layers of the propagation network G(V, E) to r _2/1 (as in equation (1)), that is, the number of nodes in the second layer of propagation network n _V (2) and the number of nodes in the first layer of propagation network n The ratio of _V (1);

(2) The characteristic distance a of the nodes of the propagation network G(V, E) is used to fit the distance distribution between all pairs of nodes in the propagation network. The specific fitting equation is shown in Equation (2), where y is the distribution probability and x is the separation distance of the node pair, b is the fitting constant;

(3) The homogeneity index h of the propagation network G(V, E), that is, the difference between the logarithm values of the homogeneity H _G of the propagation network G(V, E) and the homogeneity H _star of the star network of the same scale, calculate The method is as formula (3), and the calculation method of the homogeneity of the propagation network is as formula (4);

h=log(H _star )-log(H _G )(3)

(式中N为网络节点总数，k为节点度) （4)

(where N is the total number of network nodes, and k is the node degree) (4)

其中所取值0表示该信息完全不可信，所取值1表示该信息完全可信，设定历史社交网络信息Event_history的传播节点可信度评价指标Credibility_vec(Event_history)、传播拓扑可信度评价指标 Credibility_struc(Event_history)与综合可信度评价指标相同；Step B3: Set the credibility evaluation index of the historical social network information Event _history according to the summary of historical facts and the certification of the authority

The value of 0 indicates that the information is completely unreliable, and the value of 1 indicates that the information is completely credible. Set the credibility evaluation indicators of the propagation nodes of the historical social network information Event _history , Credibility _vec (Event _history ), and the propagation topology credible The degree evaluation index Credibility _struc (Event _history ) is the same as the comprehensive credibility evaluation index;

Step B4: For the collected Num pieces of historical social network information Event _history , the historical information dissemination node attribute F _vec (Event _history ), the historical information dissemination network structure attribute F _struc (Event _history ), and the historical information credibility index Credibility ( Event _history ) is added to the historical social network information dissemination attribute database DS (Event _history ).

Wherein, the "calculation of the similarity of the propagation attributes of the target social network information and the historical social network information, so as to evaluate the information credibility of the target social network" described in step C, the specific method is as follows: Calculate the target social network information Event Sim _vec is the similarity between _obj and the propagation node of historical social network information Event _history and obtains the propagation node credibility of the target social network information Event _obj , Credibility _vec (Event _obj ), and then calculates the target social network information Event _obj and historical social network information Event The similarity quantile of the communication structure of _history , and then obtain the credibility of the communication structure Credibility _struc (Event _obj ), assign the credibility of each communication attribute to calculate the weight, and calculate the credibility of the target social network information; the specific steps of the above process are as follows :

Step C1: Using a set similarity calculation method such as the Jaccard similarity method (Equation (5)), respectively calculate the target social network information Event _obj and the propagation node attributes of all historical information Event _history in the historical information propagation attribute database DS (Event _history ) Similarity Sim _vec (Event _obj , Event _history ), then select the largest propagation node attribute similarity value Sim _{vec_max} in the calculation result and its corresponding historical information Event _{vec_max} , record the credibility of the corresponding historical information Credibility _vec (Event vec_max ) _{vec_max} ), then the credibility of the propagation node of the target social network information Event _obj is Credibility _vec (Event _obj )=Sim _{vec_max} ×Credibility _vec (Event _{vec_max} );

In the formula: F _vec (Event _obj ) represents the attribute of the target information dissemination node, and F _vec (Event _history ) represents the attribute of the historical information dissemination node;

Step C2: For each historical information Event _history in the historical information dissemination attribute database DS (Event _history ), based on the size of the M sub-attributes (including but not limited to r _2/1 , a, and h) in the dissemination structure attribute F _struc , according to Sort by the same sorting method (ascending or descending), and get M historical information sequences after sorting:

个历史信息，随后比较目标社交网络信息的第i传播结构子属性值与第i个重排历史信息序列中各历史信息Event_history的传播结构子属性值，找到最相似的历史信息Event_history并将其所属区间序号 k_{struc_i}作为目标社交网络信息Event_obj传播结构子属性的相似性分位，对 i＝1,2,...,M个传播结构属性子属性重复此步骤，得到目标社交网络信息Event_obj各传播结构子属性的相似性分位；Step C3: For the i-th rearranged historical information sequence obtained by C2, divide the historical information into K intervals with equal numbers, then each interval has

and then compare the i-th propagation structure sub-attribute value of the target social network information with the propagation structure sub-attribute value of each historical information Event _history in the i-th rearranged historical information sequence, find the most similar historical information Event _history The interval sequence number k _{struc_i} to which it belongs is used as the similarity quantile of the sub-attributes of the target social network information Event _obj propagation structure, and this step is repeated for i=1, 2, ..., M sub-attributes of the propagation structure attributes to obtain the target social network information Similarity quantiles of sub-attributes of each propagation structure of Event _obj ;

分别为目标社交网络信息Event_obj所属的此传播结构子属性的分位k_i中，可信度为0、1的历史信息Event_history个数；Step C4: Calculate the propagation structure sub-attribute Credibility _{struc_i} of the target social network information Event _obj and the historical social network information Event _history according to formula (6), where i is the ith in the M propagation structure sub-attributes,

Respectively, in the quantile k _i of the sub-attribute of this propagation structure to which the target social network information Event _obj belongs, the number of historical information Event _history whose reliability is 0 and 1;

Step C5: Allocate calculation weights w _vec , w _{struc_1} ,..., w _{struc_M} to the propagation node credibility Credibility _vec and the M propagation structure sub-attribute credibility Credibility _{struc_i} , i=1,...,M, where The sum of all weights is equal to 1, and the distribution method can adopt the average distribution method, the analytic hierarchy process, the fuzzy evaluation method, etc., and distribute according to the weight importance or other necessary information;

Step C6: Calculate the credibility of the target social network information by using the formula (7), and evaluate the credibility of the target social network information according to the result. The closer the calculation result is to 0, the weaker the credibility, and vice versa. 1, the more reliable it is;

(3) Merit innovation

The present invention has the following innovations:

1. Strong versatility: This patent is not a credibility evaluation method based on the similarity of propagation attributes for a specific social network information, but a kind of similarity-based propagation attribute that is common to all kinds of social network information. It is a reliable evaluation method of social network information, so it has good generality.

2. Good portability: This patent does not limit the content of the dissemination attributes and the calculation method of the similarity quantile of each dissemination attribute. Therefore, the dissemination attributes can be determined according to the needs of the actual situation in the evaluation of the credibility of different social network information. The deletion and the adjustment of the calculation method, so it has good portability.

3. High degree of objectivity: This patent improves the applicability of the method by introducing the dissemination characteristics of social network information and the credibility of historical information that has been confirmed, and enables more objective credibility assessment.

4. Strong integrity: This patent is a credibility evaluation of social network information based on the similarity of dissemination attributes from the perspective of the whole process of information dissemination, so it can grasp the changes of global information and has good integrity.

To sum up, this social network information credibility evaluation method based on the similarity of dissemination attributes can combine historical information and the characteristics of information dissemination to better evaluate the credibility of a social network information, which can make up for the shortcomings of existing methods; The method of the invention is scientific, has good manufacturability, and has broad popularization and application value.

Description of drawings

FIG. 1 is a flow chart of the method framework of the present invention.

Detailed ways

In order to make the technical problems and technical solutions to be solved by the present invention clearer, the following will describe in detail with reference to the accompanying drawings and specific implementation cases. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, but not to limit the present invention.

The purpose of the present invention is to solve the problem of social network information credibility evaluation under the background of social network platform dissemination of unverified mass information. Most of the existing methods are credibility evaluation methods based on macro attributes such as publisher attributes, information content attributes, or information time and quantity, or require a lot of manpower and material resources to collect and judge the specific social network information attributes being evaluated. Little research and utilization of information dissemination attributes are conducted, and the credibility of social network information is not evaluated with reference to historical information dissemination attributes. Therefore, based on the deficiencies of the existing methods, this patent realizes the evaluation of the credibility of social network information based on the similarity of the propagation attributes from the perspective of the propagation path.

The method has the characteristics of strong versatility, good portability, high objectivity and strong integrity. The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

The embodiments of the present invention illustrate the method of the present invention by taking the credibility evaluation scenario of the target social network information Event _obj as an example. It is known that Event _obj is a piece of social network information that has caused heated discussions and widespread dissemination on social network platforms, and whether the information is credible or true has not been explained and confirmed by an official or authoritative organization. The following is the content of the Event _obj as " Eating garlic cannot eliminate the new coronavirus" as an example to illustrate the method of the present invention.

In order to achieve the above object, the technical solution adopted by the method of the present invention is: a method for evaluating the credibility of social network information based on the similarity of propagation attributes.

The "a method for evaluating the credibility of social network information based on the similarity of propagation attributes" described in the present invention is shown in Figure 1, and its steps are as follows:

Step A: extracting social network information dissemination content, and constructing a social network information dissemination network;

Step B: extracting the attributes of social network information dissemination nodes, calculating the network topology attributes of social network information dissemination, marking the credibility of historical information, and establishing a historical information dissemination attribute database;

Step C: Calculate the similarity of the propagation attributes of the target social network and the historical social network, so as to evaluate the information credibility of the target social network.

Through the above steps, the purpose of evaluating the credibility of social network information based on the similarity of dissemination attributes can be achieved. The problem.

Wherein, the method of "extracting social network information dissemination content and constructing a social network information dissemination network" described in step A is as follows: for a given number of historical social network information and target social network information with historical information credibility, Each of them selects an information publisher as the initial propagation node, and constructs the information propagation network layer by layer with the direct forwarding node of the initial propagation node and the direct forwarding directed relationship; the specific steps of the above-mentioned process are as follows:

式中：V₀表示初始传播节点V₀₀与对该初始节点的直接转发者节点V_0j的节点集合，E₀表示初始传播节点V₀₀与其直接转发者节点V_0j之间的转发关系连边集合；Step A1: According to the public content of the social network platform, select a publisher of a certain social network information Event as the initial dissemination node V ₀₀ , and the direct forwarder of the initial node as the lower node V _0j , depending on the initial node and each lower layer The direct forwarding relationship between nodes is used as a social network information propagation directed edge E ₀ (V ₀₀ , V _0j ), and the first layer of information propagation network G ₀ (V ₀ , E ₀ ) is constructed, where

In the formula: V ₀ represents the node set of the initial propagation node V ₀₀ and the direct forwarder node V _0j of the initial node, E ₀ represents the forwarding relationship between the initial propagation node V ₀₀ and its direct forwarder node V _0j The set of edges ;

Step A2: Set the threshold L of the number of information dissemination network levels, such as L=3, traverse each forwarding node of each layer, and regard it as each initial node of the layer, repeat step A1, build the information dissemination network layer by layer until the dissemination level reaches Threshold to obtain the overall dissemination network G(V,E) of the social network information Event, for example, for the initial dissemination node V ₀₀ , collect each of its direct forwarders as the lower node V _0j , j=1, 2, . . . and establish The edge relationship E ₀ (V ₀₀ , V _0j ), at this time, the number of levels of the propagation network is 1, and then for each node V _0j , j=1, 2, . The direct forwarder acts as the lower node _{V 1k} , _k = ₁ , 2, . Propagating the network until the propagation level reaches the threshold, and obtain the overall propagation network G(V, E) of the social network information Event;

Step A3: Repeat steps A1 and A2 to establish a respective information dissemination network G _history (V _history , E _history ) for the historical social network information Event _history of a given number Num (for example, Num=100), wherein the historical social network information Event _history It can be the verified content related to the content of the target social network information Event _obj on the social platform (for example: "Eating garlic can't kill the virus" is trusted information, "Drinking alcohol can kill the new coronavirus" is untrustworthy information), or It can be verified content unrelated to the content of the target social network information Event _obj (for example: "The new coronavirus killed one million people in China" is untrustworthy information, "The biological crisis is a conspiracy made by the United States" is untrustworthy information ). At the same time, steps A1 and A2 are repeated to establish a target social network information Event _obj , that is, an information dissemination network G _obj (V _obj , E _obj ) of “eating garlic cannot eliminate the new coronavirus”.

Wherein, the method of "extracting social network information dissemination node attributes, calculating social network information dissemination network topology attributes, marking the credibility of historical information, and establishing a historical information dissemination attribute database" described in step B is as follows: for a given number of Both historical social network information and target social network information with historical information credibility are extracted from each propagation node and propagation topology attribute in the constructed social network information propagation network (whether it is a historical information network or a target information network), and then marked. The credibility of each historical social network information is established, and a historical social network information dissemination attribute database is established; the specific steps of the above-mentioned process are as follows:

Step B1: Extract the unique identification information of each dissemination node V in the constructed social network information dissemination network G(V, E), such as the user number of the social platform, as the dissemination node attribute F _vec of the piece of information;

Step B2: Extract the network topology structure of the constructed social network information dissemination network G(V, E), and extract the network dissemination structure attribute F _struc using computer methods or manual statistics, including but not limited to:

(1) The ratio of the number of initial forwarding layers of the propagation network G(V, E) to r _2/1 (as in equation (1)), that is, the number of nodes in the second layer of propagation network n _V (2) and the number of nodes in the first layer of propagation network n The ratio of _V (1);

(2) The characteristic distance a of the nodes of the propagation network G(V, E) is used to fit the distance distribution between all pairs of nodes in the propagation network. The specific fitting equation is shown in Equation (2), where y is the distribution probability and x is the separation distance of the node pair;

(3) The homogeneity index h of the propagation network G(V, E) is the difference between the logarithm value of the homogeneity of the propagation network G(V, E) and the homogeneity of the star network of the same scale. The calculation method is as follows: 3), where the homogeneity calculation method of the propagation network is as in formula (4);

其中所取值0表示该信息完全不可信，所取值1表示该信息完全可信，例如对一条这样的历史社交网络信息Event1：“吃大蒜不可以消灭病毒”，根据历史事实可知该信息完全可信，则其综合可信度Credibility(Event1)＝1，那么根据综合可信度评价指标 Credibility(Event1)＝1，相应地设定历史社交网络信息Event_history的传播节点可信度评价指标Credibility_vec(Event1)、传播拓扑可信度评价指标Credibility_struc(Event1)与综合可信度评价指标相同，都为1；Step B3: According to the summary of historical facts and the certification of authoritative organizations, set the comprehensive credibility evaluation index of historical social network information Event _history

The value of 0 indicates that the information is completely unreliable, and the value of 1 indicates that the information is completely credible. For example, for a piece of such historical social network information Event1: "Eating garlic can't kill the virus", it can be seen from historical facts that the information is completely Credibility, then its comprehensive credibility Credibility(Event1)=1, then according to the comprehensive credibility evaluation index Credibility(Event1)=1, correspondingly set the historical social network information Event _history propagation node credibility evaluation index Credibility _vec (Event1), the propagation topology credibility evaluation index Credibility _struc (Event1) are the same as the comprehensive credibility evaluation index, and both are 1;

Step B4: For the above-mentioned collected number of Num=100 historical social network information Event _history , the historical information dissemination node attribute F _vec (Event _history ), the historical information dissemination network structure attribute F _struc (Event _history ), the historical information can be The reliability indicator Credibility (Event _history ) is added to the historical social network information dissemination attribute database DS (Event _history ).

Wherein, the “calculation of the similarity of the dissemination attributes between the target social network information and the historical social network information, so as to evaluate the information credibility of the target social network” described in step C is as follows: dissemination of historical social network information For each piece of information Event _history collected in the attribute database, calculate the similarity Sim _vec between the target social network information Event _obj and the propagation node of the Event _history , and obtain the propagation node Credibility _vec (Event _obj ) of the target social network information Event _obj . ), then calculate the dissemination structure similarity quantile between the target social network information Event _obj and the historical social network information Event _history , and then obtain the credibility of the dissemination structure Credibility _struc (Event _obj ), assign the credibility calculation weight of each dissemination attribute, calculate The credibility of the target social network information; the specific steps of the above-mentioned process are as follows:

Step C1: Using a set similarity calculation method such as the Jaccard similarity method (Equation (5)), respectively calculate the target social network information Event _obj and the propagation node attributes of all historical information Event _history in the historical information propagation attribute database DS (Event _history ) Similarity Sim _vec (Event _obj , Event _history ), and then select the largest propagation node attribute similarity value Sim _{vec_max} (for example, Sim _{vec_max} = 0.8) in the calculation result and its corresponding historical information Event _{vec_max} (for example, the content of this piece of information is "Eating garlic can't kill the virus"), record the credibility of the corresponding historical information Credibility _vec (Event _{vec_max} ) (known Credibility _vec (Event _{vec_max} ) = 1), then the dissemination node of the target social network information Event _obj is credible Degree Credibility _vec (Event _obj )=Sim _{vec_max} ×Credibility _vec (Event _{vec_max} )=0.8×1=0.8;

Step C2: Sort each historical information Event _history in the historical information dissemination attribute database DS (Event _history ) in descending order based on the size of the M (M=3 in this example) sub-attributes in the dissemination structure attribute F _struc , for example, Sort each historical information Event _history in descending order based on the size of the sub-attribute _{struc_1} of the propagation structure, and the order of obtaining the historical information is Event _history1 , Event _history2 ,... (Event _history ) is the largest, and the above-mentioned sorting is also performed on each historical information according to the remaining sub-attribute values, and finally a rearrangement sequence of M (for example, M=3) historical information can be obtained:

个历史信息Step C3: Rearrange the i-th (for example, i=1) historical information sequence obtained by C2, and divide the historical information into K=10 intervals by an equal number, then each interval has

historical information

个区间)，随后比较目标社交网络信息Event_obj(即“吃大蒜不可以消灭新冠病毒”)的第i 传播结构子属性值与第i个重排历史信息序列中，各历史信息Event_history的传播结构子属性值，找到最相似的历史信息Event_history并将其所属区间序号k_{struc_i}＝3 作为目标社交网络信息Event_obj传播结构子属性的相似性分位，例如对子属性struc_1，目标社交网络信息Event_obj的该属性值为0.34，与List_{struc_1}中的第23个历史信息Event_history23对应的子属性struc_1最为相似，由于Event_history23位于该序列的第3个区间，则将其所属区间序号k_{struc_1}＝3作为目标社交网络信息Event_obj传播结构子属性struc_1的相似性分位，对i＝1,2,...,M个传播结构属性子属性重复此步骤，可得到得到目标社交网络信息Event_obj各传播结构子属性的相似性分位 (3，4，2)；(For example, List _{struc_1} (Event _history1 , Event _history2 ,...) has a total of 100 historical information sorted in descending order based on the size of the sub-attribute struc_1 of the propagation structure, and the top 10 historical information with the largest sub-attribute struc_1 is divided into an interval number 1. The 11th to 20th historical information of the second largest sub-attribute struc_1 is divided into an interval and numbered 2, and the 100 historical information is divided into

interval), and then compare the i-th propagation structure sub-attribute value of the target social network information Event _obj (that is, "eating garlic cannot kill the new coronavirus") and the i-th rearranged historical information sequence, the propagation of each historical information Event _history Structural sub-attribute value, find the most similar historical information Event _history and use the interval number k _{struc_i} = 3 as the target social network information Event _obj to propagate the similarity quantile of the structural sub-attribute, for example, for the sub-attribute struc_1, the target social network information The attribute value of Event _obj is 0.34, which is most similar to the sub-attribute struc_1 corresponding to the 23rd historical information Event _history23 in List _{struc_1} . Since Event _history23 is located in the third interval of the sequence, the interval number k _{struc_1} = 3 as the similarity quantile of the target social network information Event _obj propagation structure sub-attribute struc_1, repeat this step for i=1, 2, ..., M propagation structure attribute sub-attributes to obtain the target social network information Event _obj Similarity quantile (3, 4, 2) of each propagation structure sub-attribute;

分别为目标社交网络信息Event_obj所属的此传播结构子属性的分位k_i中，可信度为0、1的历史信息Event_history个数，例如，对于第i＝1个传播结构子属性struc_1，其分位值k_{struc_1}＝3，且对应第i＝1个重排历史信息序列第3区间中可信度为0、1的历史信息Event_history个数

分别等于3、7，则目标社交网络信息的第一个传播结构子属性值

进而同理得到其他两个传播结构子属性可信度分别为0.7、0.8；Step C4: Calculate the target social network information Event _obj according to formula (6), that is, “eating garlic cannot kill the new coronavirus”, and the propagation structure sub-attribute Credibility _{struc_i} of the historical social network information Event _history , where i is M= the ith of the 3 propagation structure sub-attributes,

In the quantile k _i of this propagation structure sub-attribute to which the target social network information Event _obj belongs, the number of historical information Event _history with reliability of 0 and 1, for example, for the i-th = 1st propagation structure sub-attribute struc_1 , its quantile value k _{struc_1} = 3, and corresponds to the number of historical information Event _history whose reliability is 0 and 1 in the third interval of the i=1 rearranged historical information sequence

are equal to 3 and 7, respectively, then the first propagation structure sub-attribute value of the target social network information

In the same way, the credibility of the other two propagation structure sub-attributes is 0.7 and 0.8, respectively;

Step C5: Credibility _vec of propagation nodes, M=3 propagation structure sub-attributes

Credibility _{struc_i} , i=1,...,M assigns the calculation weights w _vec ,w _{struc_1} ,...,w _{struc_M} , where the sum of all weights is equal to 1, and the assignment method adopts the average assignment method in this example, for example

Step C6: Use formula (7) to calculate and obtain the target social network information, that is, the credibility of "eating garlic cannot eliminate the new coronavirus" is 0.75, and the credibility of the target social network information is evaluated according to the result. If it is close to 0, its credibility will be weaker. On the contrary, if it is closer to 1, its credibility will be stronger. It can be seen that the target social network information is more reliable, and its credibility value is 0.75, indicating that "eating garlic cannot eliminate the new coronavirus." ”’s social network comprehensive reliability is 0.75, there is a high probability that this piece of information can be believed, and social network users who see this piece of information on social network platforms can be considered generally credible;

Claims (4)

1. a kind of social network information credibility assessment method based on the similarity of propagation attributes, is characterized in that: its steps are as follows: Step A: extracting social network information dissemination content, and constructing a social network information dissemination network; Step B: extracting the attributes of social network information dissemination nodes, calculating the network topology attributes of social network information dissemination, marking the credibility of historical information, and establishing a historical information dissemination attribute database; Step C: Calculate the similarity of the propagation attributes of the target social network and the historical social network, so as to evaluate the information credibility of the target social network. 2. a kind of social network information credibility assessment method based on the similarity of propagation attributes according to claim 1, is characterized in that: The method of "extracting social network information dissemination content and constructing a social network information dissemination network" described in step A is as follows: for a social network information, an information publisher is selected as an initial dissemination node, and the initial dissemination node is based on the layer by layer. The direct forwarding node of the node has a directed relationship with the direct forwarding to construct the dissemination network of the information, and the above-mentioned dissemination network is constructed for a given number of historical social network information and target social network information with historical information credibility; The specific steps of the process are as follows: Step A1: According to the public information of the social network platform, select a publisher of a social network information Event as the initial dissemination node V ₀₀ , and the direct forwarder of the initial node as the lower node V _0j , depending on the initial node and each lower node The direct forwarding relationship between them is used as a social network information propagation directed edge E ₀ (V ₀₀ , V _0j ), and the first-layer information propagation network G ₀ (V ₀ , E ₀ ) is constructed, where

In the formula: V ₀ represents the node set of the initial propagation node V ₀₀ and the direct forwarder node V _0j of the initial node, E ₀ represents the forwarding relationship between the initial propagation node V ₀₀ and its direct forwarder node V _0j The set of edges ; Step A2: Set the threshold of the number of layers of the information dissemination network, traverse a plurality of forwarding nodes in each layer, and regard it as a plurality of initial nodes of the layer, repeat step A1, build an inter-layer information dissemination network layer by layer until the dissemination level reaches the threshold, and obtain: The overall dissemination network G(V, E) of the social network information Event; in the formula: V represents the node set in the network, and E represents the node forwarding relationship in the network; Step A3: Repeating steps A1 and A2, establishing a respective information dissemination network G _history (V _history , E _history ) for a given amount of Num of historical social network information Event _history , and establishing an information dissemination network G _obj of the target social network information Event _obj (V _obj , E _obj ); in the formula: G _history (V _history , E _history ) represents the information dissemination network of each historical social network information Event _history , V _history represents the set of nodes in the network, and E _history represents the nodes in the network forwarding relationship, and G _obj (V _obj , E _obj ) represents the information dissemination network of the target social network information Event _obj , V _obj represents the node set in the network, and E _obj represents the node forwarding relationship in the network. 3. a kind of social network information credibility evaluation method based on the similarity of propagation attributes according to claim 1, is characterized in that: In the step B of "extracting the node attributes of social network information dissemination, calculating the topology attributes of social network information dissemination network, marking the credibility of historical information, and establishing a database of historical information dissemination attributes", the method is as follows: extracting the constructed social network Each dissemination node and dissemination topology attribute in the information dissemination network, then mark the credibility of each historical social network information, and establish a historical social network information dissemination attribute database; the specific steps of the above-mentioned process are as follows: Step B1: Extract the unique identification information of each dissemination node V in the constructed social network information dissemination network G(V, E) as the dissemination node attribute F _vec of the information; Step B2: Extract the network topology structure of the constructed social network information dissemination network G(V, E), and extract the network dissemination structure attribute F _struc , including but not limited to: (1) The ratio r _2/1 of the initial forwarding level of the propagation network G(V, E), that is, formula (1), that is, the number of nodes in the second-layer propagation network n _V (2) and the number of nodes in the first-layer propagation network n _V (1) ratio;

(2) The characteristic distance a of the nodes of the propagation network G(V, E) is used to fit the distance distribution between all pairs of nodes in the propagation network. The specific fitting equation is shown in formula (2), where y is the distribution probability, x is the separation distance of the node pair;

(3) The homogeneity index h of the propagation network G(V, E) is the difference between the logarithm value of the homogeneity of the propagation network G(V, E) and the homogeneity of the star network of the same scale. The calculation method is as follows: 3), where the homogeneity calculation method of the propagation network is as in formula (4); h=log(H _star )-log(H _G ) (3)

where N is the total number of network nodes, and k is the node degree (4) Step B3: Set the credibility evaluation index of the historical social network information Event _history according to the summary of historical facts and the certification of the authority

The value of 0 indicates that the trust is completely unreliable, and the value of 1 indicates that the trust is completely credible. Set the credibility evaluation indicators of the dissemination nodes of the historical society update network information Event _history , Credibility _vec (Event _history ), Credibility _struc (Event _history ) is the same as the comprehensive credibility evaluation index; Step B4: For the collected Num pieces of historical social network information Event _history , the historical information dissemination node attribute F _vec (Event _history ), the historical information dissemination network structure attribute F _struc (Event _history ), and the historical information credibility index Credibility ( Event _history ) is added to the historical social network information dissemination attribute database DS (Event _history ). 4. a kind of social network information credibility assessment method based on the similarity of propagation attributes according to claim 1, is characterized in that: In step C, "calculate the similarity of the propagation attributes of the target social network and the historical social network, so as to evaluate the information credibility of the target social network", the specific method is as follows: Calculate the target social network information Event _obj and the history The dissemination node similarity Sim _vec of the social network information Event _history and the credibility of the dissemination node Credibility _vec (Event _obj ) of the target social network information Event _obj are obtained, and then the target social network information Event _obj and the The similarity quantile of the dissemination structure of the historical social network information Event _history , and then obtain the credibility of the dissemination structure Credibility _struc (Event _obj ), assign the credibility of each dissemination attribute to calculate the weight, and calculate the credibility of the target social network information; The specific steps of the process are as follows: Step C1: Using a set similarity calculation method such as the Jaccard similarity method, that is, formula (5), respectively calculates that the target social network information Event _obj is similar to the dissemination node attributes of all historical information Event _history in the historical information dissemination attribute database DS (Event _history ). Sim _vec (Event _obj , Event _history ), then select the largest propagation user attribute similarity value Sim _{vec_max} and its corresponding historical information Event _{vec_max} in the calculation result, and record the credibility of the corresponding historical information Credibility _vec (Event _{vec_max} ), then the credibility of the propagation node of the target social network information Event _obj (Credibility _vec (Event _obj )=Sim _{vec_max} ×Credibility _vec (Event _{vec_max} );

In the formula: F _vec (Event _obj ) represents the attribute of the target information dissemination node, and F _vec (Event _history ) represents the attribute of the historical information dissemination node; Step C2: For each historical information Event _history in the historical information dissemination attribute database DS (Event _history ), based on the size of the M sub-attributes in the dissemination structure attribute F _struc , sort them according to the same sorting method, and obtain M historical information after sorting sequence:

Step C3: For the i-th rearranged historical information sequence obtained by C2, divide the historical information into K intervals with equal numbers, then each interval has

and then compare the corresponding i-th dissemination structure sub-attribute values of the historical information Event _history in the historical information dissemination attribute database DS (Event _history ), find the most similar historical information Event _history and use the interval number k _{struc_i} to which it belongs as the target Similarity _quantiles of the sub-attributes of the communication structure of the social network information Event _obj , repeat this step for i=1, 2, . similarity quantile; Step C4: Calculate the propagation structure sub-attribute Credibility _{struc_i} of the target social network information Event _obj and the historical social network information Event _history according to formula (6), where i is the ith in the M propagation structure sub-attributes,

Respectively, in the quantile k _i of the sub-attribute of this propagation structure to which the target social network information Event _obj belongs, the number of historical information Event _history whose reliability is 0 and 1;

Step C5: _{Allocate calculation weights w vec ,} w _{struc_1 , .} The sum of all weights is equal to 1, and the distribution method can adopt the average distribution method, the analytic hierarchy process and the fuzzy evaluation method, and distribute according to the weight importance and other necessary information; Step C6: Calculate the credibility of the target social network information by using the formula (7), and evaluate the credibility of the target social network information according to the result. The closer the calculation result is to 0, the weaker the credibility, and vice versa. 1, the more reliable it is;

Images