A method for analyzing user influence in social networks based on relationship strength calculation

Abstract

The present invention discloses a method for analyzing the influence of social network users based on relationship strength calculation, comprising the following steps: S1, relationship strength calculation, calculating the relationship strength of social network users from three dimensions of interactive behavior types, the number of occurrences of interactive behavior types and the interactive behavior time between virtual identities in the social network, and using the calculated relationship strength to discover the closeness and distance between users in the social network and the emotional tendency; S2, combining the relationship strength between social network users calculated in step S1 with the LeaderRank algorithm to calculate the social network influence of social network users, etc. The present invention can better discover the closeness and distance between user pairs in the network, the connection strength relationship, and the better calculate the social network influence of users, thereby improving the accuracy of the algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of social network data analysis, and more specifically, to a method for analyzing the influence of social network users based on relationship strength calculation. Background Art

[0002] With the birth and development of various social networks, people around the world can use various virtual identities to share, comment or discuss any topic at any time. People in social networks are not only users of information, but also creators of information. The explosive growth of social network data has brought many opportunities for data mining and knowledge discovery. A large number of social network-related studies have gradually emerged, including network community discovery, user influence evaluation and network information dissemination.

[0003] User influence analysis in social networks is one of the most important research points. It aims to evaluate the influence of each user on other users in the social network and select a group of users who are most conducive to the spread of influence. It plays an important role in saving resource allocation, accelerating information diffusion, and suppressing the spread of inflammatory and false information.

[0004] To sum up, conducting relevant analysis on the influence of social network users has certain scientific research value and research significance for real life. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a social network user influence analysis method based on relationship strength calculation, which can better discover the closeness and connection strength between user pairs in the network, and then incorporate this strength information into the traditional LeaderRank algorithm, so as to better calculate the user's social network influence and improve the accuracy of the algorithm.

[0006] The object of the present invention is achieved through the following solutions:

[0007] A method for analyzing the influence of social network users based on relationship strength calculation includes the following steps:

[0008] S1, relationship strength calculation, calculates the relationship strength of social network users from three dimensions: the interaction behavior type, the number of occurrences of the interaction behavior type, and the interaction behavior time between virtual identities in the social network. The calculated relationship strength is used to discover the closeness and emotional tendency between users in the social network;

[0009] S2, combining the relationship strength between social network users calculated in step S1 with the LeaderRank algorithm to calculate the social network influence of the social network users.

[0010] Furthermore, in step S1, according to the interaction behavior type, the following formula is constructed to calculate the first relationship strength:

[0011]

[0012] Where g is the impact of each interaction behavior type on the relationship strength, g max Is user u i and user u j The strongest interaction between them has an impact on the strength of the relationship, where ω is a constant.

[0013] Further, in step S1, according to the number of occurrences of the interactive behavior type Construct the following formula to calculate the second relationship strength under this interactive behavior type

[0014]

[0015] Furthermore, in step S1, according to the interactive behavior time T i,j , construct the following formula to calculate the third relationship strength

[0016]

[0017] Where year(T i,j ) indicates the year in which the interaction occurred, year_base indicates the current year, Indicates the number of times each type of interaction between users occurs in a certain year.

[0018] Further, the steps include:

[0019] After calculating the impact of different types of interactive behaviors in different years on the relationship strength, the relationship strength R between the two users is obtained by accumulation. i,j , the expression is as follows:

[0020]

[0021] Based on the above expression, the time of the last interaction behavior is incorporated to reflect whether the users are still in contact. The closer the last interaction behavior is, the stronger the relationship strength is. The expression is as follows:

[0022]

[0023] in The influence of the time of the last interaction on the strength of the relationship.

[0024] Furthermore, in step S2, the node v is calculated according to the following formula:i The PR value at time t is:

[0025]

[0026] Among them, t represents time, parameter w ji is a node v in a directed weighted graph i Points to node v j The edge weight is equal to the calculated relationship strength R of the social network users. i,j , For node v j The out-degree of ji Indicates whether there is an edge from j to i. If so, then a ji =1, otherwise a ji =0,LR i (t) represents the LeaderRank score value at the tth moment.

[0027] The beneficial effects of the present invention are:

[0028] The user relationship strength calculation formula constructed by the present invention can better discover the closeness and connection strength between pairs of users in the network, and then incorporate this strength information into the traditional LeaderRank algorithm, which can better calculate the user's social network influence and improve the accuracy of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0030] Figure 1 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0031] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.

[0032] like Figure 1 As shown, the social network user influence analysis method based on relationship strength calculation includes the following steps:

[0033] S1, relationship strength calculation, calculates the relationship strength of social network users from three dimensions: the interaction behavior type, the number of occurrences of the interaction behavior type, and the interaction behavior time between virtual identities in the social network. The calculated relationship strength is used to discover the closeness and emotional tendency between users in the social network;

[0034] S2, combining the relationship strength between social network users calculated in step S1 with the LeaderRank algorithm to calculate the social network influence of the social network users.

[0035] The present invention proposes a method for calculating the influence of users in social networks based on the interactive information between users. This method deeply studies the interactive behavior between users, formulates and distributes questionnaires to various types of users in social networks, and constructs a user relationship strength function based on the collected real user scoring data. It can better discover the closeness and connection strength between user pairs in the network, and then integrate this strength information into the traditional LeaderRank algorithm to calculate the influence of users in the group. The method mainly includes the following two modules.

[0036] 1. Social network user relationship strength calculation module

[0037] Social network accounts are virtual identities that people create on the Internet to express themselves, connect, and share information with each other. The social network identity user relationship strength calculation module analyzes the interactive information such as replies, comments, forwarding, and likes contained in the information release attributes of the network virtual identity to perform social group mining such as character clustering, and combines real user scoring data to analyze the relationship strength between social network users from three dimensions: the type of interactive behavior of the virtual identity, the number of interactive behaviors, and the time of interactive behaviors. All the expressions constructed below are fitted based on a large amount of user survey data and are consistent with the real data distribution. Three decisive interactive information can be observed in this module:

[0038] Interaction behavior type

[0039] The stronger the interactive behavior, the stronger the relationship strength of the virtual identity. For example, user A's comment on user B's status is a more active and targeted behavior than user A's praise of user B's status. This stronger interactive behavior indicates that the relationship strength between the two users is stronger. Therefore, user u i and user u j The relationship strength R i,j The value depends on the interaction type G i,j .

[0040] By analyzing the operation characteristics of each social network, it can be found that for the same status, there are the following types of interactions between users: commenting, forwarding, liking, @ing each other's users, etc. The present invention defines an intensity level for each interactive behavior, as shown in the following table:

[0041] Table 1 Types of interactive behaviors between users

[0042]

[0043]

[0044] If there is only "weak" interaction between two users, such as liking the same status, it is possible that the two users are not familiar with each other or do not know each other, but just have common friends or follow the same topics, etc. If another group of users has both stronger interaction behaviors such as forwarding and @ing each other's users, as well as "weak" interaction behaviors, it means that this group of users is closer in the social network and is more likely to have a close relationship in real life.

[0045] Therefore, the present invention not only focuses on the single interaction behavior type of two users, but also focuses on the most intimate interaction behavior type between the two users. The stronger the most intimate interaction behavior, the greater the impact of their interaction behavior on the relationship strength. The expression for calculating the relationship strength according to the interaction behavior type is as follows:

[0046]

[0047] Where g is the impact of each interaction behavior type on the relationship strength, g max Is user u i and user u j The strongest interaction between them affects the relationship strength, where ω is a constant, which is set to 0.25 here. The calculated relationship strength is The value is between 0 and 1.

[0048] Number of interactive behaviors

[0049] The more frequent the interactions between users, the stronger their relationship strength is. i and user u j The number of interactions between i,j Affects the relationship strength value R i,j .

[0050] The present invention counts the number of occurrences of each type of interaction between users. Calculate the relationship strength under this interaction type The expression is as follows:

[0051]

[0052] Interaction behavior time

[0053] The strength of the relationship between users is not static, but changes over time. For example, the impact of interaction behavior 10 years ago on the strength of the relationship is significantly weaker than that of interaction behavior in recent times. i,j It also affects the relationship strength value R i,j The present invention adds a time factor into the calculation expression, as follows:

[0054]

[0055] Where year(T i,j ) indicates the year in which the interaction occurred, and year_base indicates the current year. Indicates the number of times each type of interaction between users occurs in a certain year.

[0056] Relationship Strength

[0057] After calculating the impact of different types of interactive behaviors in different years on the relationship strength, the relationship strength R between the two users is obtained by accumulation. i,j , the expression is as follows:

[0058]

[0059] Based on this expression, the present invention incorporates the time when the last interaction occurred to reflect whether the users are still in contact. The closer the time when the last interaction occurred, the stronger the relationship strength. The expression is as follows:

[0060]

[0061] in The influence of the time of the last interaction on the strength of the relationship.

[0062] 2. Social network user influence calculation module

[0063] The user relationships of social networks are directed, so their network structure can be represented by a directed weighted graph. Each user is a node in the graph, and the relationships between users are edges in the graph. The relationship strength calculated by the previous module is the weight on the edge. If user u pays more attention to user v, that is, the relationship strength value is greater, then u will give its own influence to v. If u's influence is high, then v will get more influence from u. Based on this, this module improves the traditional LeaderRank influence ranking algorithm, incorporates the user's emotional tendencies, and further improves the accuracy of the algorithm.

[0064] PageRank

[0065] PageRank is a classic web page ranking algorithm based on the link structure of web pages. The algorithm has such a cognition: the importance of a page in the World Wide Web depends on the number and quality of other pages pointing to it. If a page is pointed to by many high-quality pages, the quality of this page is also high. At the initial moment, each node (web page) is assigned the same PR value, and then iterates. Each iteration divides the current PR value of each node equally among all the nodes it points to. The new PR value of each node is the sum of all the PR values ​​it has obtained, so the node v i The PR value at time t is:

[0066]

[0067] in For node v j The out-degree of a ji Indicates whether there is an edge from j to i. If so, then a ji =1, otherwise a ji = 0. c is the jump probability. When a user visits a page, he can randomly jump to other web pages through the address bar with probability c, and jump to other pages through the hyperlinks in the web page with probability 1-c. The above formula is iterated until the PR value of each node reaches a stable state.

[0068] LeaderRank

[0069] In the PageRank algorithm, the probability of random jumps to each node is the same, that is, starting from any web page, the probability of accessing other web pages by entering the URL is equal. However, in reality, this operation does not conform to people's daily habits. When people browse popular web pages with rich content, the probability of using the address bar to jump to the page is much lower than that of web pages with less information. In order to effectively solve this problem, the LeaderRank algorithm replaces the jump probability c in the PageRank algorithm by adding a background node (Ground node) and bidirectional edges between this node and all nodes in the network during the random walk of the directed network, thereby obtaining a parameter-free algorithm that is simpler and more elegant in form.

[0070] At the initial moment, except for the background node v g Allocate unit resources to other nodes other than LR i (0) = 1; LR a (0) = 0. The following iteration process is performed until the steady state is reached:

[0071]

[0072] Note that during the iteration, the adjacency matrix is ​​n+1 order (including background nodes). In steady state, the score value LR of the background node is g (t c ) is equally divided among the other n nodes, so we get node v i The final LeaderRank score is:

[0073]

[0074] The LeaderRank algorithm has excellent performance in evaluating the influence of nodes in social networks, hence its name. Compared with the traditional PageRank, the LeaderRank algorithm has the following advantages: (1) the addition of background nodes reduces the radius of the entire network, and converges faster than PageRank; (2) it can better identify influential nodes in the network, and the mined important nodes can spread the network flow faster and wider; (3) it is more robust than PageRank in resisting spam user attacks and random interference. These advantages have made the LeaderRank algorithm widely concerned.

[0075] However, the original LeaderRank algorithm only considers the relationship between nodes (0 or 1), that is, the structure of the network. The LR value of each node is evenly distributed to the nodes that the node follows, and the emotional tendencies between users are not taken into account, which may lead to errors in the calculated influence ranking.

[0076] Improved LeaderRank

[0077] User sentiment tendency, i.e., the strength of the relationship between users, is an indispensable factor in calculating the influence ranking of users in the network. Therefore, in view of the shortcomings of the traditional LeaderRank algorithm, the present invention combines the strength of the relationship between users with the traditional LeaderRank to further improve the accuracy of the algorithm. i The PR value at time t is:

[0078]

[0079] Among them, the newly added parameter w ji is a node v in a directed weighted graph i Points to node v j The weight of the edge is equal to the user relationship strength R calculated in the previous module. i,j .

[0080] The data set of the embodiment of the present invention comes from the tweet information of 15 public homepages, 12 personal homepages and the homepages of friends of these personal users crawled by Facebook, including tweet content, URL, time, name, ID, number of shares, @users, comments, likes information, etc. The database uses Neo4j graph database.

[0081] For the social network user relationship strength estimation module, this technology calculates the relationship strength value of the user pair based on the interaction behavior type, number of interaction behaviors and interaction behavior time between users in the database. There may be two directed lines between two virtual identities and the relationship strength scores on them may be different. This is because the relationship strength between the two virtual identities is asymmetric. For example: For every tweet posted by A, B has comments, likes or reposts. But A never has any interaction with the tweets posted by B. Then it can be seen that A has a great influence on B, and the relationship strength is also great. But B has little influence on A, and the relationship strength is also small.

[0082] For the social network user influence estimation module, this technology uses PageRank, LeaderRank and improved LeaderRank to rank the user influence on the crawled Facebook data. The following table shows the ranking results of the top 10 influential users.

[0083] Table 2 Influence ranking results calculated by three methods

[0084]

[0085] The first row and second column in the table indicate that the user ID ranked first in influence using the PageRank algorithm is '100019059701459', and so on for the other columns. From the table, we can see that the rankings obtained by the three methods have overlapping areas, including {'100019059701459', '611202736', '100001718256387', '100000300156957'}. In addition, there are also non-overlapping areas. For example, the improved LeaderRank algorithm calculates the top ten user IDs {'100001868401625', '100004316735963', '125474120835697'} which cannot be obtained by the other two algorithms. In response to the above phenomenon, we specifically checked the reasons for this difference and found that {'100001868401625', '100004316735963'} users have stronger interaction types and higher interaction frequencies than {'100000300156957', '100001718256387'} users. Of course, they have greater influence in social networks. This proves that the improved LeaderRank algorithm that takes user interaction information into account can better calculate the user influence of social networks.

[0086] In addition to the above examples, those skilled in the art may obtain other embodiments based on the above disclosure or by using the knowledge or technology in the relevant field to make changes. The features of each embodiment may be interchangeable or replaced. The changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention and should be within the scope of protection of the claims attached to the present invention.

[0087] If the functions of the present invention are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium, and all or part of the steps of the methods described in various embodiments of the present invention are executed in a computer device (which can be a personal computer, a server, or a network device, etc.) and corresponding software. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, or an optical disk, and the test or actual data exists in a read-only memory (Random Access Memory, RAM), a random access memory (Random Access Memory, RAM), etc. in the program implementation.

Claims

1. A method for analyzing the influence of social network users based on relationship strength calculation. It is characterized in that Includes steps: S1, relationship strength calculation, calculates the relationship strength of social network users from three dimensions: the interaction behavior type, the number of occurrences of the interaction behavior type, and the interaction behavior time between virtual identities in the social network. The calculated relationship strength is used to discover the closeness and emotional tendency between users in the social network; S2, combining the relationship strength between social network users calculated in step S1 with the LeaderRank algorithm to calculate the social network influence of the social network users; In step S1, according to the interaction behavior type, the following formula is constructed to calculate the first relationship strength: Where g is the impact of each interaction behavior type on the relationship strength, g max Is user u i and user u j The strongest interaction between them has an impact on the relationship strength, where ω is a constant; In step S1, according to the number of occurrences of the interactive behavior type Construct the following formula to calculate the second relationship strength under this interactive behavior type In step S1, according to the interactive behavior time T i,j , construct the following formula to calculate the third relationship strength Where year(T i,j ) indicates the year in which the interaction occurred, year_base indicates the current year, Indicates the number of times a certain type of interaction between users occurred in a certain year; After calculating the impact of different types of interactive behaviors in different years on the relationship strength, the relationship strength R between the two users is obtained by accumulation. i,j , the expression is as follows: Based on the above expression, the time of the last interaction behavior is incorporated to reflect whether the users are still in contact. The closer the last interaction behavior is, the stronger the relationship strength is. The expression is as follows: in The time of the last interaction behavior affects the strength of the relationship; In step S2, the node v is calculated according to the following formula i The PR value at time t is: Among them, t represents time, parameter w ji is a node v in a directed weighted graph i Points to node v j The edge weight is equal to the calculated relationship strength R of the social network users. i,j , For node v j The out-degree of ji Indicates whether there is an edge from j to i. If so, then a ji =1, otherwise a ji =0,LR i (t) represents the LeaderRank score value at the tth moment.

Images