An influence blocking maximization algorithm based on overlapping interest community detection

Abstract

This invention proposes an influence blocking maximization algorithm based on overlapping interest community detection. It includes a multi-factor competitive propagation model and an algorithm based on this model and overlapping interest community detection, used for modeling the competitive propagation process of multiple information in social networks and selecting positive information seed nodes. Modeling the competitive propagation process and maximizing influence blocking are hot topics in social network analysis. Traditional models, based on IC and LT, are difficult to accurately depict the real-world propagation process; traditional community-based algorithms divide communities based on topological structure, ignoring user interests and preferences. This invention comprehensively considers user interests, information interaction latency between users, and trust levels, enabling better modeling of the competitive propagation process; it extends the traditional community concept to interest communities, improving community detection accuracy; and it guides the selection of positive information seed nodes through the overlapping interest community structure, achieving the desired negative information influence blocking effect in a short time.

Description

Technical Field

[0001] This invention belongs to the field of social network analysis, and specifically proposes a competitive propagation model that considers multiple factors and an influence blocking maximization algorithm based on this model and overlapping interest community detection. Background Technology

[0002] In recent years, social media platforms such as Weibo, Facebook, and Twitter have developed rapidly, with a continuous increase in registered and active users. Users establish social connections through social behaviors such as following each other, thus building massive social networks. The emergence of social networks has facilitated information interaction and sharing, resulting in rapid and wide-ranging information dissemination. Given that user decisions are influenced by their neighbors, targeting a small group of users within a social network and leveraging the complex social connections among them can achieve widespread information dissemination, giving rise to the problem of Information Maximization (IM). While the rapid development of social networks brings convenience, some negative information also spreads rapidly and widely. Given the urgency of containing the influence of negative information, it is necessary to minimize its impact within a certain time constraint. Against this backdrop, a variant of the IM problem, the Influence Blocking Maximization under Time constraint (T-IBM) problem, has gained attention and been widely studied. The problem is described as follows: Given a set of seed nodes for negative information that has been spreading in a social network for some time, seed nodes are selected to publish positive information in the social network. The two compete to spread the information in the social network according to a specified propagation model. Within a given time constraint, the goal is to make the positive information dominate the propagation. This involves considering both the inhibitory effect of the positive information on the negative information and the influence range of the positive information itself.

[0003] The T-IBM problem involves the competitive propagation of positive and negative information in social networks. Traditional single-information propagation models (independent cascade model, linear threshold model) are no longer applicable. A multi-information propagation model is needed to describe the simultaneous competitive propagation of multiple messages in a social network. Existing multi-information propagation models are mainly extensions of the classic single-information propagation model: namely, multi-information propagation models based on independent cascade models and multi-information propagation models based on linear threshold models. However, in some social network environments: users may be interested in specific topics, and the degree of interest in different topics varies, resulting in a certain topic bias in the information propagated in the social network; when a neighbor forwards a message to a user, the user does not receive it immediately, but there is a certain time delay; trust is established between users through information interaction, requiring a trust calculation mechanism to assess the degree of trust between users. Furthermore, social networks typically exhibit a community structure: close connections between users within the community and sparse connections with users outside the community.

[0004] For solving the T-IBM problem, there are currently some algorithmic studies that block the spread of negative information by publishing positive information. Similar to the IM problem, these can be divided into three categories: approximate algorithms that can provide performance guarantees but whose efficiency needs improvement; community-based algorithms that effectively utilize community structure to improve scalability; and heuristic algorithms that are highly efficient but cannot provide performance guarantees. Existing community-based algorithms usually only consider the topology of social networks, ignoring the influence of user interests and preferences on the formation of community structure. Communities that consider interest preferences can be defined as interest communities: users within an interest community are closely connected and have high similarity in interests and preferences, while users outside the interest community are sparsely connected and have low similarity in interests and preferences; at the same time, users may belong to multiple interest communities, meaning that interest communities may overlap.

[0005] In summary, existing multi-information competitive propagation models do not comprehensively consider user interests and preferences, information interaction latency between users, and trust levels among users. Furthermore, community-based algorithms ignore the impact of user interests and preferences on the structure of communities (interest-based communities) in social networks. Therefore, researching a general competitive propagation model that closely resembles real-world social networks and an interest-based community-based algorithm has significant practical implications. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes an influence blocking maximization algorithm based on overlapping interest community detection. This algorithm comprises a multi-factor competitive propagation model and an influence blocking maximization algorithm based on this model and overlapping interest community detection. It improves the generality of the competitive propagation model and the accuracy of community detection, guiding the selection of positive information seed nodes to effectively suppress the propagation of negative information under given time constraints.

[0007] Firstly, a competitive propagation model considering multiple factors is provided, including: social network modeling, which mathematically describes users in the social network, social connections between users, and the heterogeneous characteristics of the social network to obtain a social network model; calculating the trust degree between users based on the heterogeneous characteristics of the social network to improve the social network model; defining all possible states of users in the social network and the transition mechanism between states based on the social network model; and describing the multi-information competitive propagation process based on the social network model and the user state transition mechanism.

[0008] In one optional implementation, the heterogeneity features exhibited by the social network include: user heterogeneity and social connection heterogeneity. User heterogeneity includes user interests and preferences, modeled using topic preference vectors. Social connection heterogeneity includes the latency of information interaction between users, the number of information interactions, and the level of trust between users. The latency of information interaction between users is determined by the encounter probability m. uv Modeling, m uv This represents the probability that user u and user v meet at any time step. After user u forwards information to v, the corresponding meeting probability is checked at each time step. When the meeting probability is satisfied, user u and user v meet, that is, user v has received the information forwarded by user u.

[0009] In one optional implementation, the calculation of trust levels between users based on the heterogeneity characteristics of the social network is shown in the following formula:

[0010] t uv =sx_t uv +s uv +reli uv

[0011] Among them, t uv sx_t represents the degree of trust that user v has in its neighbor u. uv The level of trust built on familiarity depends on the frequency of information exchange between users. The more frequent the information exchange, the more familiar the users are with each other, and the higher the level of trust built on familiarity. uv The trust level generated by similarity depends on the similarity of interests and preferences between users. Higher similarity indicates more common ground between users, making them more receptive to information forwarded by each other, thus leading to higher trust. If user v does not interact frequently with neighbor u, but neighbor u itself has high trustworthiness (manifested by extremely frequent interaction with some neighbors or maintaining interaction with a large number of neighbors), then user v is also inclined to trust neighbor u. uv This represents the degree of trustworthiness of neighbor u for user v, and its calculation depends on the information interaction between neighbor u and other neighbors.

[0012] In one optional implementation, defining all possible user states in a social network and the transition mechanism between states includes: defining the possible user states based on the user's acceptance and forwarding of information in the social network: ① neither accepting nor forwarding, ② accepting a message but not forwarding, ③ accepting and forwarding a message. States ② and ③ need to be defined differently for different types of information. For the sake of uniformity, accepting and forwarding information are separated here, i.e., accepting ≠ forwarding, because in real life, users also have certain interest preferences when forwarding information. The state transition mechanism should include: when a user is in state ①, if they receive a message, what is the probability that the user will accept the message and forward it? If they receive multiple messages, which message will the user choose to accept and forward? When a user is in state ②, if they receive another message that they have already accepted, what is the probability that they will choose to forward it? If they receive other messages besides those they have already received, will they accept the new message and forward it? All probabilities need to be calculated by comprehensively considering the user's interest preferences and the level of trust between users.

[0013] In one optional implementation, the description of the multi-information competition propagation process based on the social network model and user state transition mechanism includes: the seed node set of each piece of information has a node state of ③, and the other nodes have a state of ①; the multi-information competition propagation process unfolds in a discrete manner, at each time step, the node that has newly transitioned to state ③ in the previous time step forwards the corresponding information to all neighboring nodes that are not in state ③, and the information arrives at the neighboring nodes after a corresponding time delay. After receiving the information, the neighboring nodes make a state transition decision based on the information forwarded by other nodes; when the preset time constraint is reached or the termination condition is met (no node has newly transitioned to state ③ and there are no nodes waiting to make a decision), the information propagation process ends.

[0014] Secondly, an influence blocking maximization algorithm based on the multi-factor competitive propagation model and overlapping interest community detection is provided, including: identifying a complete subgraph composed of users with high centrality and similarity of interests as an initial community based on the social network model; detecting overlapping interest community structures in the social network based on label propagation based on the social network model and the initial community; extracting overlapping and non-overlapping nodes based on the social network model and the overlapping interest community structure and constructing a candidate seed node set according to influence values; and selecting a final seed node set based on the multi-factor competitive propagation model and the candidate seed node set.

[0015] In one optional implementation, identifying a complete subgraph composed of users with high centrality and similarity in interests as an initial community includes: arranging nodes in the social network in descending order of centrality and visiting them sequentially to construct an initial community; if the centrality of the currently visited node is greater than the average centrality, adding it to the current initial community; evaluating the importance of neighboring nodes to the currently visited node based on their similarity in interests and centrality, selecting the neighboring node with the highest importance to join the initial community, and retaining the common neighboring nodes of the currently visited node and the newly added neighboring nodes; continuously searching for the node with the highest importance to the current initial community from the common neighboring nodes to join the initial community and deleting non-common neighboring nodes, until the set of common neighboring nodes is empty.

[0016] In one optional implementation, the label-based propagation detection of overlapping interest community structures in a social network includes: assigning a unique label to the initial community, with nodes within the initial community sharing the same label and an affiliation coefficient of 1; in each iteration, updating the label set of each node in the current iteration based on the label set of neighboring nodes from the previous iteration, until the number of nodes belonging to each label reaches its minimum and no longer changes. The step of updating the label set of each node in the current iteration based on the label set of neighboring nodes from the previous iteration includes: summing all labels appearing in the label sets of neighboring nodes and adding their corresponding affiliation coefficients, and weighting the affiliation coefficients according to the strength of the relationship between the node and the community corresponding to the label (the number of social connections between the node and the community) and the similarity of interest preferences.

[0017] In one optional implementation, the step of extracting overlapping and non-overlapping nodes and constructing a candidate seed node set based on influence values ​​includes: extracting overlapping and non-overlapping nodes based on the overlapping interest community structure, and calculating their respective influence values ​​on the information dissemination process, wherein overlapping nodes are simultaneously associated with multiple communities (C v (A set of communities connected to overlapping nodes) is located at the hub of information dissemination among multiple communities, and its influence value depends on the strength of its relationship with each community. v→c And the average value s of the similarity of interests and preferences with each node in the community. B→c ;

[0018]

[0019] Non-overlapping nodes are located within the community. Consider the impact of their two-hop neighbors on the information propagation process, including trust values ​​and information interaction latency.

[0020] In summary, compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] 1) This invention proposes a competitive propagation model that considers multiple factors, comprehensively analyzing the impact of user interests and preferences, information interaction delays between users, and trust between users on information propagation. In particular, it proposes a novel method for calculating trust between users by combining the principles of trust generation between people in social psychology. By reasonably designing the node state transition mechanism, the process of multiple messages competing for propagation in a social network can be effectively modeled.

[0022] 2) This invention uses tag propagation to detect overlapping interest communities, extending the traditional overlapping community detection to overlapping interest community detection, and effectively utilizing user interest preferences to improve detection efficiency and accuracy.

[0023] 3) This invention proposes an influence blocking maximization algorithm based on a competitive propagation model and overlapping interest community detection. Based on the detection of overlapping interest community structure in 2), it analyzes the role and influence of overlapping and non-overlapping nodes, proposes different influence calculation methods, and selects positive information seed nodes. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. Wherein:

[0025] Figure 1 This is a node state transition diagram in the competitive propagation model considering multiple factors in this invention;

[0026] Figure 2 Here is a flowchart of the influence blocking maximization algorithm based on overlapping interest communities in this invention:

[0027] Figure 3 This is a flowchart of the initial community identification algorithm in this invention;

[0028] Figure 4 This is a flowchart of the overlapping interest community detection algorithm based on tag propagation in this invention;

[0029] Figure 5 The results show the performance improvement of this invention compared to commonly used heuristic algorithms on four real-world social network datasets. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The embodiments are only used to illustrate and explain the present invention, and do not limit the present invention. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0031] Figure 1 This paper illustrates the node state transition diagram in the competitive propagation model considering multiple factors according to the present invention. For example, it assumes two pieces of information, A and B, are simultaneously competing for propagation in a social network (this can be extended to more than two pieces of information). The model defines that a node in a social network may be in the following five states:

[0032] inactive: The node has not received, accepted, or forwarded any information;

[0033] accept-A(accept-B): The user receives message A(B) and accepts the message but does not forward it;

[0034] active-A(active-B): User forwards information A(B).

[0035] For example, in combination Figure 1 Taking information A (and similarly information B) as an example, the state transition process of a node is introduced:

[0036] inactive->accept-A->active-A: Initially, the node is in an inactive state. Influenced by information A, it needs to determine whether to accept and forward information A. Consider the combined trust level of all neighboring nodes that forward information A, as well as the node's topic compatibility with information A. If simultaneously influenced by information A and information B, they must be considered separately.

[0037] accept-A -> active-A: Initially, the node is in the accept-A state, meaning that when the node accepts information A, its topic preference for information A alone is insufficient to make it forward information A. Now, it is again influenced by information A, and it needs to determine whether to forward information A. This takes into account the number of neighboring nodes that have forwarded information A to the node up to this point, i.e., the cumulative impact.

[0038] accept-A->accept-B->active-B: Initially, the node is in the accept-A state. Influenced by information B, it needs to determine whether to switch to accepting information B and further forward it. First, the topic compatibility between the node and information A and B is compared. If the topic compatibility with information A is higher, to make the node switch to accepting information B, the combined trust level of all neighboring nodes that forward information B needs to compensate for the difference in topic compatibility between information A and B; otherwise, it's a gain.

[0039] Figure 2 The algorithm for maximizing the impact blocking based on the multi-factor competitive propagation model and overlapping interest community detection is presented. The algorithm consists of four stages: the first stage is initial community identification; the second stage is overlapping interest community detection based on label propagation; the third stage is constructing a candidate seed node set; and the fourth stage is selecting the final seed node set.

[0040] In the first stage, based on the aforementioned social network model, a complete subgraph composed of users with high centrality and similar interests is identified as the initial community. For example, node degree is used as the centrality measure. First, all nodes in the social network are sorted in descending order of degree and visited sequentially. If a node's degree is greater than the average degree of nodes in the social network, it is added to the current initial community, and the node with the highest importance to the currently visited node among its neighbors is selected and added to the current initial community. A common neighbor node set for the current initial community is constructed. If the common neighbor node set is not empty, the node with the highest importance to the current initial community is found in the common neighbor node set and added to the current initial community. The common neighbor node set is updated, and the above process is repeated until the common neighbor node set is empty. Finally, it is determined whether the number of nodes in the current initial community is greater than 3 (if the number of nodes is too small, it is not representative). If so, the nodes are added to the initial community set.

[0041] In the second stage, based on the aforementioned social network model and initial communities, overlapping interest community structures in the social network are detected using label propagation. First, a unique label is assigned to each initial community, with nodes within the initial community sharing the same label and having an attribution coefficient of 1. The node order is shuffled, and a random access sequence is obtained, with nodes accessed sequentially. The label set of the currently accessed node in this iteration is updated based on the label set of its neighbors from the previous iteration, including: summarizing all labels appearing in the label set of the currently accessed node's neighbors and adding the corresponding attribution coefficients; the attribution coefficients are weighted according to the relationship strength and interest preference similarity between the currently accessed node and the communities corresponding to each label, resulting in a weighted attribution coefficient; the maximum weighted attribution coefficient is recorded, and labels with a weighted attribution coefficient / maximum weighted attribution coefficient < a given threshold are deleted; if all nodes have been visited, it is determined whether the termination condition is met: the minimum number of members for each interest community label remains unchanged between the two iterations; if this condition is met, nodes holding the same label are added to the same interest community. Some nodes may hold multiple labels and belong to multiple interest communities, thus forming an overlapping interest community structure.

[0042] In the third stage, based on the aforementioned social network model and overlapping interest community structure, overlapping and non-overlapping nodes are extracted, and a candidate seed node set is constructed according to their influence values. The influence value of an overlapping node depends on the node's relationship with its connected community C. v Relationship strength r v→c and topic preference similarity s B→c .

[0043]

[0044] For example, the relationship strength is obtained by comparing the number of edges connecting the overlapping node to the community with the total number of edges of the overlapping node and the total number of edges of the nodes in the community. The topic preference similarity is the average of the topic preference similarity between the overlapping node and all nodes in the community. For non-overlapping nodes, the trust level and encounter probability of their two-hop neighbors are considered. The higher the trust level, the more likely the neighboring nodes are to accept the forwarded information and expand the spread. The higher the encounter probability, the shorter the information interaction delay, so as to achieve a larger spread in a shorter time.

[0045] In the fourth stage, the final seed node set is selected based on the multi-factor competitive propagation model and the candidate seed node set. For example, a greedy strategy is used to continuously select the node with the largest marginal gain from the candidate seed node set and add it to the final seed node set. Regarding the calculation of the influence range, Monte Carlo simulation is used, inputting the seed node sets of each piece of information into the multi-factor competitive propagation model, running it 1000 times, and calculating the average value.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some or all of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the protection scope of this invention.

Claims

1. A method for maximizing public opinion suppression based on overlapping interest communities, characterized in that, This method involves constructing a competitive propagation model that considers multiple factors, the competitive propagation model being constructed based on the following social network heterogeneity features: (1) Model social networks based on user interests and preferences, the latency of information interaction between users, and the number of information interactions; (2) Calculate the trust level between users based on the heterogeneity characteristics of social networks, and the trust level t of user v to neighbor u. uv Calculated as: t uv =sx_t uv +s uv +reli uv ;where sx_t uv The degree of trust derived from familiarity, s uv Reli represents the level of trust generated by similarity. uv This indicates the degree of trustworthiness of neighbor u for user v. (3) Based on the improved social network model, define the user's state in the social network and the state transition mechanism between states. The state transition mechanism defines the possible states of a user based on the user's information reception and forwarding behavior in the social network: ① Do not accept or forward; ② Accept a message but do not forward it; ③ Accept a message and forward it. Here, accepting the message and forwarding the message are separated, that is, accepting ≠ forwarding. (4) Construct a public opinion suppression maximization algorithm based on overlapping interest communities to describe the multi-information competition propagation process. The public opinion suppression maximization algorithm includes the following four stages: Phase 1: Identify complete subgraphs composed of users with high centrality and similar interests as initial communities; Phase 2: Detecting overlapping interest community structures in social networks based on label propagation; The third stage: Extracting overlapping and non-overlapping nodes based on the overlapping interest community structure and constructing a candidate seed node set according to the influence value; Fourth node: Select the final seed node set. Use a greedy strategy to continuously select the node with the largest marginal gain from the candidate seed node set and add it to the final seed node set. For the calculation of the influence range, use Monte Carlo simulation. Input the seed node set of each piece of information into the competitive propagation model that considers multiple factors and run it 1000 times to calculate the average value. The public opinion suppression maximization algorithm further includes extracting an overlapping node set and a non-overlapping node set based on an overlapping interest community structure. The overlapping nodes are simultaneously connected to multiple communities and are located in multiple communities C. v The hub position for information dissemination between communities, whose influence value depends on the strength of its relationship with each community. v→c And the average value s of the similarity of interests and preferences with each node in the community. B→c ; The non-overlapping nodes are located within the community. Considering the impact of their two-hop neighbor nodes on the information propagation process, including trust level and encounter probability, the higher the trust level, the more likely the neighbor node is to accept the forwarded information and expand the propagation range. The higher the encounter probability, the shorter the information interaction delay, so as to achieve a larger propagation range in a shorter time.

2. The method for maximizing public opinion suppression based on overlapping interest communities according to claim 1, characterized in that, This method combines the principles of trust formation between people in social psychology to analyze the trust level of user v towards neighbor u from three dimensions: sx_t uv The level of trust built on familiarity depends on the frequency of information exchange between users. The more frequent the information exchange, the more familiar the users are with each other, and the higher the level of trust built on familiarity. uv The level of trust derived from similarity depends on the similarity of interests and preferences between users. Higher similarity indicates more common ground between users, making them more receptive to information forwarded by each other, thus leading to a higher level of trust. uv The degree of trustworthiness of neighbor u for user v is higher when neighbor u interacts with some neighbors very frequently or maintains information interaction with a large number of neighbors. Even if user v does not interact with neighbor u frequently, user v tends to trust neighbor u.

3. The method for maximizing public opinion suppression based on overlapping interest communities according to claim 1, characterized in that, In the state transition mechanism, a node in state ① has not received any information. When it receives a single message, it decides whether to accept the message based on the combined trust of the neighboring nodes that forwarded the message and the topic compatibility between the node and the message. Furthermore, it decides whether to forward the message based on the topic compatibility. If a node in state ① receives multiple messages, it judges them separately in the above manner. If a node in state ② receives a message but does not forward it, and then receives the same message again, it will decide whether to forward it based on the number of neighboring nodes that forwarded the message to the node, i.e., the impact accumulates. If a node in state ② receives other messages besides the one it has already received, it needs to determine whether to discard the one it has already received and instead accept the new message. This requires calculating the combined trust strength of the neighboring nodes that forwarded the new message to the node and comparing the topic fit between the node and the one it has already received and the one it has received. If the topic fit between the node and the one it has already received is higher, then the combined trust strength is needed to make up for the difference in topic fit between the node and the two messages. Conversely, the combined effect of trust levels has a beneficial effect.

4. The method for maximizing public opinion suppression based on overlapping interest communities according to claim 1, characterized in that, The aforementioned public opinion suppression maximization algorithm includes the following process: Arrange the nodes in the social network in descending order of centrality, and visit them sequentially to build the initial community; If the centrality of the currently accessing node is greater than the average centrality, join the current initial community; The importance of neighboring nodes to the current visiting node is evaluated based on their similarity to the current visiting node's interests and their centrality. The neighboring node with the highest importance is selected to join the initial community, and the common neighboring nodes of all nodes in the current initial community are retained. Continuously search for the node with the highest importance to the current initial community from the public neighbor nodes, add it to the initial community, and delete non-public neighbor nodes until the public neighbor node set is empty.

5. The method for maximizing public opinion suppression based on overlapping interest communities according to claim 1, characterized in that, The suppression maximization algorithm sets a unique label for the initial community, and nodes within the initial community share the same label with a membership coefficient of 1. In each iteration, the label set of each node in the current iteration is updated based on the label set of the neighboring nodes in the previous iteration. This includes: summing all the labels appearing in the label sets of the currently visited node's neighboring nodes, adding the corresponding attribution coefficients, weighting the attribution coefficients according to the relationship strength and interest similarity between the currently visited node and the community corresponding to each label to obtain a weighted attribution coefficient, recording the maximum weighted attribution coefficient, and deleting labels whose weighted attribution coefficient / maximum weighted attribution coefficient is less than a given threshold; until the number of nodes belonging to each label reaches its minimum and no longer changes. Nodes with the same label belong to the same interest community. Some nodes may hold multiple labels and belong to multiple interest communities, thus forming an overlapping interest community structure.

6. The method for maximizing public opinion suppression based on overlapping interest communities according to claim 1, characterized in that, The mathematical representation of the set of overlapping nodes extracted based on the structure of overlapping interest communities is:

Images