Social network key node selection method and system based on influence attenuation and storage medium

By constructing a communication dynamic model in social networks, using second-order neighbor information and iteratively updating the influence contribution value, the problem of high time complexity and overlapping influence of the influence maximization algorithm in large networks is solved, and the problem of more Wide coverage of communication range and influence.

CN118916637APending Publication Date: 2024-11-08HARBIN INST OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411042681.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing influence maximization algorithms have high time complexity in large networks and fail to fully consider the impact overlap between nodes.

Method used

Using the method of selecting key nodes of social networks based on influence decay, the communication dynamics model is constructed, the expected influence of nodes is calculated using second-order neighbor information, and the contribution value of node influence is updated iteratively to reduce the overlap of influence between nodes. Select the initial propagation node to achieve a wider propagation range.

Benefits of technology

It realizes the identification of key nodes with low time complexity in large networks, reduces the overlap of influence between nodes, and expands the spread range and influence coverage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916637A_ABST
    Figure CN118916637A_ABST
Patent Text Reader

Abstract

The invention discloses a social network key node selection method and system based on influence attenuation and a storage medium, relates to the field of social networks, and aims to solve the problems that an existing method is high in time complexity and not suitable for a large network, and influence overlapping between nodes is not considered. Comprising the steps that S1, a propagation dynamics model is constructed, initial propagation nodes are selected, the propagation probability between the nodes is beta, and meanwhile the nodes are converted into a recovery state with the probability gamma = 1; s2, calculating the expected influence of the nodes according to the second-order neighbor information, and adding the node with the maximum influence into the initial propagation node set; s3, after the determined initial propagation node is deleted from the network, discount updating is carried out on the influence contribution of the remaining nodes, the influence of the remaining nodes of the network is calculated according to the updated node influence contribution value, and the node with the maximum influence is added into the initial propagation node set; and S4, repeatedly executing the step S3, and iterating until the size of the node set meets the requirement to obtain a key node set.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of social network, and in particular to a method, system and storage medium for selecting key nodes of a social network based on influence decay. Background Art

[0002] With the continuous development of information technology, a large number of online social platforms such as Weibo, WeChat, and Facebook have emerged on the Internet. With its efficient and convenient information transmission, online social platforms have quickly become an indispensable part of people's daily work and life. Since they are not restricted by time and space, online social networks generate and transmit information at all times, ultimately generating a large amount of social network data. At present, the issue of maximizing the influence of online social networks has attracted more and more attention in its wide application in viral marketing, public opinion control, and other aspects.

[0003] The influence maximization problem is a key research area in social networks. The main purpose is to identify a fixed set of nodes from the network so that when the node set is selected as the initial propagation source, the widest propagation range can be achieved. At present, there are three main influence maximization algorithms: greedy algorithm, heuristic algorithm and community-based influence maximization algorithm.

[0004] Among the current influence maximization algorithms, although the greedy algorithm has the highest recognition accuracy, it is extremely limited in solving the influence maximization problem of large-scale networks due to its extremely high time complexity. Community-based algorithms and heuristic algorithms using the Top-k strategy avoid the high time complexity of the greedy algorithm, but they do not consider the influence overlap between nodes when selecting nodes. The iterative degree discount algorithm only considers the first-order neighbor information of the node when calculating the node influence, resulting in limited accuracy of the calculation result, and does not fully reduce the influence overlap between nodes during the iteration process. Summary of the invention

[0005] The technical problems to be solved by the present invention are:

[0006] Existing methods have high time complexity, are not suitable for large networks, and do not take into account the overlap of influence between nodes when calculating node influence.

[0007] The present invention adopts the following technical solutions to solve the above technical problems:

[0008] The present invention provides a method for selecting key nodes in a social network based on influence decay, comprising the following steps:

[0009] S1. Construct a propagation dynamics model; select the initial propagation node, the propagation probability between nodes is β, and the node itself changes to the recovery state with probability γ=1;

[0010] S2, calculate the expected influence of the node based on the second-order neighbor information, and add the node with the greatest influence to the initial propagation node set;

[0011] S3, after deleting the determined initial propagation node from the network, the influence contribution of the remaining nodes is discounted and updated, and then the influence of the remaining nodes in the network is calculated based on the updated node influence contribution value, and the node with the greatest influence is added to the initial propagation node set;

[0012] S4. Repeat S3 until the node set size meets the requirement and obtain the key node set.

[0013] Furthermore, the expected influence of the nodes in S2 is calculated as follows:

[0014]

[0015] where d ij is the shortest path length between node i and its second-order neighbor node j, g ij represents the number of shortest paths between node i and its second-order neighbor node j, CD j is the influence contribution value of node j, the initial CD value of all nodes is set to 1, and β is the propagation probability between node i and the second-order node j.

[0016] Furthermore, the calculation method for discounting the influence contribution of the remaining nodes in S3 is:

[0017]

[0018] Where S is the current initial node set.

[0019] Furthermore, the calculation method for recalculating the expected influence of the remaining nodes in S3 according to the updated influence contribution value is:

[0020]

[0021] Furthermore, the transmission dynamics model described in S1 is based on a standard infectious disease model.

[0022] Furthermore, a redundancy index is used to evaluate the redundancy degree of influence of seed nodes, and the redundancy index is:

[0023]

[0024] Among them, R v (t c) represents the influence range of node v as a single propagation source.

[0025] Furthermore, the average shortest distance ASD of the seed node set is used to evaluate the degree of influence overlap of the seed node set, that is:

[0026]

[0027] Where K is the number of nodes in the seed set.

[0028] A system for selecting key nodes in a social network based on influence decay, the system having a program module corresponding to the steps of the method described in any of the above technical solutions, and executing the steps in the above-mentioned method for selecting key nodes in a social network based on influence decay during operation.

[0029] A computer-readable storage medium stores a computer program, wherein the computer program is configured to implement the steps of the method for selecting key nodes in a social network based on influence decay described in any one of the above technical solutions when called by a processor.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention provides a method for selecting key nodes in a social network based on influence attenuation. By collecting the second-order neighborhood information of the node, the contribution of the second-order neighboring nodes to the influence of the target node is fully considered; and the influence contribution reduction of the selected node to the surrounding nodes is further considered, and the influence overlap between the final nodes is reduced through iterative updating. The present invention has low time complexity and is suitable for large networks. It can be used for the identification of key nodes in large networks to achieve a larger propagation range and a larger range of influence coverage. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A sample network graph of 12 nodes and 15 edges in an embodiment of the present invention;

[0033] Figure 2 It is a graph showing the change of the diffusion range of the initial node set of the six algorithms in the embodiments of the present invention over time;

[0034] Figure 3 It is a graph showing the variation of the diffusion range of the initial node set of the six algorithms in the embodiments of the present invention with the infection rate;

[0035] Figure 4 is a graph showing the variation of the average shortest path length of the initial node set of the six algorithms in the embodiments of the present invention with the size of the propagation set;

[0036] Figure 5Graph showing the change in initial node set redundancy of the six algorithms in the embodiments of the present invention as the size of the propagation set changes. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the scheme of the present invention, exemplary implementations or embodiments of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described implementations or embodiments are only implementations or embodiments of a part of the present invention, not all of them. Based on the implementations or embodiments of the present invention, all other implementations or embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0039] The present invention provides a method for identifying key node sets in a social network based on group communication, comprising the following steps:

[0040] S1. Construct a propagation dynamics model G = (V, E), where V represents the node set and E represents the edge set of the network. Set the initial propagation node set to empty and the influence contribution value of all nodes to 1. Select the initial propagation node, the propagation probability between nodes is β, and the node itself is transformed into the recovery state with a probability of γ = 1.

[0041] S2, calculate the expected influence of the node based on the second-order neighbor information, and add the node with the greatest influence to the initial propagation node set;

[0042] Assuming that there is a connected path between node i and node j, and the propagation probability is β, then we can know that the probability of node i influencing j through one of the paths will decrease exponentially with the length of the path. Since the propagation process between different paths is independent of each other, the probability of node i influencing j through different paths can be directly added together to finally get the probability of node i influencing j. Furthermore, by accumulating the probability of node i influencing all other nodes in the network, we can get the expected influence of node i:

[0043]

[0044] where d ij is the shortest path length between node i and its second-order neighbor node j, d ij <3,g ij represents the number of shortest paths between node i and its second-order neighbor node j, CD j The influence contribution value of node j.

[0045] S3. After deleting the determined initial propagation node from the network, the influence contribution of the remaining nodes is discounted and updated, that is, if the neighbor j of node i is selected as the initial propagation node, then node i has a probability of being affected by β. If node k can also affect node i, the contribution of node i to the influence of node k should be discounted at a rate of (1-β). During the propagation process, i may be affected by j with a probability of β and become a new propagation node. At this time, the influence of node k on i is invalid. Therefore, the influence contribution of node i is discounted and adjusted. After deleting the initial propagation node, the network nodes are re-sorted, and the node with the greatest influence is selected to be added to the initial propagation node set. The update range considers the neighbor nodes in the second-order neighborhood of the node.

[0046] The calculation method of node j’s influence contribution discount is:

[0047]

[0048] Where S is the selected initial node set, (d jk <3).

[0049] Calculate the influence of the remaining nodes in the network based on the updated node influence contribution value:

[0050]

[0051] Add the most influential nodes to the initial propagation node set.

[0052] S4. Repeat S3 until the node set size meets the requirement and obtain the key node set.

[0053] The specific process of the algorithm in this embodiment is as follows:

[0054]

[0055]

[0056] The method (algorithm) for identifying key node sets in a social network based on group communication proposed in the present invention is the underlying technical core of the present invention, and various products can be derived based on the algorithm.

[0057] Based on the method proposed in the present invention, a social network key node set identification system based on group propagation is developed using a programming language. The system has program modules corresponding to the steps of the above-mentioned technical solution, and executes the steps in the above-mentioned social network key node set identification method based on group propagation during runtime.

[0058] The computer program of the developed system (software) is stored on a computer-readable storage medium, and the computer program is configured to implement the steps of the above-mentioned method for identifying key nodes of a social network based on group propagation when called by a processor, that is, the present invention is materialized on a carrier to become a computer program product.

[0059] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0060] The computer programs (also referred to as programs, software, software applications, or codes) of the present invention include machine instructions for programmable processors, and these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or device (e.g., disk, optical disk, memory, programmable logic device PLD) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0061] Example 1

[0062] This embodiment builds a propagation dynamics model based on the standard infectious disease model SIR (Susceptible-Infected-Recovered), and divides the states of nodes into three categories: Susceptible state S (Susceptible), in the initial state, all nodes are not infected with the virus, the node is in a susceptible state, and after contacting an infected node, it is infected with the virus with a probability of β. Infected state I (Infected), after being infected by the virus, the node is converted to an infected state, and the virus is transmitted to neighboring nodes with a probability of β. Recovery state R (Recovered), the infected node will recover with a probability of γ and enter the recovery state (immune state). The node in this state will no longer react to the virus, and will not be infected even if it contacts an infected node, which is equivalent to removing the node from the infectious network. At the beginning of the propagation, the propagation source is selected, and these nodes are set to the infected state. Then, in each time step, the infected node infects its neighboring nodes with a probability of β, and at the same time, it itself is converted to a recovery state with a probability of γ=1, until there are no infected nodes in the entire network, and the propagation process ends. The number of recovered nodes R(t) is used to represent the influence diffusion range of the initial node set. R(t)=N I (t)+N R (t).N I (t) represents the number of infected nodes in the network at time t, N R (t) represents the scale of recovered nodes in the network at time t. At the end of the propagation process, the number of recovered nodes in the network is the final influence diffusion range.

[0063] The average shortest distance ASD of the seed node set is used to evaluate the degree of overlap of the influence of the seed node set, that is:

[0064]

[0065] The redundancy index is used to evaluate the redundancy degree of influence of seed nodes. The redundancy index is:

[0066]

[0067] R v (t c ) represents the influence range when node v is the single propagation source.

[0068] like Figure 1 As shown, this embodiment constructs a simple unweighted and undirected network, selects 5 initial propagation nodes, and compares the recognition results of the method of the present invention (MSCD) with degree centrality (DC), betweenness centrality (BC), K-Shell algorithm (KS), PageRank algorithm (PR) and degree discount algorithm (DD).

[0069] Taking node 9 as an example, the first-order neighbors of this node are nodes 0, 3, and 4, and the second-order neighbors are 1, 2, 8, and 11. The number of shortest paths from node 9 to the second-order neighbors 1, 2, 8, and 11 are 1, 2, 1, and 1 respectively. The calculation shows that when the infection rate β = 0.4284 (1.2 times the transmission threshold, the transmission threshold <k>is the average number of neighbors of network nodes, <k 2 > is the average of the squares of the number of network nodes), the initial expected influence of node 9 is:

[0070] T9=1+3*β+(1+2+1+1)*β 2 =3.2028

[0071] Table 1 shows the influence levels of example network nodes obtained based on different identification methods.

[0072] Table 1

[0073]

[0074]

[0075] After obtaining the initial node influence, add the top-ranked node 4 to the initial node set, delete node 4, and then update the influence contribution discounts of the remaining nodes. For example, if the first-order neighbor of node 4 is 1 and the second-order neighbor is 0, the number of shortest paths from node 4 to these two nodes is 1 and 2 respectively, and the updated influence contribution discount is calculated as:

[0076] CD1=(1-β)=0.5716

[0077] CD0=(1-β) 2*2 =0.1068

[0078] The latest expected influence of node 9 is calculated based on the updated influence contribution discount:

[0079] IT9=1+β(CD0+CD3)+β 2 (CD2+CD8+CD 11 )

[0080] IT9=1+0.4284*(0.1068+0.5716)+0.1835*(0.5716+1+0.3267)

[0081] IT9=1.6389

[0082] Table 2 shows the top four node numbers of the sample network influence ranking based on different identification methods.

[0083] Table 2

[0084]

[0085] From the results in Tables 1 and 2, we can see that for the node with the largest degree value and located in the center of the network, the algorithm of the present invention, like other algorithms, gives it a relatively high influence value, so that the node has a high probability of being added to the initial nodes when it is selected for the first time, indicating that the method of the present invention and other methods have a high consistency when only considering the influence of a single node; however, in the final node selection process, the present invention will reduce the influence value of the nodes around the selected node, thereby tending to select distant nodes, so that the final seed node set has a wider distribution range.

[0086] Example 2

[0087] To further verify the performance of the algorithm of the present invention, this embodiment selects 6 real networks for propagation simulation, namely:

[0088] (1) CA-HepTh, a collaborative network for high-energy theoretical physics papers on the preprint website arxiv;

[0089] (2)facebook_combined, a Facebook "circle" network;

[0090] (3) athletes_edges, a network of sports-related pages on Facebook;

[0091] (4) government_edges, a network of government-type pages on Facebook;

[0092] (5) company_edges, a network of company-type pages on Facebook;

[0093] (6) lastfm_asia, a social network for Asian users of the software LastFM.

[0094] The data are all from the Stanford Public Network Dataset.

[0095] The basic structural properties of the sample network are shown in Table 3, where N is the sample size, M is the number of edges, <k>is the average degree of the network, L is the average path length, LCC is the maximum connected component of the network, C is the average clustering coefficient, is the propagation threshold.

[0096] Table 3

[0097]

[0098] Five algorithms, namely degree centrality (DC), betweenness centrality (BC), K-Shell algorithm (KS), PageRank algorithm (PR) and degree discount algorithm (DD), were selected as benchmark algorithms for comparative analysis. In the same sample, if different nodes obtain the same score according to the identification method, the nodes with the same score will be assigned the same ranking, and the nodes in the same ranking will be randomly selected as the propagation source. To ensure the stability of the experimental results, the propagation results of each group of samples are taken as the average of 1000 repeated diffusion experiments.

[0099] like Figure 2 As shown in the figure, the final diffusion range of the proposed method (MSCD) is higher than that of the other five benchmark algorithms, and in most cases the propagation speed of the proposed method is close to that of the other five algorithms. Figure 2 In the results shown in (e), the propagation speed of the (MSCD) algorithm is slightly lower than that of other algorithms. This is because when selecting the initial seed set, the MSCD algorithm tends to disperse the seed nodes as much as possible in the network, resulting in the initial propagation speed lagging behind other algorithms. However, through the collective influence of the concentrated seed nodes, the final propagation range of the MSCD algorithm is still higher than that of other algorithms. This shows that the MSCD algorithm can achieve a larger influence diffusion range and has a faster diffusion speed.

[0100] In order to explore the stability of the MSCD algorithm under different conditions, further tests were conducted when the infection rate was β th ~2β th The final diffusion range of the MSCD algorithm when the range fluctuates within 100%, and is compared with the above five benchmark algorithms. Figure 3 As shown in FIG. 1 , the final diffusion range curve of the MSCD algorithm of the present invention is above the curves of the other five benchmark algorithms in most scenarios, indicating that the MSCD algorithm performs better than the other five benchmark algorithms in most cases and has very stable performance. Figure 3 In the results shown in (d), the degree discount algorithm has a final diffusion range close to that of the MSCD algorithm when the infection rate is low, and is better than the other four algorithms. However, as the infection rate continues to rise, the growth momentum of the degree discount algorithm gradually begins to lag behind, and is slightly weaker than the MSCD algorithm in terms of influence stability.

[0101] like Figure 4 As shown in the figure, when the initial propagation set is small, the average path length of the seed set of the MSCD algorithm does not have a large gap with the other five benchmark algorithms. However, as the seed set continues to grow, the average shortest path length of the node set of the MSCD algorithm increases significantly faster than the other five benchmark algorithms. Figure 4 In (c) and (d), the average shortest path length of the seed set of the K-shell algorithm and the PageRank algorithm is higher than that of the MSCD algorithm, but the performance of the MSCD algorithm is still better than that of the other four algorithms. In general, the MSCD algorithm can stably and evenly disperse the seed set nodes in the network to avoid overlapping influence within the seed set.

[0102] like Figure 5 As shown in the figure, as the seed node set continues to grow, the degree of overlap of node influence within the set continues to deepen. Simply increasing the size of the initial propagation set will result in diminishing marginal utility, and the increase in the size of the initial propagation set will contribute less and less to the expansion of the influence range. Except for the facebook_combined network and the lastfm_asia network, the redundancy of the MSCD algorithm is better than the benchmark algorithm. Even in the lastfm_asia network, the performance of the MSCD algorithm is very close to the best performing betweenness centrality algorithm, and eventually exceeds the betweenness centrality algorithm as the set size increases. This shows that the MSCD algorithm is more stable than the benchmark algorithm in avoiding overlap of influence within the seed set.

[0103] Finally, the time complexity of the MSCD algorithm is analyzed. The main process of the MSCD algorithm can be divided into three steps: pre-sorting process, influence contribution discount update, and node influence update. The pre-sorting process is similar to the node influence update process. Both of them need to traverse the second-order neighbors of each node, so the time complexity of these two steps is O(n <k> 2 ),in <k>is the average degree of network nodes, and n is the total number of network nodes. When the seed set size is c, the influence contribution discount needs to update the second-order neighbor node information of c nodes, and the time complexity is about O(c <k> 2 ). Therefore, the time complexity of the MSCD algorithm is O(n <k> 2 ), which shows that the MSCD algorithm has a lower time complexity.

[0104] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.< / k> < / k> < / k> < / k> < / k> < / k>

Claims

1. A method for selecting key nodes in a social network based on influence decay, characterized in that: The steps include: S1. Construct a propagation dynamics model; select the initial propagation node, the propagation probability between nodes is β, and the node itself changes to the recovery state with probability γ=1; S2, calculate the expected influence of the node based on the second-order neighbor information, and add the node with the greatest influence to the initial propagation node set; S3, after deleting the determined initial propagation node from the network, the influence contribution of the remaining nodes is discounted and updated, and then the influence of the remaining nodes in the network is calculated based on the updated node influence contribution value, and the node with the greatest influence is added to the initial propagation node set; S4. Repeat S3 until the node set size meets the requirement and obtain the key node set.

2. The method for selecting key nodes in a social network based on influence decay according to claim 1 is characterized in that: The expected influence of a node in S2 is calculated as: where d ij is the shortest path length between node i and its second-order neighbor node j, g ij represents the number of shortest paths between node i and its second-order neighbor node j, CD j is the influence contribution value of node j, the initial CD value of all nodes is set to 1, and β is the propagation probability between node i and the second-order node j.

3. The method for selecting key nodes in a social network based on influence decay according to claim 2 is characterized in that: The calculation method for discounting the influence contribution of the remaining nodes in S3 is: Where S is the current initial node set.

4. The method for selecting key nodes in a social network based on influence decay according to claim 3 is characterized in that: The calculation method for recalculating the expected influence of the remaining nodes in S3 based on the updated influence contribution value is:

5. The method for selecting key nodes in a social network based on influence decay according to claim 1 is characterized in that: The transmission dynamics model described in S1 is based on standard infectious disease models.

6. The method for selecting key nodes in a social network based on influence decay according to claim 1 is characterized in that: The redundancy index is used to evaluate the redundancy degree of influence of seed nodes. The redundancy index is: Among them, R v (t c ) represents the influence range of node v as a single propagation source.

7. The method for selecting key nodes in a social network based on influence decay according to claim 1 is characterized in that: The average shortest distance ASD of the seed node set is used to evaluate the degree of overlap of the influence of the seed node set, that is: Where K is the number of nodes in the seed set.

8. A social network key node selection system based on influence decay, characterized in that: The system has a program module corresponding to the steps of the method described in any one of claims 1 to 7, and executes the steps of the above-mentioned method for selecting key nodes in a social network based on influence decay when running.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps in the method for selecting key nodes in a social network based on influence decay according to any one of claims 1 to 7 when called by a processor.

Citation Information

Cited By

  • Message control method, product, electronic equipment and medium

    CN119211182A