Research method and system for influence node identification in random high-order network
By introducing extended influence node identification indicator EII and weighted extended influence node identification indicator WEII in higher-order networks, the problem that traditional indicators cannot capture the advanced interaction mode is solved, and more accurate influence on nodes in higher-order networks is achieved Evaluation and robustness analysis.
Patent Information
- Application Number
- CN202411227642.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2024-12-03
AI Technical Summary
It is difficult for the prior art to fully identify and evaluate the influence of nodes in advanced networks, and traditional node-centric indicators cannot effectively capture the complex interaction patterns and structural characteristics of nodes in advanced networks.
It proposes the extended influence node identification index EII and the weighted extended influence node identification index WEII. By considering the participation of nodes in simplex in different dimensions and the number of neighbor nodes, combined with the sub-face structure information of simplex, it provides more accurate influence. Force assessment.
It can more comprehensively evaluate the influence of nodes in higher-order networks, improve the accuracy and robustness of identification of important nodes in higher-order networks, and is suitable for the mining of key nodes and network optimization of complex systems.
Smart Images

Figure CN119071135A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of evaluation of influential node identification in random high-order networks, and specifically to a research method and a research system for influential node identification in random high-order networks. Background Art
[0002] In modern scientific research, complex networks, as a method to characterize the interactions between various systems, have made significant progress in the study of their topological properties and dynamic behaviors. The heterogeneity of the network causes some nodes to play an important role in the network. At the same time, these nodes will have a huge impact on the structure and function of the network. For example, the opinions and behaviors of "opinion leaders" in social networks will affect the behavior of other users in the network. These opinion leaders can guide the behavior of other users, influence the process of information dissemination and opinion formation, and thus form a significant influence in the network. Therefore, how to accurately evaluate the importance of each node in the network and thus mine the high-influence nodes in the network is of great significance to understanding and optimizing complex systems.
[0003] In recent years, significant progress has been made in mining important nodes in common complex networks. Among them, classic important node identification methods including centrality, betweenness centrality, eigenvector centrality, PageRank, etc. have been widely used in common networks. For example, some researchers proposed a network node importance identification algorithm based on adjacency information entropy by studying the association between nodes and their direct and indirect neighboring nodes. This algorithm effectively characterizes the importance of nodes in the network by calculating the adjacency and information entropy of nodes. In addition, in real systems, in addition to measuring the importance of a single node, it is often necessary to simultaneously identify multiple important nodes to cope with the challenges in complex networks. To this end, some researchers proposed a new algorithm MINE (maximum influential neighbor extension) to effectively cope with the challenges of network analysis in a big data environment by hierarchically identifying key nodes in complex networks. However, with the deepening of the understanding of network structure and function, researchers have found that traditional network models can no longer fully capture the higher-order interaction relationships beyond binary in complex systems. Therefore, the emergence of high-order networks has broken through the limitations of traditional network models for low-order interactions and has shown unique advantages in revealing the deep organizational structure and functions within the system. Due to the essential difference between ordinary networks and high-order networks, traditional important node identification methods are no longer applicable. To this end, researchers have carried out research based on the identification of important nodes in high-order networks to address the challenges faced in mining key nodes in high-order networks. In order to mine high-influence nodes in high-order networks based on simplex structures, the present invention takes into account the differences in simplices of different dimensions in high-order networks and the sub-face structure characteristics of different simplex structures. On the basis of the basic degree centrality index DC, the degree centrality index is given reasonable weights and related parameters through high-order structural characteristics, and two influential node identification indicators in high-order networks of the present invention are proposed.
[0004] Existing research results show that networks with different structures also have differences in identifying high-influence nodes or important nodes. At present, the technology for mining and identifying important nodes in ordinary networks has become increasingly mature, but in high-order networks beyond binary interactions, there is relatively little research on the identification of important nodes. Traditional node centrality indicators such as degree centrality and betweenness centrality often perform poorly in these networks because they cannot effectively capture the complex interaction patterns and structural characteristics of nodes in high-order networks. Therefore, researchers have begun to explore new measurement methods and algorithms to adapt to the special properties of high-order networks. For high-order networks such as hypernetworks based on hypergraphs, some researchers have extended the K-shell method to hypernetworks from the perspective of network structure, and proposed a new indicator using Euclidean distance combined with hyperdegree and K-shell value. To identify the key nodes in the hypernetwork; then, the researcher evaluated the importance of the nodes by considering their direct and indirect relationships, providing a comprehensive perspective on the influence of nodes in the hypernetwork; some researchers have also proposed a hypergraph important node identification method based on von Neumann entropy, which identifies key nodes by integrating high-order information and shows the complex relationship of nodes in the hypernetwork. In short, the hypernetwork important node identification method reveals to us more detailed associations and action patterns in complex systems, which has important theoretical significance and practical application value.
[0005] On the other hand, the mining of important nodes in high-order networks is based on high-order networks with simplex structures, but the identification of important nodes in such high-order networks is still in its infancy. For example, some researchers have improved the ability to identify key nodes in the network by combining the multi-order graph representation method of high-order binary graphs and classical pair graphs, through the high-order enhanced random walk model HoRW. In addition, some researchers have used simple complex structures to quantitatively characterize high-order propagation dynamics, and defined propagation centrality to identify key nodes, providing a new method to quantify the influence of node propagation in high-order networks. Although the existing methods for identifying important nodes in high-order networks with simplex structures have been innovated from the perspective of propagation and structure, they are still insufficient in fully capturing the multi-dimensional interactions of nodes in high-order networks and evaluating their influence on the overall network. Therefore, we need to develop new methods and indicators to more accurately identify and evaluate important nodes in high-order networks with simplex structures. The present invention considers a new method suitable for identifying important nodes in high-order networks from the perspective of high-order topological structure of high-order networks. The extended influence index EII evaluates the importance of a node by considering the node's participation in the d-dimensional simplex and the number of neighboring nodes it influences in the simplex; while the weighted extended influence index WEII enables it to more accurately measure the specific contribution of the simplex structure of each dimension to the influential node in the high-order network. This index adds weight distribution to different simplex dimensions on the basis of the index EII, and comprehensively considers the complexity of the node in the simplex of different dimensions through information such as the number of sub-faces of the d-dimensional simplex and its sub-face dimensions, so as to more meticulously capture the key role of the node in the high-order network structure. In the end, the two indicators proposed by the present invention can more comprehensively evaluate the influence of nodes in high-order networks compared with the basic indicator DC. At the same time, this experiment not only verifies the effectiveness of these two indicators, but also highlights their superiority in the problem of high-order network node identification.
[0006] In summary, the current work on mining important nodes is mainly focused on complex networks with pairwise interactions and high-order networks such as hypergraph-based hypernetworks. The research method is mainly to analyze the structural characteristics of the network and the network dynamics characteristics, but the identification of important nodes in high-order networks with simplex structures is still in the initial exploration stage. Summary of the Invention
[0007] In view of this, the purpose of the present invention is to provide a research method and system for identifying influential nodes in a random high-order network, so as to solve the technical problem in the prior art that it is impossible to comprehensively understand the influence of nodes in a complex high-order network that conforms to reality. The method of the present invention combines the idea of the traditional node centrality index DC with the emerging high-order network analysis technology, and particularly focuses on capturing more complex high-order interaction patterns and high-order structure information between nodes.
[0008] To achieve the above object, the present invention provides the following technical solutions:
[0009] A research method for identifying influential nodes in a random high-order network, the method comprising the following steps:
[0010] Construct a random high-order network;
[0011] Propose an extended influential node identification index EII and a weighted extended influential node identification index WEII based on the node degree centrality index DC;
[0012] Evaluate and verify the two proposed influential node identification indexes and the basic index degree centrality index DC through network robustness, and prove the effectiveness and accuracy of the proposed indexes.
[0013] As a further technical solution of the present invention, the process of constructing the random high-order network is mainly: according to different simplex structures, different connection probabilities are assigned to different-dimensional simplices to construct a random high-order network including 1-simplex, 2-simplex, and 3-simplex structures. The steps of constructing the random high-order network include:
[0014] Step 1: Initial setting of the network: The network is initialized to include N independent nodes, where N is the preset network scale; in addition, the parameter M is set as the total number of simplices in the entire high-order network;
[0015] Step 2: Generation of 1-simplex structure: On the basis of Step 1, by traversing all node pairs in the network The nodes are connected to each other with probability p1 to form 1-simplices, that is, ordinary edge connections. For each pair of independent nodes in the network, a random number r between 0 and 1 is generated. If r < p1, a connection is established between these two nodes. Therefore, Step 1 forms the basic structure of the random high-order network;
[0016] Step 3: Generation of 2-simplex structure: On the basis of Step 2, the network adds 2-simplex structures with probability p2, that is, "full" triangles, by traversing all possible node triples in the high-order network Generate a random number \(r\) for each triple again. If \(r < p2\), construct a completely connected closed triangle among these three nodes;
[0017] Step 4. Generation of 3 - simplex structure: Based on Step 3, the network adds 3 - simplex structures, that is, tetrahedrons, with probability \(p3\). By traversing all possible node quadruples and generating a random number \(r\) for each quadruple. If \(r < p3\), construct a completely connected tetrahedron among these four nodes;
[0018] After gradually constructing 1 - simplex, 2 - simplex, and 3 - simplex, a random high - order network with dimension \(d = 3\) is finally constructed. This \(d\) - dimensional random high - order network not only includes basic point - to - point connections (1 - simplex), but also contains "full" triangles (2 - simplex) and tetrahedrons (3 - simplex).
[0019] As a further technical solution of the present invention, considering the rich and multi - dimensional characteristics of simplex structures in the random high - order network, the present invention considers simplex structures of different dimensions included in the random high - order network, that is, 1 - simplex, 2 - simplex, and higher - dimensional 3 - simplex structures, and proposes two node influence indicators to identify high - influence nodes in the random high - order network. The proposed indicators comprehensively and fully give indicators for measuring node influence from information such as the generalized degree of nodes in the high - order network, the simplex structures of different dimensions where the nodes are located, the number of neighbor nodes around the nodes, and the structure of simplex sub - faces in simplex structures of different dimensions where the nodes are located, namely the extended influence node identification index EII and the weighted extended influence node identification index WEII.
[0020] The present invention first proposes a new index considering high - order network structure parameters, that is, the extended influence node identification index EII, for evaluating the role and importance of nodes in the high - order network structure. The EII index determines the importance of a node by measuring the influence of the node on neighbor nodes in each simplex it is associated with, so as to discover high - influence nodes in the high - order network under high - order interactions. Specifically, this index fully considers the relative position of the node in simplexes of different dimensions and the number of neighbor nodes it affects. The extended influence node identification index EII proposed by the present invention is defined as the following formula:
[0021]
[0022] where \(d\) represents the dimension of the simplex in the high - order network, \(d = 0,1,\cdots,n\); represents node \(v\) i the ratio of the number of neighbor nodes other than node \(v\) i in the currently associated \(d\) - simplex to the number of nodes in the \(d\) - simplex; \(k\)d,0 (v i ) represents node v i The number of associated d-simplices. For the simplex structure in the network, the higher the dimension, the more high-order information the simplex structure contains, and it will affect more neighboring nodes, so the node is more important.
[0023] As a further technical solution of the present invention, the extended influential node identification index provides a powerful measure for identifying high-influence nodes in high-order networks. It evaluates the overall role of nodes by comprehensively considering the influence of nodes in simplexes of different dimensions and the influence of nodes on neighboring nodes of different dimensions. However, the EII index still depends on the number of simplexes of different dimensions, and fails to fully reflect the differential contribution of simplexes of different dimensions to the influence of the network. For example, high-dimensional simplexes have higher-order multivariate interactions, so high-dimensional simplexes can often carry more important information in high-order networks.
[0024] In order to solve this problem and further refine the evaluation of node influence, this section proposes the weighted extended influence node identification index WEII. This indicator adds weighted processing of simplexes of different dimensions on the basis of the EII indicator to more accurately distinguish the differential contributions of simplexes of different dimensions to the network. The WEII indicator not only takes into account the participation of nodes in simplexes of different dimensions, but also further distinguishes the contribution of simplexes of different dimensions to the influence of the network by introducing the weights of simplexes of different dimensions. Simplexes of higher dimensions usually involve more nodes, thus carrying more complex information and functions in the network. Therefore, this paper defines the weighted extended influence node identification index WEII as follows:
[0025]
[0026] Where d represents the dimension of the simplex in the high-order network, d = 0, 1, ..., n; Represents node v i The current associated d-simplex except node v i The ratio of the number of neighboring nodes outside to the number of d-simplex nodes; k d,0 (v i ) represents node v i The number of associated d-simplices. d represents the weight of a d-simplex of dimension d, W d It is defined as follows:
[0027]
[0028] Among them, C d represents the structural complexity of the d-simplex; and F represents the parameter of the k-dimensional subsurface of the d-simplex; d,k represents the number of k-dimensional subsurfaces of the d-simplex. In addition, k represents the dimension of the k-dimensional subsurfaces under the d-simplex in the high-order network.
[0029] The WEII index introduces a weight factor based on the EII index to more accurately identify key nodes in high-order networks. This weight setting not only takes into account the dimension of the simplex associated with the node, but also the structural information of the sub-facets inside the simplex. Specifically, the higher the dimension of the simplex, the richer the sub-facet structural information it contains, and the corresponding weight is larger. For example, consider two nodes v1 and v2, both of which are associated with 3 simplex structures. According to the traditional degree centrality index (DC), if only the number of simplices associated around the node is considered, and the differences between different simplices are ignored, the DC values of points v1 and v2 are both 3, which seem to be equally important. However, if v1 is associated with 3 1-simplices, and v2 is associated with 3 higher-dimensional 2-simplices or 3-simplices, then the importance of these two nodes is obviously different. Through the weighted method of the WEII index, we can distinguish the difference in importance between v1 and v2 and more accurately evaluate the different influences of each node in the network.
[0030] As a further technical solution of the present invention, the steps of evaluating and verifying the proposed two influential node identification indicators and the basic indicator degree centrality indicator DC through network robustness to prove the effectiveness and accuracy of the proposed indicators include:
[0031] According to the constructed random high-order network and the two influential node identification indicators proposed, an evaluation indicator suitable for evaluating the robustness of high-order networks is proposed, namely the high-order simplex retention rate HOR. HOR is the ratio of the weighted number of various simplices remaining in the network after the network is attacked to the weighted number of all simplices before the attack. It is used to measure the structural integrity and functional retention of the network after being damaged. The calculation formula of HOR is as follows:
[0032]
[0033] Among them, M'1, M'2 and M'3 represent the number of 1-simplex, 2-simplex and 3-simplex remaining in the network after the attack, respectively, and M1, M2 and M3 represent the number of 1-simplex, 2-simplex and 3-simplex in the network before the attack, respectively; W1, W2 and W3 are the weight values given to 1-simplex, 2-simplex and 3-simplex when evaluating the importance of network structure, respectively, reflecting the contribution of different simplices to network function; here, the weights of simplices of different dimensions are set as 1-simplex weight W1 = 1, 2-simplex weight W2 = 2, 3-simplex weight W3 = 6. HOR quantifies the current robustness of the network after attack by evaluating the HOR obtained by EII index and WEII index under different attack intensities.
[0034] Later, through the posterior robustness measure S q , based on the changes of simplexes of different dimensions. Specifically, for the posterior measure HOR considering high-order network damage, the robustness measure S q It can be written as follows:
[0035]
[0036] The posterior robustness measure S q It is actually the weighted average result of the HOR value, which is used to analyze how the overall performance or state of the network changes over time or other factors. The posterior robustness metric S q It provides a global perspective to observe the robustness of the network, which helps to understand the average performance of the network in the long term or under changing conditions.
[0037] Preferably, the evaluation index of high-order network robustness is applied to the high-order network, and the high-order network robustness is analyzed by combining two improved indexes EII and WEII and basic index DC, so as to obtain the effectiveness of the proposed index, specifically:
[0038] According to the experimental results, the threshold is set. When the attack or failure causes the robustness value S in the network q When it drops to the corresponding threshold, the number of high-dimensional structures in the high-order network decreases with the destruction of the network, which means that the network is no longer connected; at the same time, the high-order network will become sparse from the initial state as the nodes are removed: the lower the threshold, the more the network can maintain its function with less loss of high-order structures, indicating that the high-order network is more robust. By evaluating the robustness value S calculated by the EII and WEII indicators under deliberate attacks q , the robustness of the network can be quantified, thus proving the effectiveness of the proposed indicator.
[0039] Another object of the present invention is to provide a research system for identifying influential nodes in a random high-order network, the system comprising:
[0040] The first building block builds a random high-order network;
[0041] The second building block proposes the extended influence node identification index EII and the weighted extended influence node identification index WEII based on the node degree centrality index DC;
[0042] In the third application module, the two proposed influential node identification indicators and the basic indicator degree centrality indicator DC are evaluated and verified through network robustness to prove the effectiveness and accuracy of the proposed indicators.
[0043] Compared with the prior art, the present invention has the following beneficial effects:
[0044] In a random high-order network, the importance of a node can be measured by the high-dimensional simplex structure and its number associated with it. Specifically, if a node is associated with more high-dimensional simplices (such as 2-simplices and 3-simplices), it is considered that this node has greater influence in the network. This is because high-dimensional simplices tend to contain more nodes and edges, so they play a more significant role as bridges in the network structure. When performing network robustness analysis, if these nodes associated with high-dimensional simplices are removed first, the structural integrity of the network will change rapidly, resulting in a rapid deterioration of network function or even global collapse.
[0045] In addition, the connection probability of various simplexes in the network will also significantly affect the stability of the network. For example, when the connection probabilities p2 and p3 of 2-simplexes and 3-simplexes are high, there are more high-dimensional structures in the high-order network, making the network tightly connected, which may also increase the sensitivity of the network to node failure. Therefore, once key nodes begin to be removed, such a high-dimensionally connected network may experience a collapse process faster than a network with lower connection probabilities or a high-order network containing more 1-simplex structures.
[0046] The present invention constructs a random high-order network and the proposed influential node identification index, applies it to the random high-order network, and evaluates the effectiveness and accuracy of the proposed index through a network robustness index. The present invention proposes two identification indexes for identifying influential nodes in a high-order network, and applies them to a random high-order network. According to the effectiveness of the evaluation index of the proposed high-order network robustness index, effective experimental results are obtained through experimental analysis. For the effectiveness and applicability of the proposed node influence identification index, the high-order network should include three types of simplex structures, namely 1-simplex, 2-simplex and 3-simplex, as much as possible. In addition, the index will be more effective when the network is dense. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1It is a flowchart of a research method for identifying influential nodes in a random high-order network according to an exemplary embodiment.
[0048] Figure 2 It is an algorithm flow chart of a random high-order network important node identification index EII according to an exemplary embodiment.
[0049] Figure 3 It is an algorithm flow chart of a random high-order network important node identification index WEII according to an exemplary embodiment.
[0050] Figure 4 According to an exemplary embodiment, different indicators in a random high-order network under deliberate attack, S under different q q Plot of its relationship with q.
[0051] Figure 5 is a graph showing, according to an exemplary embodiment, different indicators in a random high-order network under deliberate attack, S q Data graph of its relationship with q.
[0052] Figure 6 is a relationship diagram of the number of 1-simplices, 2-simplices and 3-simplices removed under a deliberate attack for different indicators in a random high-order network according to an exemplary embodiment.
[0053] Figure 7 3 is a graph showing the influence of the total number of simplexes M on the network robustness under different indicators in a random high-order network under a deliberate attack when the ratio of removed nodes is q=0.5 according to an exemplary embodiment.
[0054] Figure 8 is a graph showing the WEII indicator under deliberate attack in different cases of 1-simplex, 2-simplex and 3-simplex probabilities according to an exemplary embodiment. q Plot of its relationship with q.
[0055] Fig. 9 is a relationship diagram between Sq and q of the EII indicator under a deliberate attack in different cases of 1-simplex, 2-simplex and 3-simplex probabilities according to an exemplary embodiment.
[0056] Fig.10 It is a system diagram for studying the identification of influential nodes in random high-order networks according to an exemplary embodiment. DETAILED DESCRIPTION
[0057] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.
[0058] Embodiment 1
[0059] Figure 1 is a flowchart of a research method for identifying influential nodes in a random high-order network according to an exemplary embodiment. Figure 1 As shown, the method includes:
[0060] Step S01, constructing a random high-order network;
[0061] Step S02, proposing two extended influence node identification indicators EII and weighted extended influence node identification indicators WEII based on the degree centrality indicator DC;
[0062] Step S03: The two proposed indicators and the basic indicator degree centrality indicator DC are evaluated and verified through network robustness to prove the effectiveness and accuracy of the proposed indicators.
[0063] It should be noted that the technical solution provided in this embodiment is applicable to the identification and analysis of influential nodes and the evaluation of robustness applied to scientific research cooperation, movie actors, biology, transportation, protein and other systems.
[0064] It can be understood that the technical solution provided in this embodiment, by constructing a random high-order network and two indicators suitable for identifying influential nodes in high-order networks, and evaluating and verifying the two proposed indicators and the basic indicator degree centrality indicator DC through network robustness, proves the effectiveness and accuracy of the proposed indicators. The present invention is based on the constructed random high-order network based on simplex structures of different dimensions, and according to the dimensional differences and different contribution degrees of different simplex structures in the network, proposes two extended influence node identification indicators EII and weighted extended influence node identification indicators WEII based on the degree centrality indicator DC suitable for mining high-order networks based on simplex structures, and evaluates the effectiveness of the indicators based on the proposed high-order network robustness indicator, and obtains effective experimental results through experimental analysis. For the effectiveness and applicability of the proposed node influence identification indicator, the high-order network should try to include three types of simplex structures, namely 1-simplex, 2-simplex, and 3-simplex, and the two proposed indicators will be more effective when the network is dense.
[0065] It should be noted that the “EII index for important node identification in random high-order networks” can be found in the specific practice. Figure 2 , Figure 2 The flowchart of the algorithm of the random high-order network important node identification index EII according to an exemplary embodiment includes:
[0066] Taking into account the richness and multidimensional characteristics of the simplex structure in the random high-order network, either through paired interactions or through group interactions, that is, through high-order interactions. In order to study the identification method of influential nodes in random high-order networks, the present invention takes into account the simplex structures of different dimensions contained in the constructed random high-order network, namely 1-simplex, 2-simplex and higher-dimensional 3-simplex structures. Two node influence indicators are proposed to identify highly influential nodes in random high-order networks. The proposed indicators are mainly considered from three aspects, including the generalization of nodes in high-order networks, the number of neighboring nodes affected by the nodes in the simplex structures of different dimensions, and the structural information of the simplex sub-faces in the simplex structures of different dimensions where the nodes are located. Based on these three aspects, the extended influence node identification index EII and the weighted extended influence node identification index WEII are proposed.
[0067] To this end, the present invention first proposes a new indicator, namely the extended influence node identification index EII, which is used to evaluate the role and importance of nodes in high-order network structures. The EII indicator determines the influence of a node by measuring the influence of the node on its neighboring nodes in each simplex to which it is associated, thereby mining high-influence nodes in the network. Specifically, this indicator takes into account the relative position of the node in simplexes of different dimensions and the number of affected neighboring nodes. Therefore, the present invention defines the extended influence node identification index as follows:
[0068]
[0069] Where d represents the dimension of the simplex in the high-order network, d = 0, 1, ..., n; Represents node v i The current associated d-simplex except node v i The ratio of the number of neighboring nodes outside to the number of d-simplex nodes; k d,0 (v i ) represents node v i The number of associated d-simplices. For the simplex structure in the network, the higher the dimension, the more high-order information the simplex structure contains, and it will affect more neighboring nodes, so the node is more important.
[0070] The algorithm flow of the high-order network important node identification index EII is as follows:
[0071] S11, initializing the network structure and the initial state of the nodes;
[0072] S12. According to step S11, for each node in the network, evaluate its influence in each dimensional simplex. Calculate the number of neighboring nodes influenced by each node in different dimensional simplexes. Use the index EII to calculate the extended influence of each node. The index EII is calculated by integrating the position and influence of the node in each dimensional simplex.
[0073] S13. According to step S12, all nodes in the network are sorted in descending order using the results calculated by the indicator EII to determine the influence of each node.
[0074] S14. According to step S13, the indicator EII is applied to the high-order network robustness analysis, and the robustness of the network under the indicator and the effectiveness of the indicator EII are evaluated according to the proposed robustness evaluation indicator.
[0075] It should be noted that the “WEII index for important node identification in random high-order networks” can be found in the specific practice. Figure 3 , Figure 3 The flowchart of the algorithm of the random high-order network important node identification index WEII according to an exemplary embodiment includes:
[0076] The Extended Influential Node Identification (EII) provides a powerful metric for identifying high-influence nodes in high-order networks. It evaluates the overall role of nodes by integrating the influence of nodes in simplexes of different dimensions and the influence of nodes on neighboring nodes of different dimensions. However, the EII indicator still depends on the number of simplexes of different dimensions, and fails to fully reflect the differential contributions of simplexes of different dimensions to network influence. For example, high-dimensional simplexes have higher-order multivariate interactions, so high-dimensional simplexes can often carry more important information in high-order networks.
[0077] In order to solve this problem and further refine the evaluation of node influence, this section proposes the weighted extended influence node identification index WEII. This index adds weighted processing of simplexes of different dimensions on the basis of the EII index, and more accurately distinguishes the differential contributions of simplexes of different dimensions to the network. The WEII index not only takes into account the participation of nodes in simplexes of each dimension, but also further distinguishes the contribution of simplexes of different dimensions to the influence of the network by introducing the weights of simplexes of different dimensions. Simplexes of higher dimensions usually involve more nodes, thereby carrying more complex information and functions in the network. Therefore, the present invention defines the weighted extended influence node identification index as follows:
[0078]
[0079] Where d represents the dimension of the simplex in the high-order network, d = 0, 1, ..., n; Represents node v i The current associated d-simplex except node v i The ratio of the number of neighboring nodes outside to the number of d-simplex nodes; k d,0 (v i ) represents node v i The number of associated d-simplices.
[0080] W d represents the weight of a d-simplex of dimension d, W d It is defined as follows:
[0081]
[0082] Among them, C d represents the structural complexity of the d-simplex; and F represents the parameter of the k-dimensional subsurface of the d-simplex; d,k represents the number of k-dimensional subsurfaces of the d-simplex. In addition, k represents the dimension of the k-dimensional subsurfaces under the d-simplex in the high-order network.
[0083] The WEII index introduces a weight factor based on the EII index to more accurately identify key nodes in high-order networks. This weight setting not only takes into account the dimension of the simplex associated with the node, but also the structural information of the sub-facets inside the simplex. Specifically, the higher the dimension of the simplex, the richer the sub-facet structural information it contains, and the corresponding weight is larger. For example, consider two nodes v1 and v2, both of which are associated with 3 simplex structures. According to the traditional degree centrality index DC, if only the number of simplices associated around the node is considered, and the differences between different simplices are ignored, then the DC values of points v1 and v2 are both 3, which seem to be equally important. However, if v1 is associated with 3 1-simplices, and v2 is associated with 3 higher-dimensional 2-simplices or 3-simplices, then the importance of these two nodes is obviously different. Through the weighted method of the WEII index, we can distinguish the difference in importance between v1 and v2 and more accurately evaluate the network influence of each node.
[0084] The algorithm flow of the high-order network important node identification index WEII is as follows:
[0085] S11, initializing the network structure and the initial state of the nodes;
[0086] S12. According to step S11, for each node in the network, the weighted influence in each dimensional simplex is evaluated. The influence of each node in the simplex of different dimensions is weighted, and the weight depends on the dimension and structural complexity of the simplex. The weighted extended influence of each node is calculated according to the indicator WEII, which takes into account that the role of the node in the high-dimensional simplex is more significant.
[0087] S13. According to step S12, all nodes in the network are sorted by the result calculated by the indicator WEII to determine the influence of each node.
[0088] S14. According to step S13, the indicator WEII is applied to the robustness analysis of the high-order network, and the robustness of the network under the indicator and the effectiveness of the indicator WEII are evaluated according to the proposed robustness evaluation indicator.
[0089] According to a second aspect of an embodiment of the present invention, an evaluation index suitable for evaluating the robustness of a high-order network is proposed, namely, the high-order simplex retention rate HOR. HOR is the ratio of the weighted number of each type of simplex remaining in the network after the network is attacked to the weighted number of all simplexes before the attack. It is used to measure the structural integrity and functional retention of the network after being damaged. The HOR calculation formula is as follows:
[0090]
[0091] Among them, M'1, M'2 and M'3 represent the number of 1-simplex, 2-simplex and 3-simplex remaining in the network after the attack, respectively, and M1, M2 and M3 represent the number of 1-simplex, 2-simplex and 3-simplex in the network before the attack, respectively; W1, W2 and W3 are the weight values given to 1-simplex, 2-simplex and 3-simplex when evaluating the importance of network structure, respectively, reflecting the contribution of different simplices to network function; here, the weights of simplices of different dimensions are set as 1-simplex weight W1 = 1, 2-simplex weight W2 = 2, 3-simplex weight W3 = 6. HOR quantifies the current robustness of the network after attack by evaluating the HOR obtained by EII index and WEII index under different attack intensities.
[0092] Later, through the posterior robustness measure S q , based on the changes of simplexes of different dimensions. Specifically, for the posterior measure HOR considering high-order network damage, the robustness measure S q It can be written as follows:
[0093]
[0094] The posterior robustness measure S qIt is actually the weighted average result of the HOR value, which is used to analyze how the overall performance or state of the network changes over time or other factors. The posterior robustness metric S q It provides a global perspective to observe the robustness of the network, which helps to understand the average performance of the network in the long term or under changing conditions.
[0095] Preferably, the evaluation index of high-order network robustness is applied to the high-order network, and the high-order network robustness is analyzed by combining two improved indexes EII and WEII and basic index DC, so as to obtain the effectiveness of the proposed index, specifically:
[0096] According to the experimental results, the threshold is set. When the attack or failure causes the robustness value S in the network q When it drops to the corresponding threshold, the number of high-dimensional structures in the high-order network decreases with the destruction of the network, which means that the network is no longer connected; at the same time, the high-order network will become sparse from the initial state as the nodes are removed: the lower the threshold, the more the network can maintain its function with less loss of high-order structures, indicating that the high-order network is more robust. By evaluating the robustness value S calculated by the EII and WEII indicators under deliberate attacks q , the robustness of the network can be quantified, thus proving the effectiveness of the proposed indicator.
[0097] It should be noted that the present invention gradually generates 1-simplex, 2-simplex, and 3-simplex according to different random probabilities set for different simplex structures, and finally constructs a random high-order network with these three simplex structures, and uses PYTHON analysis software to simulate and analyze the process of identifying influential nodes in the above random high-order network. In the simulation experiment, the above random high-order network was analyzed in a deliberate attack mode, and the basic indicator degree centrality index (DC), the proposed high-order network important node identification index EII and WEII indicators were compared. Under these three indicators, the ratio of removed nodes q and the robustness value S q The relationship diagram between the removed node ratio q and the robustness value S under different indicators is analyzed. q The data relationship diagram between them; the relationship diagram of the number of 1-simplex, 2-simplex and 3-simplex removed by different indicators under deliberate attack in random high-order networks; the relationship diagram of the influence of the total number of simplex M on the network robustness under different indicators under deliberate attack when the node ratio is q=0.5; the relationship diagram of the influence of WEII indicator S under deliberate attack under different probabilities of 1-simplex, 2-simplex and 3-simplex q The relationship diagram of EII index under deliberate attack is analyzed, and S is analyzed under different probabilities of 1-simplex, 2-simplex and 3-simplex. q vs. q. Figures 4 to 9 , we get: In the interval of node removal ratio q from 0 to 0.1, when the important nodes identified by WEII index are removed one by one, the collapse speed of random high-order network is significantly accelerated. The reason is that the high sensitivity of WEII index is particularly obvious for high-dimensional simplex with complex connections in the network, which makes it particularly good at node identification in the initial stage. However, when the node removal ratio reaches about 0.12, the EII index begins to outperform WEII and DC indexes in identifying the remaining key nodes in the network. In addition, when the removal ratio reaches 0.2, the performance of EII index continues to be better than other indexes, while WEII index gradually fails in distinguishing the importance of the remaining nodes in the network. This is because with the continuous removal of nodes, the high-order structure in the network gradually decreases, and the network structure becomes more and more sparse, which eventually leads to the advantage of WEII index is no longer obvious, while EII index performs better in this case of network sparseness, that is, the network collapses faster under EII index. It can be seen that both WEII index and EII index of high-order network can accurately identify important nodes in the network. The former can effectively identify key nodes in the early stages of a network, while the EII indicator can effectively and continuously identify important nodes when the network becomes sparse, and has a wider range of applicability.
[0098] In addition, we found that the simplex scale also has a slight effect on the robustness of the network. As the simplex scale M increases from 1000 to 20000, S(q) of both indicators show an upward trend. When the simplex scale is large, the network robustness reflected by the EII and WEII indicators tends to be consistent, but the WEII indicator shows a slightly higher stability as a whole, showing its strong adaptability to large-scale network structure changes. Secondly, the connection probability of different simplexes will also affect the overall robustness of the network; the same conclusion is shown under the EII and WEII indicators. If the 3-simplex probability in the network is set larger, the network will collapse faster. This shows that the number of high-dimensional simplex structures in the network will have a great impact on the robustness of high-order networks.
[0099] Therefore, through simulation experiments, we concluded that in order to improve the accuracy of identifying high-influence nodes in random high-order networks, the three types of simplex structures, 1-simplex, 2-simplex, and 3-simplex, should be included as much as possible in the high-order network, especially to ensure that there are a large number of high-dimensional simplex structures in the network to ensure the density of the high-order network, so as to more accurately identify nodes with high influence.
[0100] It is understandable that the technical solution provided by this embodiment provides two new indicators for identifying important nodes in high-order networks. Compared with the node identification indicators in traditional complex networks, the indicators of this embodiment are more suitable for mining high-influence nodes in high-order networks based on different simplex structures. In addition, the proposed indicators not only take into account the types and numbers of simplexes, but also comprehensively consider their distribution probability in the network, which is particularly critical when evaluating network robustness and node importance. Therefore, these indicators have significant advantages in accuracy and practicality, and provide important theoretical support and practical guidance for high-order network analysis and design.
[0101] Embodiment 2
[0102] See also Fig.10 This embodiment provides a research system 100 for identifying influential nodes in a random high-order network, the system comprising:
[0103] The first construction module 101 is used to construct a random high-order network;
[0104] The second construction module 102 is used to propose an influential node identification index applicable to a high-order network, an extended influential node identification index EII and a weighted extended influential node identification index WEII;
[0105] The application module 103 is used to apply the proposed influential node identification index to the random high-order network, and evaluate the effectiveness and accuracy of the proposed index through network robustness.
[0106] It should be noted that the technical solution provided in this embodiment is applicable to the identification and analysis of influential nodes and the evaluation of robustness applied to scientific research cooperation, movie actors, biology, transportation, protein and other systems.
[0107] It can be understood that the technical solution provided in this embodiment includes a first construction module 101, which constructs a random high-order network; a second construction module 102, which is used to propose an influence node identification index suitable for a high-order network, an extended influence node identification index EII and a weighted extended influence node identification index WEII; and a third application module 103, which is used to apply the proposed influence node identification index to the random high-order network, and evaluate the effectiveness and accuracy of the proposed index through network robustness. According to the constructed random high-order networks based on different simplex structures, and according to the dimensional differences and different contribution degrees of different simplex structures in the network, the present invention proposes two extended influence node identification indicators EII and weighted extended influence node identification indicators WEII based on degree centrality, which are suitable for mining high-order networks based on simplex structures. According to the effectiveness of the proposed high-order network robustness indicator evaluation indicator, effective experimental results are obtained through experimental analysis. In order to improve the accuracy of identifying high-influence nodes in random high-order networks, three types of simplex structures, namely 1-simplex, 2-simplex and 3-simplex, should be included as much as possible in the high-order network, and the density and structural complexity of the network should be improved at the same time, so as to more accurately identify nodes with high influence.
[0108] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.
[0109] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.
[0110] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.
[0111] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0112] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0113] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0114] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.
[0115] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0116] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A research method for identifying influential nodes in random high-order networks, characterized by: The method comprises the following steps: Construct random high-order networks; The extended influence node identification index EII and the weighted extended influence node identification index WEII based on the node degree centrality index DC are proposed; The two proposed influential node identification indicators and the basic indicator degree centrality index DC are evaluated and verified through network robustness to prove the effectiveness and accuracy of the proposed indicators.
2. The research method for identifying influential nodes in a random high-order network according to claim 1 is characterized in that: The process of constructing the random high-order network is mainly: according to different simplex structures, different connection probabilities are assigned to construct a random high-order network with a dimension of d=3 including 1-simplex, 2-simplex, and 3-simplex structures. The steps of constructing the random high-order network include: Step 1: Initialization of the network: The network is initialized to contain N independent nodes, where N is the pre-set network size; in addition, the parameter M is set to the total number of simplexes in the entire high-order network; Step 2, Generation of 1-simplex structure: Based on Step 1, by traversing all node pairs in the network Nodes are connected to each other with probability p1 to form 1-simplex, that is, ordinary edge connection. For each pair of independent nodes in the network, a random number r between 0 and 1 is generated. If r < p1, a connection is established between these two nodes. Therefore, Step 1 forms the basic structure of the random high-order network; Step 3, Generation of 2-simplex structure: Based on Step 2, the network adds 2-simplex structures, i.e., "full" triangles, with probability p2 by traversing all possible node triples in the higher-order network For each triple, generate a random number r again. If r < p2, construct a fully connected closed triangle among these three nodes; Step 4, Generation of 3-simplex structure: Based on Step 3, the network adds 3-simplex structures, i.e., tetrahedrons, with probability p3 by traversing all possible node quadruples and generating a random number r for each quadruple. If r < p3, a fully connected tetrahedron is constructed among these four nodes; After completing the step-by-step construction of 1-simplex, 2-simplex and 3-simplex, a random high-order network with dimension d=3 is finally constructed. The random high-order network with dimension d=3 includes not only basic point-to-point connections, but also "full" triangles and tetrahedrons.
3. The research method for identifying influential nodes in a random high-order network according to claim 2 is characterized in that: The extended influence node identification index EII is defined as follows: Where d represents the dimension of the simplex in the high-order network, d = 0, 1, ..., n; Represents node v i The current associated d-simplex except node v i The ratio of the number of neighboring nodes outside to the number of d-simplex nodes; k d,0 (v i ) represents node v i The number of associated d-simplices.
4. The research method for identifying influential nodes in a random high-order network according to claim 3 is characterized in that: The weighted extended influence node identification index WEII is defined as follows: Where d represents the dimension of the simplex in the high-order network, d = 0, 1, ..., n; Represents node v i The current associated d-simplex except node v i The ratio of the number of neighboring nodes outside to the number of d-simplex nodes; k d,0 (v i ) represents node v i The number of associated d-simplices, W d represents the weight of a d-simplex of dimension d, W d Defined as: Among them, C d represents the structural complexity of the d-simplex; and The parameter representing the k-dimensional subsurface of the d-simplex; F d,k represents the number of k-dimensional subsurfaces of the d-simplex. In addition, k represents the dimension of the k-dimensional subsurfaces under the d-simplex in the high-order network.
5. The research method for identifying influential nodes in a random high-order network according to claim 4 is characterized in that: The steps of evaluating and verifying the proposed two influential node identification indicators and the basic indicator degree centrality indicator through network robustness to prove the effectiveness and accuracy of the proposed indicators include: According to the constructed random high-order network and the two influential node identification indicators proposed, an evaluation indicator suitable for evaluating the robustness of high-order networks is proposed, namely the high-order simplex retention rate HOR. HOR is the ratio of the weighted number of various simplices remaining in the network after the network is attacked to the weighted number of all simplices before the attack. It is used to measure the structural integrity and functional retention of the network after being damaged. The calculation formula of HOR is as follows: Among them, M'1, M'2 and M'3 represent the number of 1-simplex, 2-simplex and 3-simplex remaining in the network after the attack, respectively, and M1, M2 and M3 represent the number of 1-simplex, 2-simplex and 3-simplex in the network before the attack, respectively; W1, W2 and W3 are the weight values given to 1-simplex, 2-simplex and 3-simplex when evaluating the importance of network structure, respectively, reflecting the contribution of different simplices to network function; here, the weights of simplices of different dimensions are set as 1-simplex weight W1=1, 2-simplex weight W2=2, 3-simplex weight W3=6, and HOR is obtained by evaluating the HOR obtained by EII index and WEII index under different attack intensities, thereby quantifying the current robustness of the network after attack; Later, through the posterior robustness measure S q , based on the changes of simplexes of different dimensions, specifically, for the posterior measure HOR considering high-order network damage, the robustness measure S q Written as follows: The a posteriori robustness measure S q It is actually the weighted average result of the HOR value, which is used to analyze how the overall performance or state of the network changes with time or other factors. The posterior robustness metric S q It provides a global perspective to observe the robustness of the network, which helps to understand the average performance of the network in the long term or under changing conditions.
6. A research system for identifying influential nodes in random high-order networks, characterized by: The system comprises: The first building block builds a random high-order network with dimension d=3; The second building block proposes the extended influence node identification index EII and the weighted extended influence node identification index WEII based on the node degree centrality index DC; In the third application module, the two proposed influential node identification indicators and the basic indicator degree centrality indicator DC are evaluated and verified through network robustness to prove the effectiveness and accuracy of the proposed indicators.
Citation Information
Patent Citations
Multi-layer super network influence node identification method based on evidence theory
CN116248523A
Information propagation control method based on network high-order structure
CN117061574A
Method and system for analyzing network scale and robustness of multi-layer high-order network
CN118200159A
Method and apparatus for evaluating the importance of complex network nodes
JP7224700B1
Cited By
Graph neural network anomaly detection method based on high-order topology and personalized PageRank
CN120633707A
