Label propagation overlapping community detection method based on core community attraction mechanism

Abstract

The invention provides a label propagation overlapping community detection method based on a core community attraction mechanism. The label propagation overlapping community detection method comprises the steps of constructing a network model and preprocessing; identifying core communities in the network model and distributing a unique label for each core community; spreading the label and adjusting the spreading weight; the nodes are allowed to belong to a plurality of core communities at the same time, label distribution of the nodes is adjusted according to the association degree of the nodes and different core communities, and an overlapped core community structure is formed; evaluating a label propagation result; and according to an evaluation result, adjusting a core community attraction mechanism and label propagation parameters, and optimizing a core community detection effect. According to the method, the core communities in the network are identified, and the label propagation process is guided by using the core communities, so that the label propagation precision is enhanced; according to the method, a self-adaptive mechanism is adopted, the label propagation path and the weight are adjusted in real time according to the interaction frequency and the correlation degree between the nodes, the unnecessary calculation overhead is reduced, and the propagation efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of data mining, and in particular to a label propagation overlapping community detection method based on a core community attraction mechanism. Background Art

[0002] With the continuous development of social informatization, complex networks are increasingly used in various fields. Complex networks are not only widely used in social networks, biological networks, communication networks, etc., but also play an important role in information dissemination, virus diffusion, economic systems, etc. In these complex networks, the core community structure is one of its core characteristics. Core community detection, that is, identifying highly connected groups of nodes in the network, is one of the important means to understand the structure and function of the network. Traditional core community detection methods mostly focus on non-overlapping core communities, but in many networks in the real world, such as social networks and biological networks, nodes often belong to multiple core communities at the same time. Therefore, overlapping core community detection has become a difficult problem that needs to be solved urgently.

[0003] Among the existing core community detection methods, label propagation algorithms have attracted widespread attention due to their simplicity and high efficiency. Traditional label propagation algorithms detect core community structures by mutual propagation and updating of labels between nodes. However, this method has problems of insufficient accuracy and efficiency when dealing with large-scale complex networks. Especially in overlapping core community detection, nodes may belong to multiple core communities at the same time. Traditional label propagation algorithms are difficult to accurately distinguish and detect these overlapping core communities, resulting in inaccurate results.

[0004] The application of traditional label propagation algorithms in large-scale complex networks also faces some challenges. First, due to the complexity and diversity of the network, the label propagation process is easily interfered by noise, which affects the accuracy of core community detection. Secondly, as the scale of the network increases, the amount of calculation increases sharply, resulting in a slower convergence of the algorithm and affecting the overall performance. In addition, it is difficult for traditional methods to achieve efficient calculation and fast convergence while ensuring accuracy, which brings great limitations in practical applications. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a label propagation overlapping community detection method based on the core community attraction mechanism. By introducing the core community attraction mechanism, the present invention can more accurately identify and distinguish overlapping core communities, reduce noise interference, and improve detection accuracy. At the same time, by optimizing the calculation process, the convergence speed of the algorithm is significantly improved to achieve efficient calculation.

[0006] The technical solution of the present invention is: a label propagation overlapping community detection method based on a core community attraction mechanism, comprising the following steps:

[0007] S1), construct a network model by representing nodes as individuals and edges as relationships between individuals;

[0008] S2), preprocessing the network model G;

[0009] S3), identifying the core community in the network model G by a method based on node degree centrality;

[0010] S4), initialize the label propagation process and assign a unique label to each core community;

[0011] S5), starting from the core community, propagating labels according to the degree of association between nodes and the core community, and adjusting the propagation weight based on the number and weight of edges;

[0012] S6), gradually covering the entire network model G, ensuring that all nodes are assigned labels;

[0013] S7) Allowing nodes to belong to multiple core communities at the same time, adjusting the node label assignment according to the degree of association between the node and different core communities, and forming an overlapping core community structure;

[0014] S8), evaluate the label propagation results, and evaluate the core community division results through modularity and core community density;

[0015] S9) According to the evaluation results, adjust the core community attraction mechanism and label propagation parameters to optimize the core community detection effect;

[0016] S10), repeat the label propagation and evaluation optimization process until the core community detection result reaches the preset standard.

[0017] Preferably, in step S1), the construction of the network model G specifically includes the following steps:

[0018] S11), define the node set V and edge set E of the network model; wherein, the node v∈V represents an individual, and the edge e∈E represents the relationship between individuals;

[0019] S12), by collecting network data, determine the initial set of nodes v and edges e, and record the attribute information of each node and edge;

[0020] S13), construct an adjacency matrix A, where the element A of the adjacency matrix A ij Represents node v i and node v j If there is an edge e ij , then A ij =1, otherwise A ij =0;

[0021] S14), for each edge eij Assign weight w ij ;

[0022] S15) Construct the graph structure of the network model, namely:

[0023] G = (V, E, W);

[0024] Among them, V represents the node set, E represents the edge set, and W represents the edge weight set.

[0025] Preferably, in step S14), the weight w ij Determined by the strength or frequency of the relationship between nodes, including using the number of connections or interaction frequency to determine the weight w ij .

[0026] Preferably, in step S2), preprocessing the network model G includes removing noise nodes and edges, and normalizing the weights of the edges.

[0027] Preferably, in step S3), the core community in the network model G is identified based on the node degree centrality method, which specifically includes the following steps:

[0028] S31), calculating the degree d(v) of each node v∈V, where the degree d(v) represents the number of edges directly connected to the node v;

[0029] S32), based on the degree centrality of the nodes, determine the initial core node set C, where nodes with higher degree centrality are preferentially selected as core nodes;

[0030] S33), further screening the core nodes by calculating the weighted connectivity between the nodes to form the final core community set;

[0031] S34), introduce weighted modularity Q w As an evaluation indicator, determine the boundaries of the core community;

[0032] S35), according to the weighted modularity Q w The calculation results are used to adjust the scope of the core community to ensure that the core community has strong internal connectivity and obvious boundaries;

[0033] S36), using the identified core community as the starting point of the label propagation process, and assigning a unique label to each core community.

[0034] Preferably, in step S4), each core community is assigned a unique label, which specifically includes the following steps:

[0035] S41) For each core community C i ∈C assigns a unique label L i, where i is the index of the core community;

[0036] S42), setting the initial labels of all nodes in each core community to the label L of the core community i , that is, for the core community C i All nodes v∈C i , set L(v) = L i ;

[0037] S43), initializing a label set L(v) on each node in the network model G, wherein the label set L(v) includes the label of the core community to which the node belongs;

[0038] S44), record each node and its corresponding label set L(v) to form an initial label distribution.

[0039] Preferably, in step S5), starting from the core community, the label is propagated according to the association between the node and the core community, and the propagation weight is adjusted based on the number and weight of the edges, which specifically includes the following steps:

[0040] S51), from each core community C i Initially, nodes in the core community are selected as the starting point for label propagation;

[0041] S52), calculate each node v and the core community C i The correlation R(v,C i );

[0042] S53), according to the node v and the core community C i The correlation R(v,C i ), determine the priority of label propagation, and nodes with higher priority will propagate labels first;

[0043] S54) During the label propagation process, dynamically adjust the label propagation weight W based on the number of edges and the weight of the edge p (v);

[0044] S55), label L i Spread to the nodes directly connected to the core community nodes, and according to the propagation weight W p (v) Dynamically update the label set L(v) of these nodes;

[0045] S56), repeat the label propagation process until all nodes receive labels from the core community.

[0046] Preferably, in step S7), a node is allowed to belong to multiple core communities at the same time, and the label allocation of the node is adjusted according to the association degree between the node and different core communities to form an overlapping core community structure, which specifically includes the following steps:

[0047] S71), during the label propagation process, record the label set L(v) of each node v, and the initial label set comes from the labels of the core community;

[0048] S72), according to the correlation R(v,C i ), computing node v belongs to each core community C i The probability P(v,C i );

[0049] S73), according to the probability P(v,C i ), adjust the label set L(v) of each node v to ensure that the node can belong to multiple core communities at the same time;

[0050] S74), in each iteration, dynamically updating the label set L(v) of each node based on the interaction between the node and its neighboring nodes;

[0051] S75), repeating the label propagation and adjustment process until the label set L(v) of all nodes is stable and an accurate overlapping core community structure is formed;

[0052] S76), finally outputting the label set L(v) of each node, and evaluating and verifying the overlapping core community structure formed by the label set.

[0053] Preferably, in step S8), the label propagation result is evaluated, and the core community division result is evaluated by modularity and core community density, which specifically includes the following steps:

[0054] S81), perform a preliminary evaluation on the label propagation results, and record the final label set L(v) of each node v;

[0055] S82), calculating the modularity Q of the core community division;

[0056] S83) Calculate the core community density D (C i );

[0057] S84), further evaluate the core community division results, calculate the average core community density Dˉ and the average modularity Qˉ; and comprehensively evaluate the label propagation results based on the average core community density Dˉ and the average modularity Qˉ;

[0058] S85) According to the evaluation results, necessary adjustments and optimizations are made to the label propagation process, and finally the evaluated and optimized core community division results are output, and a corresponding evaluation report is generated.

[0059] Preferably, in step S9), according to the evaluation results, the core community attraction mechanism and label propagation parameters are adjusted to optimize the core community detection effect, which specifically includes the following steps:

[0060] S91) According to the core community division evaluation results, analyze the modularity Q and core community density D (C i ) changes;

[0061] S92), adjust the core community attraction mechanism by recalculating the degree centrality and weighted connectivity W(v i ,v j ), optimize the selection and allocation of core communities and redefine the core community set C;

[0062] S93), according to the new core community set C, reinitialize the label propagation process and assign a unique label to each core community;

[0063] S94), adjusting label propagation parameters, including propagation weight and propagation probability;

[0064] S95) Optimize the label propagation algorithm and adopt an adaptive mechanism to adjust the label propagation path and weight in real time according to the interaction frequency and correlation between nodes;

[0065] S96), based on the correlation between nodes and core communities R(v,C i ) and the propagation weight W p (v), dynamically adjust the label set L(v) of each node;

[0066] S97) Through multiple iterations, the label propagation process is continuously adjusted and optimized.

[0067] The beneficial effects of the present invention are:

[0068] 1. The present invention enhances the accuracy of label propagation by identifying core communities in the network and using these core communities to guide the label propagation process;

[0069] 2. Based on node degree centrality and weighted connectivity, the present invention ensures that the selected core communities have strong internal connectivity and obvious boundaries, thereby accurately reflecting the actual structure of the network, and allows nodes to belong to multiple core communities at the same time, and adjusts label allocation according to the association between nodes and different core communities to form an overlapping core community structure;

[0070] 3. The present invention calculates the association between nodes and core communities and makes dynamic adjustments based on propagation weights and propagation probabilities, making the label propagation process more accurate and ultimately obtaining more detailed core community division results;

[0071] 4. The present invention adopts an adaptive mechanism to adjust the label propagation path and weight in real time according to the interaction frequency and correlation between nodes, reducing unnecessary computing overhead and improving propagation efficiency;

[0072] 5. The present invention dynamically adjusts the propagation parameters so that the algorithm can converge quickly when processing large-scale complex networks. During the label propagation process, the propagation weight and propagation probability are accurately adjusted based on the number of edges and weights to ensure the effectiveness and accuracy of label propagation, thereby reducing the amount of calculation and improving the algorithm performance.

[0073] 6. The present invention comprehensively evaluates the label propagation results through indicators such as modularity and core community density to ensure the accuracy and rationality of the core community division results;

[0074] 7. The present invention uses modularity to evaluate the overall quality of core community division and uses core community density to evaluate the compactness within the core community, thereby performing comprehensive quality control on the core community detection results. According to the evaluation results, the core community attraction mechanism and label propagation parameters are dynamically adjusted to further optimize the core community detection effect; through multiple iterations and continuous optimization, the final core community division results are ensured to have high accuracy and effectiveness;

[0075] 8. The present invention significantly reduces the amount of calculation by introducing a core community attraction mechanism and dynamic adjustment and optimization during the label propagation process, so that the algorithm can run efficiently in large-scale complex networks. The adaptive label propagation mechanism enables the algorithm to converge quickly, reduces the number of iterations, and improves the overall performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0077] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings:

[0078] like Figure 1 As shown, this embodiment provides a label propagation overlapping community detection method based on a core community attraction mechanism, including the following steps:

[0079] S1) Constructing a network model G by representing nodes as individuals and edges as relationships between individuals; specifically comprising the following steps:

[0080] S11), define the node set V and edge set E of the network model; wherein, the node v∈V represents an individual, and the edge e∈E represents the relationship between individuals;

[0081] S12), by collecting network data, determine the initial set of nodes v and edges e, and record the attribute information of each node and edge;

[0082] S13), construct an adjacency matrix A, where the element A of the adjacency matrix A ij Represents node v i and node v j If the relationship between node v i and node v j There is an edge e between ij , then A ij =1, otherwise A ij =0; i and j represent the index of the node respectively;

[0083] S14), for each edge e ij Assign the corresponding weight w ij , the weight w ij Determined by the strength or frequency of the relationship between nodes, including using the number of connections or interaction frequency to determine the weight w ij ;

[0084] S15) Construct the graph structure of the network model, namely:

[0085] G = (V, E, W);

[0086] Among them, V represents the node set, E represents the edge set, and W represents the edge weight set.

[0087] S2), preprocessing the network model G;

[0088] The preprocessing of this embodiment is to clean and standardize the network data by formulating a set of preset rules, which are determined based on the properties of nodes and edges and the relationships between them; including removing noise nodes and edges, and standardizing the weights of edges.

[0089] In addition, this embodiment can also monitor the network status in real time, identify available resources and interference conditions, clean up and standardize network data, and through the process of spectrum sensing, help identify which parts of the network have interference and which parts have richer resources, thereby providing reliable data support for subsequent core community identification.

[0090] S3) Identify the core community in the network model G by using a method based on node degree centrality; specifically, the following steps are included:

[0091] S31), calculating the degree d(v) of each node v∈V, where the degree d(v) represents the number of edges directly connected to the node v;

[0092] S32), based on the degree centrality of the nodes, determine the initial core node set C, where nodes with higher degree centrality are preferentially selected as core nodes;

[0093] S33), further screening core nodes by calculating the weighted connectivity between nodes to form a final core community set; the weighted connectivity W (v i , v j ) is calculated by the following formula:

[0094]

[0095] Among them, A ij Represents node v i and node v j The relationship between ij For node v i and node v j The weight of the edge between i ) and d(v j ) are nodes v i and node v j degree;

[0096] S34), introduce weighted modularity Q w As an evaluation indicator, the boundaries of the core community are determined; among them, the weighted modularity Q introduced w The expression is:

[0097]

[0098] Where mw is the total weight of all edges in the network; A ij Represents node v i and node v j The relationship between i and w j The nodes v i and node v j The total weight of δ(c i , c j ) represents the indicator function, which is used to determine the node point v i and node v j Whether they belong to the same core community; when node v i and node v j The value is 1 when they belong to the same core community, otherwise it is 0; c i 、c j They are node points v i and node v j Community tags.

[0099] S35), according to the weighted modularity Q wThe calculation results are used to adjust the scope of the core community to ensure that the core community has strong internal connectivity and obvious boundaries;

[0100] S36), using the identified core community as the starting point of the label propagation process, and assigning a unique label to each core community.

[0101] S4) Initialize the label propagation process and assign a unique label to each core community; specifically, the steps include:

[0102] S41) For each core community C i ∈C assigns a unique label L i , where i is the index of the core community;

[0103] S42), setting the initial labels of all nodes in each core community to the label L of the core community i , that is, for the core community C i All nodes v∈C i , set L(v) = L i ;

[0104] S43), initializing a label set L(v) on each node in the network model G, wherein the label set L(v) includes the label of the core community to which the node belongs;

[0105] S44), record each node and its corresponding label set L(v) to form an initial label distribution.

[0106] S5) Starting from the core community, the label is propagated according to the association between the node and the core community, and the propagation weight is adjusted based on the number and weight of the edges; specifically, the steps include:

[0107] S51), from each core community C i Initially, nodes in the core community are selected as the starting point for label propagation;

[0108] S52), calculate each node v and the core community C i The correlation R(v, C i ); the correlation R(v, C i ) is determined based on the number of edges between the node and the core community and the weight of the edge. The association R(v, C i ) is calculated as:

[0109]

[0110] Among them, A uv is the relationship between node u and node v. If there is an edge e between node u and node v uv , then A uv=1, otherwise A uv =0; w uv is the weight of the edge between node u and node v, d(u) and d(v) are the degrees of node u and node v respectively;

[0111] S53), according to the node v and the core community C i The correlation R(v, C i ), determine the priority of label propagation, and nodes with higher priority will propagate labels first;

[0112] S54) During the label propagation process, dynamically adjust the label propagation weight W based on the number of edges and the weight of the edge p (v); The propagation weight W of the label p The expression of (v) is:

[0113]

[0114] Where N(v) represents the set of neighboring nodes directly connected to node v; R(u, C i ) represents the relationship between node u and core community C i The degree of association; A uv is the relationship between node u and node v; d(u) is the degree of node u; w uv is the weight of the edge between node u and node v;

[0115] S55), label L i is propagated to the nodes directly connected to the nodes in the core community and is calculated according to the propagation weight W p (v) Dynamically update the label set L(v) of these nodes;

[0116]

[0117] Where N(v) represents the set of neighbor nodes directly connected to node v; A uv is the relationship between node u and node v, w uv is the weight of the edge between node u and node v;

[0118] S56), repeat the label propagation process until all nodes receive labels from the core community.

[0119] S6), gradually covering the entire network model G, ensuring that all nodes are assigned labels;

[0120] S7) allows a node to belong to multiple core communities at the same time, and adjusts the node label allocation according to the degree of association between the node and different core communities to form an overlapping core community structure; specifically includes the following steps:

[0121] S71), during the label propagation process, record the label set L(v) of each node v, and the initial label set comes from the labels of the core community;

[0122] S72), according to the association degree R(v, C i ), computing node v belongs to each core community C i The probability P(v, C i ),Right now:

[0123]

[0124] Among them, ∑ j R(v,C j ) represents the total association between node v and all core communities;

[0125] S73), according to the probability P(v, C i ), adjust the label set L(v) of each node v to ensure that the node can belong to multiple core communities at the same time;

[0126] L(v)=∪ i {L i ·P(v,C i )}

[0127] In the formula, ∪ i Represents the set L of all possible core community labels i and the corresponding probability P(v, C i ) performs a union operation;

[0128] S74), in each iteration, dynamically updating the label set L(v) of each node based on the interaction between the node and its neighboring nodes;

[0129] S75), repeating the label propagation and adjustment process until the label set L(v) of all nodes is stable and an accurate overlapping core community structure is formed;

[0130] S76), finally outputting the label set L(v) of each node, and evaluating and verifying the overlapping core community structure formed by the label set.

[0131] S8) Evaluate the label propagation results and evaluate the core community division results through modularity and core community density; specifically, the following steps are included:

[0132] S81), perform a preliminary evaluation on the label propagation results, and record the final label set L(v) of each node v;

[0133] S82), calculating the modularity Q of the core community division; the calculation formula of the modularity Q is:

[0134]

[0135] Where m is the total number of edges in the network model; A ij Represents node v i and node v j The relationship between d(v i ) represents node v i degree; d(v j ) represents node v j degree; δ(c i , c j ) represents the indicator function, which is used to determine the node point v i and node v j Whether they belong to the same core community; when node v i and node v j The value is 1 when they belong to the same core community, otherwise it is 0; C i 、c j They are node points v i and node v j Community tags.

[0136] S83) Calculate the core community density D (C i ); the core community density D(C i ) is calculated as:

[0137]

[0138] In the formula, |E(C i )|For the core community C i The number of edges in |V(C i )|For the core community C i The number of nodes in .

[0139] S84) Further evaluate the core community division results and calculate the average core community density and average modularity And based on the average core community density and average modularity Conduct a comprehensive evaluation of label propagation results;

[0140] The average core community density and average modularity The calculation formulas are:

[0141]

[0142]

[0143] Where C is the total number of core communities; D(C i ) is the core community C i The density of Q(C i ) is the core community C i The modularity.

[0144] S85) According to the evaluation results, necessary adjustments and optimizations are made to the label propagation process, and finally the evaluated and optimized core community division results are output, and a corresponding evaluation report is generated.

[0145] S9) According to the evaluation results, adjust the core community attraction mechanism and label propagation parameters to optimize the core community detection effect; specifically, the following steps are included:

[0146] S91) According to the core community division evaluation results, analyze the modularity Q and core community density D (C i ) changes in ΔQ and ΔD (C i );Right now:

[0147]

[0148]

[0149] In the formula, C is the total number of core communities; m is the total number of edges in the network model; A ij Represents node v i and node v j The relationship between d(v i ) represents node v i degree; d(v i ) represents node v i degree; d(v j ) represents node v j degree; |E(C i )|For the core community C i The number of edges in |V(C i )|For the core community C i The number of nodes in

[0150] S92), adjust the core community attraction mechanism by recalculating the degree centrality and weighted connectivity W(v i , v j ), optimize the selection and allocation of core communities and redefine the core community set C;

[0151]

[0152] Where wij is the node v i and node v j The weight of the edge between ij Represents node vi and node v j The relationship between d(v i ) represents node v i degree; d(v i ) represents node v i degree; d(v j ) represents node v j degree; w uv is the weight of the edge between node u and node v;

[0153] S93), according to the new core community set C, reinitialize the label propagation process and assign a unique label to each core community;

[0154] S94), adjusting label propagation parameters, including propagation weight and propagation probability;

[0155] S95) Optimize the label propagation algorithm and adopt an adaptive mechanism to adjust the label propagation path and weight in real time according to the interaction frequency and correlation between nodes;

[0156]

[0157] Where AdaptiveWeight(v) is the adaptive weight of the node v;

[0158] S96), based on the correlation between the node and the core community R(v, C i ) and the propagation weight W p (v), dynamically adjust the label set L(v) of each node;

[0159]

[0160] In the formula, ∪ i Indicates; L i For the core community C i Label; R(v,C i ) is the node v and the core community C i The correlation degree of j R(v,C j ) represents the total association between node v and all core communities; A uv is the relationship between node u and node v, w uv is the weight of the edge between node u and node v;

[0161] S97) Through multiple iterations, the label transmission process is continuously adjusted and optimized; namely:

[0162]

[0163] Where T is the number of iterations and σ is a control parameter used to balance the changes in modularity and core community density.

[0164] S10), repeat the label propagation and evaluation optimization process until the core community detection result reaches the preset standard.

[0165] The above embodiments and descriptions are only for illustrating the principles and best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, all of which fall within the scope of the present invention to be protected.

Claims

1. A label propagation overlapping community detection method based on core community attraction mechanism, characterized in that: The steps include: S1), building a network model; S2), preprocessing the network model; S3), identifying core communities in network models through a method based on node degree centrality; S4), initialize the label propagation process and assign a unique label to each core community; S5), starting from the core community, propagating labels according to the correlation between nodes and the core community and adjusting the propagation weight; S6), gradually covering the entire network model to ensure that all nodes are assigned labels; S7) Allowing nodes to belong to multiple core communities at the same time, adjusting the node label assignment according to the degree of association between the node and different core communities, and forming an overlapping core community structure; S8), evaluate the label propagation results, and evaluate the core community division results through modularity and core community density; S9) According to the evaluation results, adjust the core community attraction mechanism and label propagation parameters to optimize the core community detection effect; S10), repeat the label propagation and evaluation optimization process until the core community detection result reaches the preset standard.

2. According to claim 1, a label propagation overlapping community detection method based on core community attraction mechanism is characterized by: In step S1), the construction of the network model G specifically includes the following steps: S11), define the node set V and edge set E of the network model; wherein, the node v∈v represents an individual, and the edge e∈E represents the relationship between individuals; S12), by collecting network data, determine the initial set of nodes v and edges e, and record the attribute information of each node and edge; S13), construct an adjacency matrix A, where the element A of the adjacency matrix A ij Represents node v i and node v j If there is an edge e ij , then A ij =1, otherwise A ij =0; S14), for each edge e ij Assign weight w ij ; S15) Construct the graph structure of the network model, namely: G = (V, E, W); Among them, V represents the node set, E represents the edge set, and W represents the edge weight set.

3. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 2 is characterized by: In step S3), the core community in the network model G is identified based on the node degree centrality method, which specifically includes the following steps: S31), calculating the degree d(v) of each node v∈V, where the degree d(v) represents the number of edges directly connected to the node v; S32), based on the degree centrality of the nodes, determine the initial core node set C, where nodes with higher degree centrality are preferentially selected as core nodes; S33), further screening the core nodes by calculating the weighted connectivity between the nodes to form the final core community set; S34), introduce weighted modularity Q w As an evaluation indicator, determine the boundaries of the core community; S35), according to the weighted modularity Q w The calculation results are used to adjust the scope of the core community to ensure that the core community has strong internal connectivity and obvious boundaries; S36), using the identified core community as the starting point of the label propagation process, and assigning a unique label to each core community.

4. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 3 is characterized by: In step S4), each core community is assigned a unique label, which specifically includes the following steps: S41) For each core community C i ∈C assigns a unique label L i , where i is the index of the core community; S42), setting the initial labels of all nodes in each core community to the label L of the core community i , that is, for the core community C i All nodes v∈C i , set L(v) = L i ; S43), initializing a label set L(v) on each node in the network model G, wherein the label set L(v) includes the label of the core community to which the node belongs; S44), record each node and its corresponding label set L(v) to form an initial label distribution.

5. According to claim 4, a label propagation overlapping community detection method based on core community attraction mechanism is characterized by: In step S5), starting from the core community, the label is propagated according to the association between the node and the core community, and the propagation weight is adjusted based on the number and weight of the edges, which specifically includes the following steps: S51), from each core community C i Initially, nodes in the core community are selected as the starting point for label propagation; S52), calculate each node v and the core community C i The correlation R(v, C i ); S53), according to the node v and the core community C i The correlation R(v, C i ), determine the priority of label propagation, and nodes with higher priority will propagate labels first; S54) During the label propagation process, dynamically adjust the label propagation weight W based on the number of edges and the weight of the edge p (v); S55), label L i Spread to the nodes directly connected to the core community nodes, and according to the propagation weight W p (v) Dynamically update the label set L(v) of these nodes; S56), repeat the label propagation process until all nodes receive labels from the core community.

6. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 5 is characterized by: In step S7), a node is allowed to belong to multiple core communities at the same time. According to the degree of association between the node and different core communities, the label allocation of the node is adjusted to form an overlapping core community structure, which specifically includes the following steps: S71), during the label propagation process, record the label set L(v) of each node v, and the initial label set comes from the labels of the core community; S72), according to the correlation R(v,C i ), computing node v belongs to each core community C i The probability P(v,C i ); S73), according to the probability P(v,C i ), adjust the label set L(v) of each node v to ensure that the node can belong to multiple core communities at the same time; S74), in each iteration, dynamically updating the label set L(v) of each node based on the interaction between the node and its neighboring nodes; S75), repeating the label propagation and adjustment process until the label set L(v) of all nodes is stable and an accurate overlapping core community structure is formed; S76), finally outputting the label set L(v) of each node, and evaluating and verifying the overlapping core community structure formed by the label set.

7. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 6 is characterized by: In step S8), the label propagation result is evaluated, and the core community division result is evaluated by modularity and core community density, which specifically includes the following steps: S81), perform a preliminary evaluation on the label propagation results, and record the final label set L(v) of each node v; S82), calculating the modularity Q of the core community division; S83) Calculate the core community density D (C i ); S84) Further evaluate the core community division results and calculate the average core community density and average modularity And based on the average core community density and average modularity Conduct a comprehensive evaluation of label propagation results; S85) According to the evaluation results, necessary adjustments and optimizations are made to the label propagation process, and finally the evaluated and optimized core community division results are output, and a corresponding evaluation report is generated.

8. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 7 is characterized by: In step S9), according to the evaluation results, the core community attraction mechanism and label propagation parameters are adjusted to optimize the core community detection effect, which specifically includes the following steps: S91) According to the core community division evaluation results, analyze the modularity Q and core community density D (C i ) changes; S92), adjust the core community attraction mechanism by recalculating the degree centrality and weighted connectivity W(v i , v j ), optimize the selection and allocation of core communities and redefine the core community set C; S93), according to the new core community set C, reinitialize the label propagation process and assign a unique label to each core community; S94), adjusting label propagation parameters, including propagation weight and propagation probability; S95) Optimize the label propagation algorithm and adopt an adaptive mechanism to adjust the label propagation path and weight in real time according to the interaction frequency and correlation between nodes; S96), based on the correlation between the node and the core community R(v, C i ) and the propagation weight W p (v), dynamically adjust the label set L(v) of each node; S97) Through multiple iterations, the label propagation process is continuously adjusted and optimized.

9. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 3 is characterized by: In step S3), the weighted connectivity W(v i , v j ) is calculated by the following formula: Among them, A ij Represents node v i and node v j The relationship between ij For node v i and node v j The weight of the edge between i ) and d(v j ) are nodes v i and node v j The degree of weighted modularity Q w The expression is: Among them, m w is the total weight of all edges in the network; A ij Represents node v i and node v j The relationship between i and w j The nodes v i and node v j The total weight of δ(c i , c j ) represents the indicator function, which is used to determine the node point v i and node v j Whether they belong to the same core community; when node v i and node v j The value is 1 when they belong to the same core community, otherwise it is 0; c i 、c j They are node points v i and node v j Community tags.

10. The label propagation overlapping community detection method based on core community attraction mechanism according to claim 5, characterized in that: In step S5), the correlation R(v, C i ) is calculated as: Among them, A uv is the relationship between node u and node v. If there is an edge e between node u and node v uv , then A uv =1, otherwise A uv =0; w uv is the weight of the edge between node u and node v, d(u) and d(v) are the degrees of node u and node v respectively; The label propagation weight W p The expression of (v) is: Where N(v) represents the set of neighboring nodes directly connected to node v; R(u, C i ) represents the relationship between node u and core community C i The degree of association; A uv is the relationship between node u and node v; d(u) is the degree of node u; w uv is the weight of the edge between node u and node v.

Images