Dynamic talent team discovery method, device and equipment based on complex network analysis

Abstract

The application relates to a dynamic talent team discovery method, device and equipment based on complex network analysis. The method comprises the following steps: acquiring a talent correlation matrix and a research direction matrix; performing similarity sorting on each talent and other talents according to the talent correlation matrix and the research direction matrix to obtain a sorting vector, performing information fusion on the sorting vector, then performing singular value decomposition, extracting the first column of a left singular vector, obtaining neighbors of each talent based on the similarity, and constructing a talent unified graph; performing non-negative matrix decomposition on the talent unified graph to obtain a talent affiliation indication matrix, calculating the similarity degree of the talent in team affiliation according to the talent affiliation indication matrix, minimizing the difference between the talent unified graph and the similarity degree of the talent in team affiliation, imposing a historical evolution constraint, and constructing an objective function; and performing dynamic talent team division according to the optimal talent affiliation indication matrix obtained by solving the objective function at each moment. The method can adjust team division in real time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of talent matching, in particular to a dynamic talent team discovery method, device and equipment based on complex network analysis. BACKGROUND

[0002] It is always an important research topic in the field of human resources to accurately and efficiently locate professional talents meeting the needs and quickly form a systematic professional force, i.e., the problem of talent team discovery. The problem of talent team mining based on similar research directions has always been a hot topic in this field. In actual scenarios, a group of high-tech professional talents with the same research direction or knowledge and skills are often needed to support specific work tasks. However, the key information of talents such as research direction usually has the characteristics of time evolution, which requires real-time mining of talent teams with similar research directions in a dynamically changing environment.

[0003] Traditional human resource research mostly adopts qualitative analysis, but in a dynamically changing environment, the traditional method cannot effectively support real-time team discovery. Therefore, artificial intelligence technology provides a more objective, comprehensive perspective and intelligent tool for the discovery of talent teams. In order to better develop talent teams in the information age, it is necessary to use more quantitative and intelligent means to mine talent teams, so as to provide a new decision support method for talent development and management. Based on the related theories and methods of complex networks, the limitations of intuition or personal experience are broken, and more scientific decision-making basis is provided, which will open up a new direction for the discovery of talent teams. SUMMARY

[0004] Therefore, it is necessary to provide a dynamic talent team discovery method, device and equipment based on complex network analysis to solve the above technical problems.

[0005] A dynamic talent team discovery method based on complex network analysis, the method comprising:

[0006] constructing a talent correlation matrix according to the academic correlation relationship between talents, and constructing a research direction matrix according to the research direction of talents;

[0007] performing similarity sorting on each talent and other talents according to the talent correlation matrix and the research direction matrix to obtain a sorting vector, performing multi-dimensional information fusion on the sorting vector to obtain a sorting matrix, performing singular value decomposition on the sorting matrix, extracting the first column of the left singular vector, selecting the top K talents with the highest similarity as neighbors of the current talent, and constructing a talent unified graph according to each talent and the corresponding neighbors; the nodes of the talent unified graph represent talents, and the edges represent the correlation relationship between talents;

[0008] performing non-negative matrix factorization on the talent uniform graph to obtain a talent membership indication matrix, calculating a similarity degree of talent team membership according to the talent membership indication matrix, minimizing a difference between the talent uniform graph and the similarity degree of talent team membership, obtaining a talent team division result, and imposing a historical evolution constraint on the talent team division result to construct an objective function;

[0009] optimizing and solving the objective function at each time to obtain an optimal talent membership indication matrix, and performing dynamic talent team division according to the optimal talent membership indication matrix.

[0010] A dynamic talent team discovery device based on complex network analysis, the device comprising:

[0011] a parameter acquisition module configured to construct a talent correlation matrix according to academic correlation relationships between talents, and construct a research direction matrix according to research directions of the talents;

[0012] a graph construction module configured to perform similarity degree sorting on each talent and other talents according to the talent correlation matrix and the research direction matrix to obtain a sorting vector, perform multi-dimensional information fusion on the sorting vector to obtain a sorting matrix, perform singular value decomposition on the sorting matrix, extract a first column of a left singular vector, select the first K talents with the highest similarity degrees as neighbors of a current talent, and construct a talent uniform graph according to each talent and the corresponding neighbors; the nodes of the talent uniform graph represent talents, and the edges represent correlation relationships between the talents;

[0013] a model construction module configured to perform non-negative matrix factorization on the talent uniform graph to obtain a talent membership indication matrix, calculate a similarity degree of talent team membership according to the talent membership indication matrix, minimize a difference between the talent uniform graph and the similarity degree of talent team membership, obtain a talent team division result, and impose a historical evolution constraint on the talent team division result to construct an objective function;

[0014] a model solving module configured to optimize and solve the objective function at each time to obtain an optimal talent membership indication matrix, and perform dynamic talent team division according to the optimal talent membership indication matrix.

[0015] A computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0016] constructing a talent correlation matrix according to academic correlation relationships between talents, and constructing a research direction matrix according to research directions of the talents;

[0017] According to the talent correlation matrix and the research direction matrix, similarity of each talent and other talents is sorted to obtain a sorting vector, multi-dimensional information fusion is performed on the sorting vector to obtain a sorting matrix, singular value decomposition is performed on the sorting matrix, the first column of the left singular vector is extracted, the first K talents with the highest similarity are selected as neighbors of the current talent, and a talent unified graph is constructed according to each talent and the corresponding neighbors; nodes of the talent unified graph represent talents, and edges represent the correlation between talents.

[0018] The talent unified graph is subjected to non-negative matrix factorization to obtain a corresponding talent membership indication matrix, the similarity of talents in team affiliation is calculated according to the talent membership indication matrix, the difference between the talent unified graph and the similarity of talents in team affiliation is minimized, a talent team division result is obtained, and a history evolution constraint is applied to the talent team division result to construct an objective function.

[0019] The objective function at each moment is optimized and solved to obtain a corresponding optimal talent membership indication matrix, and dynamic talent team division is performed according to the optimal talent membership indication matrix.

[0020] A computer readable storage medium has a computer program stored thereon, and the computer program is executed by a processor to implement the following steps:

[0021] A talent correlation matrix is constructed according to the academic correlation between talents, and a research direction matrix is constructed according to the research direction of talents.

[0022] According to the talent correlation matrix and the research direction matrix, similarity of each talent and other talents is sorted to obtain a sorting vector, multi-dimensional information fusion is performed on the sorting vector to obtain a sorting matrix, singular value decomposition is performed on the sorting matrix, the first column of the left singular vector is extracted, the first K talents with the highest similarity are selected as neighbors of the current talent, and a talent unified graph is constructed according to each talent and the corresponding neighbors; nodes of the talent unified graph represent talents, and edges represent the correlation between talents.

[0023] The talent unified graph is subjected to non-negative matrix factorization to obtain a corresponding talent membership indication matrix, the similarity of talents in team affiliation is calculated according to the talent membership indication matrix, the difference between the talent unified graph and the similarity of talents in team affiliation is minimized, a talent team division result is obtained, and a history evolution constraint is applied to the talent team division result to construct an objective function.

[0024] The objective function at each moment is optimized and solved to obtain a corresponding optimal talent membership indication matrix, and dynamic talent team division is performed according to the optimal talent membership indication matrix.

[0025] The aforementioned method, apparatus, and equipment for dynamic talent team discovery based on complex network analysis effectively integrate multi-dimensional data by constructing a unified talent graph containing talent association information and research direction information. This improves the analytical accuracy of talent relationships and collaboration potential. The unified talent graph undergoes non-negative matrix decomposition to obtain a corresponding talent affiliation indicator matrix. Based on this matrix, the similarity of talents in team affiliation is calculated. The difference between the unified talent graph and the similarity of talents in team affiliation is minimized to obtain the talent team division results. Historical evolution constraints are applied to these results, and an objective function is constructed. This achieves smooth evolution of team division in dynamic environments, effectively supporting adjustments to team structure at different points in time and ensuring teams can adapt to dynamic changes in talent needs. This invention avoids subjective judgment bias through data-driven quantitative analysis, ensuring more scientific and accurate team division. Embodiments of this invention can adjust team division in real time, optimize talent resource allocation, and improve team collaboration efficiency and innovation capabilities. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a dynamic talent team discovery method based on complex network analysis in one embodiment.

[0027] Figure 2 This is a schematic diagram illustrating the experimental results of the algorithm on a real network dataset Q in one embodiment.

[0028] Figure 3 This is a structural block diagram of a dynamic talent team discovery device based on complex network analysis in one embodiment;

[0029] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0031] In one embodiment, such as Figure 1 As shown, a dynamic talent team discovery method based on complex network analysis is provided, including the following steps:

[0032] Step 102: Construct a talent association matrix based on the academic connections between talents, and construct a research direction matrix based on the research directions of talents.

[0033] Construct a talent association matrix A∈R based on the pre-built talent pool. n×n Where n is the number of talents, and if there is an academic relationship between talents i and j, then Aij = 1, otherwise A ij = 0. Meanwhile, construct the talent research direction attribute matrix T ∈ R n×m , where m is the number of research directions, and T ij = 1 if talent i's research direction is j, otherwise T ij = 0. If there is an academic association between two talents, such as academic citation, it is considered that the research directions of the two talents are similar. The talent pre-set library refers to a pre-established database or system containing talent information, which is used to store and manage various information related to talents, including academic background, research direction, skill, work experience, project participation, etc.

[0034] Step 104: According to the talent association matrix and the research direction matrix, the similarity of each talent to other talents is sorted to obtain a sorting vector. The sorting vector is subjected to multi-dimensional information fusion to obtain a sorting matrix. The sorting matrix is subjected to singular value decomposition, the first column of the left singular vector is extracted, the first K talents with the highest similarity are selected as the neighbors of the current talent, and a talent unified graph is constructed according to each talent and the corresponding neighbor.

[0035] Dynamic talent team discovery involves multi-dimensional heterogeneous data, such as academic association and research direction of talents, and these data may change over time. These characteristics make it difficult for simple linear models or traditional data mining techniques to effectively solve this problem. The present invention is based on complex network analysis method, which realizes the dynamic talent team discovery by constructing a multi-dimensional network model. In the present invention, complex network analysis refers to modeling and analyzing the multiple relationships between talents through graph theory, in order to identify potential collaboration patterns, optimize team structure and realize dynamic adjustment.

[0036] By fusing the talent association matrix and the research direction matrix, a unified graph model is constructed, the input of which includes a talent set {u1, u2,..., u n}, a talent relationship view and a research direction view, each view representing part or all of n talents, the nodes of the talent unified graph representing talents, and the edges representing the association between talents. For data containing w dimensions, the nodes are denoted as U = {u1, u2,..., u n}, and each dimension of data contains all or part of the nodes. The information fusion steps are as follows:

[0037] Step 1: For each dimension of data j (1≤j≤w), calculate the similarity vector p i of u ij and other nodes in this dimension. Specifically, for each talent u i and each view j, use the similarity measure provided by the view to calculate the talent u isimilarity vector between the view and all other people in the view.

[0038] Step 2: According to the value of p ij , generate a ranking vector containing the other n-1 nodes, denoted as r ij . If all nodes are not included in dimension j, the non-existent node is assigned a sequence n' j +1, n' j is the number of nodes in this dimension.

[0039] Step 3: Arrange the w ranking vectors into a (n-1) x w ranking matrix R i , and normalize each row. Specifically, stack the ranking vectors of the two views as columns to form a ranking matrix, and normalize the columns of the ranking matrix to unit length, where w = 2.

[0040] Step 4: Calculate the SVD (singular value decomposition) of , extract the first column of the left singular vector. Arrange the vector in descending order, and select the top k people as the neighbor nodes of u i .

[0041] Step 5: When the k nearest neighbor nodes of n users are ranked, construct a directed and unweighted graph, where each node in the graph represents a talent, and the edge starts from an empty set. When u j is a neighbor of u i , add an edge from node i to node j in the graph, thereby generating a unified talent graph G that integrates talent relationships and research directions.

[0042] By constructing a unified talent graph G that contains talent association information and research direction information, the relationship between talents and the possibility of collaboration can be effectively represented, and the deep mining and analysis of multi-dimensional dynamic data and relationships can be realized. The unified graph integrates information from two dimensions of academic association and research field, reflecting the collaborative potential and cooperation path between talents. Through complex network analysis methods, the most potential talent team can be dynamically identified and discovered, supporting efficient discovery and optimal allocation of talents. In addition, as time goes by, the relationships and collaboration patterns in the graph will evolve, and the method of the present application can adjust the talent team division in real time in a changing environment, thereby providing accurate decision support.

[0043] Step 106, non-negative matrix factorization is performed on the talent unified graph to obtain a talent affiliation indication matrix, the similarity degree of talents in team affiliation is calculated according to the talent affiliation indication matrix, the difference between the talent unified graph and the similarity degree of talents in team affiliation is minimized, and the talent team division result is obtained. The talent team division result is subjected to historical evolution constraint to construct an objective function.

[0044] After obtaining the talent unified graph G that integrates the two types of information at each time step through step 104, it is expected that accurate talent team division can be obtained at each time step, and it is also expected that the talent team division will not change dramatically in a short period of time. First, the non-negative matrix factorization framework is applied to decompose the graph G at each time step into a talent affiliation indication matrix H t t n×k , where H ij,t represents the probability of talent i belonging to team j at time t, and the current talent team division result can be obtained:

[0045]

[0046] For the dynamic evolution of talent association and research direction information over time, the talent team division at consecutive time steps is smoothly evolved, and the accuracy of team discovery is improved by minimizing the consecutive talent team division results.

[0047]

[0048] In formula (2), only the case of the same number of teams can be handled, because the dimensions of the H matrix are the same. In order to better handle the case of team changes, the team evolution constraint at consecutive time steps is written as:

[0049]

[0050] Finally, the talent division at each time step and the constraint describing the talent team at consecutive time steps are combined to obtain the final objective function of the method:

[0051]

[0052] where α is a weight coefficient, used to determine the weight proportion of the current talent division quality and historical evolution.

[0053] In dynamic talent team discovery, the core goal is to dynamically divide talent teams according to the graph structure and ensure that the team division remains stable and smoothly evolves over time. By introducing a smoothing constraint, it is ensured that the team division is stable over time, avoiding dramatic fluctuations in team structure. Thus, the talent team affiliation is dynamically adjusted to adapt to the changing research directions and task requirements.

[0054] Step 108, optimize and solve the objective function at each time step to obtain the optimal talent affiliation indication matrix, and perform dynamic talent team division based on the optimal talent affiliation indication matrix.

[0055] The final objective function (5) is a non-convex function, so by fixing other variables, the optimization objective for a variable is obtained, and the update formula for the matrix H t is obtained:​​

[0056]

[0057] By iteratively optimizing the solution, the talent affiliation indicator matrix H at each time step can be obtained. t via H ic,t Assign talent i to team c corresponding to the maximum membership degree, and thus obtain the talent team assignment at each time step:

[0058]

[0059] It is understood that the method of this invention addresses the dynamic characteristics of talent development and mobility by utilizing artificial intelligence to propose a real-time dynamic talent team discovery method based on complex network analysis. This method fully leverages historical evolutionary information and, based on a non-negative matrix factorization framework, obtains accurate talent team discovery results across consecutive timeframes. This not only improves the accuracy and real-time performance of talent team segmentation but also provides scientific and reliable support for talent management decisions.

[0060] The aforementioned dynamic talent team discovery method based on complex network analysis constructs a unified talent graph containing talent association information and research direction information. This effectively integrates multi-dimensional data, improving the accuracy of talent relationship and collaboration potential analysis. Non-negative matrix decomposition is performed on the unified talent graph to obtain the corresponding talent affiliation indicator matrix. Based on this matrix, the similarity of talents in team affiliation is calculated. The difference between the unified talent graph and the similarity of talents in team affiliation is minimized to obtain the talent team division result. Historical evolution constraints are applied to the talent team division result, and an objective function is constructed. This achieves smooth evolution of team division in dynamic environments, effectively supporting adjustments to team structure at different points in time and ensuring that teams can adapt to dynamic changes in talent needs. This invention avoids the bias of subjective judgment through data-driven quantitative analysis, ensuring more scientific and accurate team division. The embodiments of this invention can adjust team division in real time, optimize talent resource allocation, and improve team collaboration efficiency and innovation capabilities.

[0061] In one embodiment, minimizing the difference between the talent unification map and the similarity of talents in team affiliation yields the following talent team segmentation results: Minimizing the difference between the talent unification map and the similarity of talents in team affiliation yields the following talent team segmentation results:

[0062]

[0063] Among them, G t H is the talent unified graph at time t. t Let H be the talent membership indicator matrix at time t. ij,t Let i be the probability that talent i belongs to team j at time t. is the square of the F-norm of a matrix, (·) T is the transpose of a matrix, n is the number of talents, and k is the number of talent teams.

[0064] In an embodiment, the historical evolution constraint is:

[0065]

[0066] wherein L2 is the historical evolution constraint, H is a talent affiliation indication matrix, t is a time, and (·) T is the transpose of a matrix, is the square of the F-norm of a matrix.

[0067] In an embodiment, the objective function is:

[0068]

[0069] wherein G t is a talent unified graph at t, H t is a talent affiliation indication matrix at t, H ij,t is the probability that talent i is affiliated to team j at t, is the square of the F-norm of a matrix, (·) T is the transpose of a matrix, n is the number of talents, k is the number of talent teams, and a is a weight coefficient.

[0070] In an embodiment, the objective function at each time is optimized to obtain the corresponding optimal talent affiliation indication matrix, including: the objective function at each time is optimized to obtain the corresponding optimal talent affiliation indication matrix according to a pre-set talent affiliation indication matrix updating rule.

[0071] In an embodiment, the dynamic talent team division is performed according to the optimal talent affiliation indication matrix, including: the probability that talents are affiliated to different teams is obtained according to the optimal talent affiliation indication matrix, talents are divided into the team corresponding to the maximum probability value to realize the dynamic talent team division.

[0072] In one embodiment, a series of experiments were conducted on two datasets to evaluate the effectiveness of the proposed method, DTTD (Dynamic Talent Team Discovery). DTTD was compared with FacetNet and SCI. FacetNet is a dynamic community detection method and SCI is a static community detection method. First, a dynamic talent dataset was constructed, which contains researchers from a university and covers multiple interdisciplinary research directions. Since there are similar research areas among disciplines, there can be academic connections among researchers. A talent association network was constructed based on the papers published by each researcher. If two researchers co-occur in the same paper, an edge is added between them. In addition, a research direction attribute network was constructed based on the research direction of each researcher. A time slice was constructed every year, so the entire dataset contains multiple time slices. As new researchers join and some researchers leave, the number of nodes and research direction attributes of the network change over time. As shown in the algorithm in Figure 2 The experimental results of the algorithm on the real network dataset Q are shown in the figure. The experimental results on the real network show that DTTD achieves stable and excellent community talent division performance. The Q value obtained by the proposed method remains at a high level in all steps. This proves that the proposed method can consistently group talents with similar research directions and closely related academic interests into the same team at all times.

[0073] In addition, the paper co-author network DBLP was extracted from dblp.org. This network contains information about authors who have published at least two papers in different fields in multiple years. The main field of the author's published papers is used as the research direction attribute of the node. As can be seen from Table 1, the proposed method achieved the best performance at 7 time points. This is mainly because the proposed method takes into account both the academic association between authors and the research direction, and also makes full use of historical information, so it obtains real-time accurate division.

[0074] Table 1 Experimental results of the algorithm on the DBLP co-author dataset Q

[0075]

[0076] It should be understood that, although Figure 1 the steps in the flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1At least one of the steps in the method can comprise a plurality of sub-steps or a plurality of stages, which sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the order of the sub-steps or stages is not necessarily sequential, but can be performed alternately or in rotation with other steps or sub-steps or stages of other steps.

[0077] In one embodiment, as shown in FIG. 1, a dynamic talent team discovery device based on complex network analysis is provided, comprising: Figure 3

[0078] The parameter acquisition module 302 is configured to construct a talent association matrix according to the academic association relationship between talents, and construct a research direction matrix according to the research direction of the talents.

[0079] The graph construction module 304 is configured to perform similarity sorting on each talent and other talents according to the talent association matrix and the research direction matrix to obtain a sorting vector, perform multi-dimensional information fusion on the sorting vector to obtain a sorting matrix, perform singular value decomposition on the sorting matrix, extract the first column of the left singular vector, select the top K talents with the highest similarity as neighbors of the current talent, and construct a talent unified graph according to each talent and the corresponding neighbors. The nodes of the talent unified graph represent talents, and the edges represent the association relationship between the talents.

[0080] The model construction module 306 is configured to perform non-negative matrix decomposition on the talent unified graph to obtain a talent membership indication matrix, calculate the similarity of the talents in team affiliation according to the talent membership indication matrix, minimize the difference between the talent unified graph and the similarity of the talents in team affiliation, obtain a talent team division result, and impose a historical evolution constraint on the talent team division result to construct an objective function. The talent membership indication matrix represents the probability of each talent belonging to different teams.

[0081] The model solving module 308 is configured to optimize and solve the objective function at each time to obtain a corresponding optimal talent membership indication matrix, and perform dynamic talent team division according to the optimal talent membership indication matrix.

[0082] The specific limitations of the dynamic talent team discovery device based on complex network analysis can be referred to the limitations of the dynamic talent team discovery method based on complex network analysis in the foregoing, which will not be repeated here. The various modules in the above dynamic talent team discovery device based on complex network analysis can be realized by software, hardware and combinations thereof, in whole or in part. The above various modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above various modules.

[0083] ​In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a dynamic talent team discovery method based on complex network analysis. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0084] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0085] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0086] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0087] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, databases, or other media in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0088] The technical features of the above embodiments can be combined in any way. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist, they should be considered as the scope of the present application.

[0089] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A dynamic talent team discovery method based on complex network analysis, characterized in that, The method comprises: According to the academic relationship between talents, a talent correlation matrix is constructed, and according to the research direction of talents, a research direction matrix is constructed; According to the talent correlation matrix and the research direction matrix, the similarity of each talent and other talents is sorted to obtain a sorting vector, multi-dimensional information fusion is performed on the sorting vector to obtain a sorting matrix, singular value decomposition is performed on the sorting matrix, the first column of the left singular vector is extracted, the top K talents with the highest similarity are selected as the neighbors of the current talent, and a talent unified graph is constructed according to each talent and the corresponding neighbor; The nodes of the talent unified graph represent talents, and the edges represent the correlation between talents. Non-negative matrix decomposition is performed on the talent unified graph to obtain a corresponding talent membership indication matrix, the similarity of talents in team affiliation is calculated according to the talent membership indication matrix, the difference between the talent unified graph and the similarity of talents in team affiliation is minimized, a talent team division result is obtained, and a historical evolution constraint is applied to the talent team division result to construct an objective function; The objective function at each moment is optimized and solved to obtain a corresponding optimal talent membership indication matrix, and dynamic talent team division is performed according to the optimal talent membership indication matrix; The objective function is: ； wherein, is the unified matrix of talents at the moment, is the membership indication matrix of talents at the moment, is the number of talents i at t the moment, j the probability of belonging to the team at the moment, is the square of the F-norm of the matrix, is the transpose of the matrix, is the number of talents, is the number of talent teams, is the weight coefficient.

2. The method of claim 1, wherein, Minimizing the difference between the talent unified graph and the similarity of talents in team affiliation to obtain a talent team division result comprises: Minimizing the difference between the talent unified graph and the similarity of talents in team affiliation to obtain a talent team division result is: ； wherein, is the unified graph of talents at time t, is the membership indication matrix of talents at time t, is the number of talents i at t time t belonging to team j , is the square of the F-norm of the matrix, is the transpose of the matrix, n is the number of talents, k is the number of teams of talents.

3. The method of claim 1, wherein, The historical evolution constraint is: ； wherein, is a historical evolution constraint, is a talent affiliation indication matrix, is a time, is a transpose of a matrix, is a square of the F-norm of a matrix.

4. The method of claim 1, wherein, Optimizing and solving the objective function at each moment to obtain a corresponding optimal talent membership indication matrix comprises: According to the pre-set talent membership indication matrix update rule, the objective function at each moment is optimized and solved to obtain a corresponding optimal talent membership indication matrix.

5. The method of claim 1, wherein, Dynamic talent team division according to the optimal talent membership indication matrix comprises: According to the optimal talent membership indication matrix, the probability that talents belong to different teams is obtained, talents are divided into the team corresponding to the maximum probability value, and dynamic talent team division is realized. 6.A dynamic talent team discovery device based on complex network analysis, characterized in that, The device comprises: A parameter acquisition module is configured to construct a talent correlation matrix according to the academic relationship between talents, and construct a research direction matrix according to the research direction of talents; A graph construction module is configured to sort the similarity of each talent and other talents according to the talent correlation matrix and the research direction matrix to obtain a sorting vector, perform multi-dimensional information fusion on the sorting vector to obtain a sorting matrix, perform singular value decomposition on the sorting matrix, extract the first column of the left singular vector, select the top K talents with the highest similarity as the neighbors of the current talent, and construct a talent unified graph according to each talent and the corresponding neighbor; The nodes of the talent unified graph represent talents, and the edges represent the correlation between talents. A model construction module is configured to perform non-negative matrix factorization on the talent unified graph to obtain a talent membership indication matrix, calculate a similarity degree of talent team membership according to the talent membership indication matrix, minimize a difference between the talent unified graph and the similarity degree of talent team membership, obtain a talent team division result, and impose a historical evolution constraint on the talent team division result to construct an objective function. A model solution module is configured to optimize and solve the objective function at each time to obtain an optimal talent membership indication matrix, and perform dynamic talent team division according to the optimal talent membership indication matrix. The objective function is as follows: ； in, for A unified talent map at all times. for Talent affiliation indicator matrix at any given time For talent i exist t Always belong to the team j The probability, The square of the F-norm of the matrix. This is the transpose of the matrix. For the number of talents, For the number of talent teams, These are the weighting coefficients. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. The processor implements the steps of the method in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method in any one of claims 1 to 5.

Images