Academic cooperation network construction method based on big data

Abstract

The invention relates to the field of academic cooperation network construction, in particular to an academic cooperation network construction method based on big data, which comprises the following steps: constructing an initial academic cooperation network of a Neo4j graph database architecture, taking scholars as nodes and taking literature signature relationships among the scholars as edges; calculating a weighted center index of the academic cooperation network based on the author signature sequence of the literature; adjusting a weighted center index based on the document publishing time point; calculating a contribution degree deviation item based on the author contribution degree weight and the contribution degree statistical value of the field to which the literature belongs; and calculating a comprehensive center index based on the contribution degree deviation term, the domain adjustment weight and the time sequence center index, and carrying out iterative construction on the initial academic cooperation network. According to the method, through contribution degree quantification, time sequence dynamic adjustment and multi-dimensional index integration, the calculation efficiency of the Neo4j graph database and the fluency of the iteration construction process are ensured.

Description

Technical Field

[0001] This invention relates to the field of academic collaboration network construction, and in particular to a method for constructing academic collaboration networks based on big data. Background Technology

[0002] With the advent of globalization and the information age, academic collaboration has become a vital force driving scientific research and knowledge innovation. Collaborative relationships among scholars are no longer confined to traditional geographical and disciplinary boundaries, but rather exhibit a diversified and dynamic trend. Based on this, academic collaboration networks, used to reveal the structure of research teams, identify core scholars, and predict research trends, have become important tools for providing precise analysis in research management, collaboration recommendations, and academic evaluation.

[0003] Academic collaboration networks are often constructed using graph databases. A scholar collaboration network can be viewed as a complex network composed of scholar nodes and collaboration relationship edges. In this network, scholars, as independent individuals, form nodes in the network. Scholars establish connections through collaborative research, joint publication of papers, and other means. These connections form edges that connect nodes in the network, and the weight of the edges reflects the degree of academic exchange and interaction between scholars.

[0004] However, the traditional process of constructing scholar collaboration networks often overlooks the unevenness of contributions among scholars in the publication process and the dynamic nature of scholarly collaborations. On the one hand, collaborations among scholars are not equal; in reality, this depends on factors such as the frequency of collaborations and the magnitude of contributions. On the other hand, scholarly collaborations change over time, with new collaborations constantly emerging and older ones potentially weakening. Therefore, the edge weights in traditional graph databases cannot accurately reflect the true relationships between scholars.

[0005] Furthermore, with the exponential growth of academic data, traditional static network models are struggling to meet the demands of real-time analysis in terms of dynamism, scalability, and computational efficiency, necessitating efficient graph database construction methods for graph computation.

[0006] Currently, the existing Neo4j architecture graph database supports dynamic addition and deletion of edges and real-time updates of node attributes. However, with large amounts of data related to academic collaborations, its transaction processing and indexing mechanisms cannot adapt to the dynamic evolution of academic collaboration networks, thus affecting data processing efficiency during iteration.

[0007] Therefore, how to solve the performance bottleneck of large-scale batch update latency in databases under the dynamic evolution scenario of academic collaboration networks is a technical problem that needs to be solved. Summary of the Invention

[0008] To address this, the present invention provides a method for constructing academic collaboration networks based on big data. By calculating multi-dimensional indicators of the Neo4j graph database based on author contribution weights according to authorship order and publication time, and iteratively constructing the Neo4j graph database based on these multi-dimensional indicators, the method enables path traversal and calculation directly based on edge weight attributes using Neo4j's built-in graph algorithm engine. This avoids the performance overhead of nested queries across multiple tables in relational databases and solves the technical problem of delayed batch updates of large datasets with tens of thousands of scholar nodes in dynamically evolving academic collaboration networks. This ensures the computational efficiency and smoothness of the iterative construction process of the Neo4j graph database.

[0009] To achieve the above objectives, this invention proposes a method for constructing an academic collaboration network based on big data, comprising: Construct an initial academic collaboration network with scholars as nodes and authorship relationships between scholars as edges, based on the Neo4j graph database architecture; The author contribution weights of the edges of the initial academic collaboration network are calculated based on the author authorship order of the literature, and the weighted centrality index of the initial academic collaboration network is calculated based on the author contribution weights. The weighted centrality index is adjusted based on the publication time of the literature to generate a time-series centrality index; The weighted citation count is calculated based on the author contribution weight, and the contribution deviation is calculated based on the statistical value of the contribution weight of all authors in the field to which the document belongs. The field adjustment weight is calculated based on the number of scholars in the field to which the document belongs. A comprehensive center index is calculated based on the contribution deviation term, the domain adjustment weight, and the time-series center index. The initial academic collaboration network is then iteratively constructed based on the comprehensive center index to generate the academic collaboration network.

[0010] Furthermore, the process of calculating the author contribution weights includes: If the scholar is not the corresponding author of the document, the first contribution weight is generated based on the ratio of the number of authors of non-corresponding authors to the sum of the number of authors of the document. If a scholar is the corresponding author of a document, the ratio of the first-ranked non-corresponding author's authorship ranking to the sum of the ranking values ​​of the authors is calculated to generate the second contribution weight. The author contribution weight is generated by summing the first or second contribution weight of the first scholar and the first or second contribution weight of the second scholar who authored the same document, and then dividing by the total number of authors.

[0011] Furthermore, the process of calculating the weighted centrality index of the initial academic collaboration network includes: A weighted degree centrality index is generated based on the sum of the author contribution weights of all edges connecting the nodes in the initial academic collaboration network. Based on the author contribution weight of the shortest path connecting two nodes in the initial academic collaboration network, a weighted intermediary center index is generated. Based on the node feature vectors of the initial academic collaboration network and the author contribution weights, a weighted feature vector center index is generated. The weighted center index is generated by weighting the weighted degree center index, the weighted intermediary center index, and the weighted feature vector center index.

[0012] Furthermore, the process of generating the weighted intermediary center index includes: Obtain the first total value of the author contribution weight of the shortest path from the first node to the second node and through the third node in the initial academic collaboration network; Obtain the second total value of the author contribution weight of the shortest path from the first node to the second node of the initial academic collaboration network; A weighted intermediary center index is generated based on the ratio of the first total value to the second total value.

[0013] Furthermore, the process of generating time-series centrality indicators includes: An exponential calculation based on the difference between the current time and the publication time of the document is used to generate a publication attenuation term; An exponential calculation based on the current time and the earliest publication time of the literature stored in the initial academic collaboration network is used to generate an overall decay term; The normalized calculation of the literature publication attenuation term based on the overall attenuation term is used to generate time-series weights; The time-series centrality index is generated by weighting the weighted centrality index and the time-series weights based on multiple time slices stored in the initial academic collaboration network.

[0014] Furthermore, the statistical values ​​include the median and standard deviation, and the process of calculating the contribution deviation term includes: The contribution deviation term is generated by subtracting the median of the contribution weights of all authors in the field of the document from the weighted citation count and dividing by the standard deviation of the contribution weights of all authors in the field of the document.

[0015] Furthermore, the process of adjusting weights in the computing domain includes: Calculate a first logarithm of the number of scholars in the field to which the literature belongs, a second logarithm of the number of scholars in the largest field, and calculate the field-adjusted weights based on the ratio of the first logarithm to the second logarithm.

[0016] Furthermore, the process of calculating the comprehensive center index includes: The domain time-series center index is generated by multiplying the weighted center index and the time-series center index. The comprehensive center index is generated by weighting the contribution deviation term and the domain time series center index.

[0017] Furthermore, the process of iteratively constructing the initial academic collaboration network based on comprehensive center indicators includes: The values ​​from the previous cycle are adjusted based on the comprehensive center index to generate the values ​​for the current cycle. The initial academic collaboration network is iteratively constructed based on the current cycle iteration value, and the iterative construction of the initial academic collaboration network is determined to stop based on the comparison between the difference between the current iteration value and the next cycle iteration value and a set threshold.

[0018] Furthermore, the process of iteratively constructing the initial academic collaboration network based on the current periodic iteration values ​​includes: The node attraction coefficient is calculated based on the current periodic iteration value of the first node and the current periodic iteration value of the second node in the initial academic cooperation network. The node attraction coefficients are passed through a Sigmoid function to generate edge generation probabilities; The initial academic collaboration network is iteratively constructed based on the edge generation probability.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention calculates the author contribution weight based on the order of authorship, introduces the mechanism of association of scholar contribution differences based on authorship order into network modeling, introduces the weighting mechanism of centrality indicators based on the publication time of the literature to generate time-series centrality indicators for the Neo4j graph database, further adjusts the centrality indicators based on contribution deviation terms and time decay terms, and iteratively constructs the academic collaboration network based on the centrality indicators. This makes the calculation of centrality indicators unnecessary for additional data format conversion, directly using the author contribution weights of the edges as path weights to quickly solve the weighted shortest path between nodes. It achieves a significant improvement in the calculation efficiency of complex centrality indicators under large data volumes through contribution quantification, dynamic time-series adjustment, and multi-dimensional indicator integration.

[0020] In particular, this invention utilizes the weighted centrality index, weighted intermediary centrality index, and weighted eigenvector centrality index of the Neo4j graph database architecture to achieve a comprehensive characterization of scholars' network roles. The weighted centrality index considers the breadth of collaboration while highlighting the value of high-contribution collaborations, avoiding the bias of quantity masking quality. The weighted intermediary centrality index quantifies the bridging role of nodes through the path contribution weight ratio, accurately identifying key intermediary scholars in cross-disciplinary collaborations. The weighted eigenvector centrality index adjusts each centrality index using authorship order contribution weights, making the weight assignment of network edges more closely reflect the actual contribution logic of academic collaborations, providing a more scientific quantitative tool for evaluating scholar influence and mining collaboration patterns. Furthermore, through multi-dimensional time modeling of time-series centrality indicators, the dynamic analysis capability of the academic collaboration network of the Neo4j graph database architecture is improved, providing a more scientifically sound quantitative tool for research evaluation, collaboration recommendation, and policy formulation.

[0021] In particular, this invention improves the fairness of cross-domain evaluation by statistical standardization of contribution deviation terms and optimizes the comprehensiveness of the comprehensive central indicators through multi-level integration. That is, it achieves three-dimensional coupling of static contribution, dynamic timeliness and domain benchmark by combining domain time-series weighted centrality with the weighted fusion of contribution deviation terms. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the method for constructing an academic collaboration network based on big data, according to an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the calculation process of the weighted centrality index in the big data-based academic collaboration network construction method of this invention. Figure 3 This is a schematic diagram illustrating the calculation process of the domain time-series centrality index in the big data-based academic collaboration network construction method of this invention. Figure 4 This is a schematic diagram illustrating the academic collaboration network construction effect of the big data-based academic collaboration network construction method according to an embodiment of the present invention. Figure 5 This invention relates to a method for constructing an academic collaboration network based on big data. Figure 4 A schematic diagram of the partial construction effect of part A in the middle. Detailed Implementation

[0023] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0024] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0025] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0026] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] like Figures 1 to 5 As shown, this invention provides a method for constructing academic collaboration networks based on big data. By calculating multi-dimensional indicators of the Neo4j graph database based on author contribution weights according to authorship order and publication time, and iteratively constructing the Neo4j graph database based on these multi-dimensional indicators, it enables path traversal and calculation directly based on edge weight attributes through Neo4j's built-in graph algorithm engine. This avoids the performance loss of nested queries in relational databases and solves the technical problem of delayed batch updates of large amounts of data with tens of thousands of scholar nodes in the dynamic evolution scenario of academic collaboration networks. This ensures the computational efficiency and smoothness of the iterative construction process of the Neo4j graph database.

[0028] like Figure 1 As shown, this embodiment proposes a method for constructing an academic collaboration network based on big data, including: Construct an initial academic collaboration network with scholars as nodes and authorship relationships between scholars as edges, based on the Neo4j graph database architecture; The author contribution weights of the edges of the initial academic collaboration network are calculated based on the author authorship order of the literature, and the weighted centrality index of the initial academic collaboration network is calculated based on the author contribution weights. The weighted centrality index is adjusted based on the publication time of the literature to generate a time-series centrality index; The weighted citation count is calculated based on the author contribution weight, and the contribution deviation is calculated based on the statistical value of the contribution weight of all authors in the field to which the document belongs. The field adjustment weight is calculated based on the number of scholars in the field to which the document belongs. A comprehensive center index is calculated based on the contribution deviation term, the domain adjustment weight, and the time-series center index. The initial academic collaboration network is then iteratively constructed based on the comprehensive center index to generate the academic collaboration network.

[0029] Understandably, the calculation of comprehensive centrality indicators only requires the graph algorithm engine built into the Neo4j graph database architecture, such as the shortest path algorithm and the PageRank algorithm. It can directly perform path traversal and calculation based on the weight attributes of the edges without additional data format conversion. This avoids the performance loss of nested queries in multiple tables in relational databases, thereby significantly improving the calculation efficiency of complex centrality indicators under large data volumes of tens of thousands of scholar nodes and tens of thousands of collaborative edges, ensuring the computational efficiency of the Neo4j graph database and the smoothness of the iterative construction process.

[0030] It should be noted that, in this embodiment, the academic collaboration network and initial academic collaboration network are preferably academic collaboration networks in the field of traditional Chinese medicine, the literature is preferably academic papers, and the author contribution weights are preferably stored as an adjacency matrix in the Neo4j graph database architecture. The construction effect of the academic collaboration network and initial academic collaboration network is as follows: Figure 4 and 5 As shown.

[0031] In this embodiment, as Figure 2 As shown, the process of calculating the author contribution weight includes: If the scholar is not the corresponding author of the document, the first contribution weight is generated based on the ratio of the number of authors of non-corresponding authors to the sum of the number of authors of the document. If a scholar is the corresponding author of a document, the ratio of the first-ranked non-corresponding author's authorship ranking to the sum of the ranking values ​​of the authors is calculated to generate the second contribution weight. The author contribution weight is generated by summing the first or second contribution weight of the first scholar and the first or second contribution weight of the second scholar who authored the same document, and then dividing by the total number of authors.

[0032] Specifically, the authors of the literature include both non-corresponding authors and corresponding authors.

[0033] Specifically, when the scholar is not the corresponding author of the document, the calculation process of the first contribution weight can be expressed as follows: ; In the formula, Let x represent the weight of the first contribution of the j-th non-corresponding author in the overall authorship ranking of this document, where j represents the non-corresponding author's position in the overall authorship ranking, k represents the total number of authors in the document (preferably the total number of corresponding and non-corresponding authors, i.e., the total number of authors), and x represents the total number of corresponding authors. This indicates the sequential number of the non-corresponding author's authorship in the j-th non-corresponding author's publication.

[0034] in The order of priority for non-corresponding authors is as follows: At this point, the sum of all position values ​​for the corresponding author and non-corresponding authors, i.e., the total sum of the author order values, is: .

[0035] Therefore, for non-corresponding authors, their contribution weight is determined by the following steps: First, calculate the ratio of the non-corresponding author's ranking in the total number of author rankings (S); then, sort all the non-corresponding author ratios in ascending order. The non-corresponding author j's ranking in the total number of authors and its corresponding first contribution weight are as follows: ; The formula for calculating the first contribution weight of the non-corresponding author can then be derived.

[0036] Specifically, when a scholar is the corresponding author of a paper, the weight of the corresponding author's second contribution is set equal to the weight of the first contribution of the first-ranked non-corresponding author, i.e., [the weight is taken as...]. The order of authorship for each corresponding author (x) is uniformly set to [number]. The corresponding author, J, satisfies The process of calculating the weight of the second contribution can be expressed as follows: k represents the total number of authors listed in the paper, x represents the total number of corresponding authors, and the first-ranked non-corresponding author has the following authorship order: The sum of the ranking values ​​of the signed scholars is .

[0037] Specifically, the process of generating the author contribution weights can be represented as follows: ; In the formula, Weighting of author contribution These represent the first or second contribution weight of the first scholar i and the second scholar j, respectively, which are determined based on whether the first scholar i and the second scholar j are the corresponding author or not. The denominator can be the total number of authors k or the total number of authors minus the number of first authors.

[0038] Understandably, when a paper only has a first-ranked non-corresponding author and several corresponding authors, the calculation process for the first contribution weight / second contribution weight and the author contribution weight can be simplified as follows: ; In the formula, Weighting of author contribution represents the first contribution weight and / or second contribution weight of scholar i and scholar j who are authors of the same document, respectively, and k represents the total number of authors of the document.

[0039] In this embodiment, as Figure 2 As shown, the process of calculating the weighted centrality index of the initial academic collaboration network includes: A weighted degree centrality index is generated based on the sum of the author contribution weights of all edges connecting the nodes in the initial academic collaboration network. Based on the author contribution weight of the shortest path connecting two nodes in the initial academic collaboration network, a weighted intermediary center index is generated. Based on the node feature vectors of the initial academic collaboration network and the author contribution weights, a weighted feature vector center index is generated. The weighted center index is generated by weighting the weighted degree center index, the weighted intermediary center index, and the weighted feature vector center index.

[0040] In this embodiment, the process of generating the weighted intermediary center index includes: Obtain the first total value of the author contribution weight of the shortest path from the first node to the second node and through the third node in the initial academic collaboration network; Obtain the second total value of the author contribution weight of the shortest path from the first node to the second node of the initial academic collaboration network; A weighted intermediary center index is generated based on the ratio of the first total value to the second total value.

[0041] In this embodiment, as Figure 2 As shown, the process of generating weighted central indicators includes: The weighted feature vector center index is generated by multiplying the feature vector of a node by the contribution weights of all authors.

[0042] Specifically, the process of generating the weighted degree centrality index can be expressed as: ; In the formula, This represents the weighted degree centrality index of node i. The author contribution weight between node i and node t, i.e., the node in the initial academic collaboration network of the Neo4j graph database architecture. and nodes The weight of the edges between them. This represents the sum of the contribution weights of all authors connected to node i.

[0043] Understandably, in the Neo4j graph database architecture, edges have different weights, and the weight of an edge represents the importance of the connection. The greater the weight of an author's contribution, the greater the weight of the edge, and the stronger the collaboration between scholars.

[0044] Specifically, the process of generating the weighted intermediary center index can be represented as follows: ; In the formula, This represents the weighted degree centrality index of node i. This indicates traversing all possible pairs of nodes in the network. and exclude or In this situation, express ,Right now This represents the sum of author contribution weights for all shortest paths from the first node s to the second node t, where the shortest path is the path that traverses the fewest nodes or does not traverse any nodes. This represents the author contribution weight from the first node s to the second node t via the third node i, i. This represents the sum of the author contribution weights along the shortest path i.

[0045] Therefore, the weighted degree centrality index can reflect the relative control of the third node i from the first node s to the second node t. That is, the higher the weighted degree centrality index, the more important the role of scholars in the information dissemination process, the more information can be transmitted through fewer paths, and the greater the collaborative influence of scholars.

[0046] Specifically, the process of generating the weighted eigenvector center index can be represented as: ; In the formula, This represents the weighted degree centrality index of node i. The eigenvector centrality measure for node i. It is a proportionality constant, preferably 0.6. This indicates the weight of the author's contribution. As an adjacency matrix The elements in W are used to solve for the eigenvectors of node j. 'n' represents the total number of nodes connected to node 'i', which can be understood as the feature vector. The computation process follows the standard computation process of the weighted Neo4j graph database architecture.

[0047] Therefore, the weighted eigenvector centrality index, in the initial academic collaboration network of the weighted Neo4j graph database architecture, not only considers the collaboration relationships between scholars, but also further considers the contribution weight of scholars in the collaboration relationships. By introducing a weighting mechanism, we can evaluate the centrality of scholars based on their actual contributions in the collaboration.

[0048] Specifically, the weighted summation of the weighted degree center index, the weighted intermediate center index, and the weighted feature vector center index can be expressed as follows: The weighting coefficients Different values ​​are set according to the type of scholar.

[0049] As a specific implementation result analysis of this embodiment, data from 34 scholars were selected for analysis. The centrality index analysis results, which consider the collaboration relationships and number of collaborations among scholars and the scholar ranking weight (author contribution weight) of this embodiment, are as follows: Firstly, the calculation results of the weighted degree centrality index are shown in the table below, where... This indicates that the weighted centrality is not considered based on the order of authorship or the number of collaborations among scholars. This indicates the weighted degree centrality considering the number of times scholars collaborate. This embodiment describes the weighted degree centrality considering author ranking (author contribution weight): ; It is evident that the weighted centrality index evaluates a scholar's influence by calculating the number of scholars directly connected to that scholar, but it cannot be simply assumed that the more collaborators a scholar has, the greater his or her influence.

[0050] Analysis of the experimental results revealed that both the number of collaborations and the author's position on the author's name affected the degree centrality index, leading to an increase in the number of scholars... Compared to All of them fluctuate to varying degrees.

[0051] Among them, Scholar 9 experienced the largest change in ranking. Considering the number of collaborations, their ranking rose from 19th to 10th, an increase of 9 places. Considering the weight of scholar ranking (contribution weight), their ranking rose to 5th, an increase of 14 places. A deeper examination of Scholar 9's academic activities revealed that they published 7 articles that year and collaborated with some scholars not just once, but multiple times, with the maximum number of collaborations reaching 4. This may have led to... Compared to The ranking has risen. Of these seven articles, the scholar served as the corresponding author in six out of nine. The corresponding author is typically responsible for key aspects of the paper, including overall design, data integrity, and submission. Therefore, this role is often considered a key indicator of academic contribution and influence, which proves... Compared to That makes it more convincing.

[0052] Secondly, the calculation results of the weighted intermediary center index are shown in the table below, where... This indicates that the centrality of mediation is not considered based on the order of authorship or the number of collaborations among scholars. This indicates the centrality of the intermediary when considering the number of collaborations. This refers to the weighted intermediary center index, which takes into account author ranking (author contribution weight), as described in this embodiment. , ; Betweenness centrality measures a node's control over the entire network's information, indicating whether a node occupies a core position in information transmission. In this calculation process, as long as an author participates in collaboration, regardless of the depth of collaboration or the size of their contribution, the betweenness centrality is calculated solely based on the number of connections. This method ignores the difference between the depth of collaboration and the actual contributions of co-authors. Analysis of the experimental results reveals that the number of collaborations and authorship positions influence the calculation of the betweenness centrality index, leading to variations in the number of collaborations and authorship positions. Compared to All of them fluctuate to varying degrees.

[0053] It is evident that the largest change in ranking when considering the number of collaborations is seen in Scholar 18, whose ranking dropped from 15th to 30th, a decrease of 15 places. Conversely, the largest change in ranking when considering author order is seen in Scholar 32, whose ranking rose from 28th to 10th, an increase of 18 places. This is because, among the established collaborations, Scholar 18's collaborations with other scholars were mostly limited to one or two times, placing him on the periphery of the network. Therefore, the weight of his edge did not positively impact the calculation of the betweenness centrality index; his mediating role was replaced by nodes in the core region, and his connecting role was weakened due to insufficient weight. Scholar 32 published 10 papers that year, serving as the corresponding author for 6 and the second author for 2, demonstrating a significant contribution. The higher weight of his collaborative edges makes the path through him shorter in the weighted network, thus enhancing his betweenness centrality. The rise in rankings demonstrates the rationality of considering author contribution weights in calculating the betweenness centrality index, which more closely reflects the actual collaborative influence of scholars.

[0054] Third, the calculation results of the weighted eigenvector center index are shown in the table below, where... This indicates the eigenvector centrality without considering the order of authorship and the number of collaborations among scholars. This indicates the eigenvector centrality considering the number of collaborations. This indicates the eigenvector centrality considering author position weights (author contribution weights): ; In the table, E is a simplified scientific notation, i.e., a × 10ⁿ, abbreviated as aE ± n, which simplifies the writing and display of maxima and minima. This is because the calculation process of eigenvector centrality relies on the adjacency matrix, and a minimum value appears after matrix iteration and multiplication. However, weighted eigenvector centrality and weighted betweenness centrality do not undergo multiple iterations, so their values ​​do not become too small.

[0055] The weighted eigenvector centrality index posits that a node's importance depends not only on the node itself but also on its neighbors; that is, a node connected to an important node is more important than one connected to a common node. In the eigenvector centrality calculation, an author's influence is considered greater if they have established collaborations with multiple influential scholars. Analysis of the results reveals that, after introducing collaboration frequency and author contribution weights, the eigenvector centrality index shows a more significant change compared to degree centrality and betweenness centrality indices. This is primarily due to two reasons: first, degree centrality only considers the number of direct connections between nodes, betweenness centrality focuses on the shortest path position, while eigenvector centrality comprehensively considers both direct connections and the importance of neighbors; second, the eigenvector centrality calculation involves iterative multiplication of the adjacency matrices of neighboring nodes, which enhances its sensitivity to changes in network structure.

[0056] It is evident that the variation in scholar 21 is quite significant. It rose 17 places. It rose 14 places. By introducing the number of collaborations as the weight of the edges, Scholar 21 frequently established collaborations with other collaborators in that year. Some of the collaborations between its collaborators reached 13 times, which is higher than the network average, forming local "hot spots". At the same time, these hot spots were continuously strengthened during the iteration process, so Scholar 21's feature vector centrality ranking increased significantly. Considering the weight of author contributions, in the 21 papers published by Scholar 21 that year, more than half of them were collaborations with important scholars as the first author or corresponding author. These high-weight edges directly connect to the core nodes of the network. Influenced by the high-weight radiation of their collaborators, the eigenvector centrality ranking of Scholar 21 was also raised. By prioritizing connections to high-value neighbor nodes, the centrality index values ​​of feature vectors were aggregated.

[0057] In summary, it can be seen that for different types of scholars, it is necessary to focus on the weighted degree center index, the weighted intermediary center index, or the weighted eigenvector center index respectively.

[0058] In this embodiment, as Figure 3 As shown, the process of generating time series centrality indicators includes: An exponential calculation based on the difference between the current time and the publication time of the document is used to generate a publication attenuation term; An exponential calculation based on the current time and the earliest publication time of the literature stored in the initial academic collaboration network is used to generate an overall decay term; The normalized calculation of the literature publication attenuation term based on the overall attenuation term is used to generate time-series weights; The time-series centrality index is generated by weighting the weighted centrality index and the time-series weights based on multiple time slices stored in the initial academic collaboration network.

[0059] Specifically, the process of generating time-series weights can be represented as follows: ; In the formula, Represents time series weights. This indicates the attenuation term of publications. Represents the overall attenuation term, where It is the time decay factor, preferably 0.6. It is the current time point. It refers to the publication date of the literature. This refers to the initial time of the academic collaboration network, that is, the earliest publication time of the documents stored in the network. The value of t is... .

[0060] Specifically, the process of generating time series centrality indicators can be represented as follows: ; In the formula, Time slice The weighted centrality index, where w represents time-series weighting. Indicates to The summation of the weighted centrality index and the time series weight of all time slices t is taken as F, where F represents the time series centrality index.

[0061] Therefore, the time-series centrality index can more accurately quantify the timeliness of collaborative relationships and thereby assess the collaborative influence of scholars at different time periods.

[0062] Understandably, incorporating temporal relationships is crucial in constructing collaborative networks because it more accurately captures the dynamics and timeliness of these relationships. Collaborative relationships are not static but evolve over time; new collaborations may emerge, while older ones may gradually diminish or disappear. Therefore, to comprehensively assess the collaborative influence of scholars, this temporal evolution must be considered. Although past collaborations may have significantly impacted a scholar's academic development, in the rapidly changing academic field, recent collaborations often better reflect a scholar's current activity and influence. To more accurately reflect this trend, a time decay factor is introduced when calculating the centrality index, and a dynamically weighted centrality index is used to measure the collaborative influence of scholars.

[0063] In particular, by utilizing the Neo4j graph database architecture's weighted centrality index, weighted intermediary centrality index, and weighted eigenvector centrality index, a comprehensive characterization of scholar network roles is achieved. The weighted centrality index considers the breadth of collaboration while highlighting the value of high-contribution collaborations, avoiding the bias of quantity masking quality. The weighted intermediary centrality index quantifies the bridging role of nodes through path contribution weights, accurately identifying key intermediary scholars in cross-disciplinary collaborations. The weighted eigenvector centrality index weights each centrality index with authorship order contribution weights, making the weight assignments of network edges more closely reflect the actual contribution logic of academic collaborations, providing a more scientific quantitative tool for scholar influence assessment and collaboration pattern discovery. Furthermore, through multi-dimensional time modeling of time-series centrality indicators, the dynamic analysis capability of academic collaboration networks is further improved, providing a more scientifically sound quantitative tool for research evaluation, collaboration recommendation, and policy formulation.

[0064] As a specific implementation result analysis of this embodiment, data from 34 scholars were selected for analysis. The results of the time-weighted centrality index analysis, which considers the collaboration relationships and number of collaborations among scholars with the scholar ranking weight (author contribution weight) of this embodiment, are as follows: Firstly, the time-weighted centrality index, which focuses on the weighted degree centrality index, is shown in the table below, where... This indicates that the weighted centrality is not considered based on the order of authorship or the number of collaborations among scholars. This indicates the weighted degree centrality considering the number of times scholars collaborate. This embodiment describes the weighted degree centrality considering author ranking weights (author contribution weights). T1 to T5 represent the first to fifth academic cycles, preferably with a five-year academic cycle. Dynamic weighting is performed using this unit, which aligns with the time span of research projects, results generation, and dissemination in the academic field. F represents the time-series centrality index, and F-ranking represents the ranking result of the time-series centrality index. ; ; Secondly, the time-weighted centrality index, which focuses on the weighted intermediate center index, is shown in the table below, where... This indicates that the centrality of mediation is not considered based on the order of authorship or the number of collaborations among scholars. This indicates the centrality of the intermediary when considering the number of collaborations. This embodiment describes the weighted betweenness centrality index, which considers author ranking weights (author contribution weights). T1 to T5 represent the first to fifth academic cycles, preferably with a five-year academic cycle. Dynamic weighting is performed using this unit, which aligns with the time span of research projects, results generation, and dissemination in the academic field. F represents the time-series centrality index, and F-ranking represents the ranking result of the time-series centrality index. ; ; ; In the table, E is a simplified form of scientific notation, i.e., a × 10ⁿ, abbreviated as aE ± n. Its purpose is to simplify the writing and display of maximum or minimum values. Since the calculation of the time series centrality index relies on past values, a minimum value E will appear.

[0065] Third, the time-weighted centrality index of the weighted eigenvector center index is shown in the table below, where... This indicates the eigenvector centrality without considering the order of authorship and the number of collaborations among scholars. This indicates the eigenvector centrality considering the number of collaborations. This indicates the eigenvector centrality considering author position weights (author contribution weights). T1 to T5 represent the first to fifth academic cycles, with a preferred five-year cycle for dynamic weighting. This aligns with the time span of research projects, results generation, and dissemination in the academic field. F represents the time-series centrality index, and F-ranking indicates the ranking result of the time-series centrality index. ; ; ; In the table, E is a simplified form of scientific notation, i.e., a × 10ⁿ, abbreviated as aE ± n. Its purpose is to simplify the writing and display of maximum or minimum values. Since the calculation of the time series centrality index relies on past values, a minimum value E will appear.

[0066] Analysis of the experimental results reveals a significant change in the degree centrality index after incorporating dynamic time-series variations. Some scholars saw their rankings rise, while others experienced a decline. For example, in year T5, scholar 4's degree centrality index was 0.04390, ranking 22nd, while scholar 14's was 0.03158, ranking 29th. However, after considering dynamic time-series variations, a significant difference emerged in their dynamically weighted degree centrality indices: scholar 4's index dropped to 0.02142, and their ranking fell to 30th, a drop of 8 places; in contrast, scholar 14's index rose to 0.06812, and their ranking jumped to 18th, an increase of 11 places. This change is because, although scholar 14's degree centrality index was lower than scholar 4's in year T5, it was higher than scholar 4's in the preceding four years (T1 to T4). This long-term accumulated academic influence was fully considered in the dynamic weighted degree centrality calculation model, thus having a positive impact on the ranking of Scholar 14.

[0067] In the time-weighted centrality index, which emphasizes the weighted betweenness centrality, taking Scholar 21 as an example, although its betweenness centrality index in year T5 was only 0.00607, ranking 16th, its betweenness centrality index in previous years was excellent. Therefore, when calculating the dynamic weighted betweenness centrality, its index value increased to 0.01613, and its ranking rose to third place. This further confirms the necessity of dynamic time-series changes for evaluating scholar centrality; a similar phenomenon exists in the analysis of eigenvector centrality. In year T5, the eigenvector centrality of Scholar 24 was 1.1840E-05, higher than Scholar 28's 4.4620E-07. However, when examining the data from T1 to T4, we found that Scholar 28's eigenvector centrality was consistently higher than Scholar 24's. Therefore, when calculating the dynamic weighted eigenvector centrality, Scholar 28's index value surpassed that of Scholar 24.

[0068] In conclusion, calculating scholar centrality indicators by considering the temporal relationship not only focuses on a scholar's current influence but also comprehensively considers their historical academic influence, thus providing a more comprehensive and reasonable evaluation method. This evaluation method helps us to more accurately understand changes in a scholar's academic status and influence, providing strong support for the improvement of the academic evaluation system.

[0069] Therefore, this embodiment first introduces author contribution weights into the collaboration network, while also considering the changes in scholar collaboration relationships over time to construct a dynamic collaboration network. Then, it improves upon traditional centrality indices—degree centrality, betweenness centrality, and eigenvector centrality—by calculating a dynamic weighted centrality index. Experimental results show that the proposed method effectively identifies scholars who have made outstanding contributions to collaborations, focusing not only on their current collaborative influence but also on their historical academic contributions. Compared to traditional centrality index calculation methods, this approach is more reasonable and convincing.

[0070] In this embodiment, the statistical values ​​include the median and standard deviation, and the process of calculating the contribution deviation term includes: The contribution deviation term is generated by subtracting the median of the contribution weights of all authors in the field of the document from the weighted citation count and dividing by the standard deviation of the contribution weights of all authors in the field of the document.

[0071] Specifically, the process of generating weighted citation counts based on author contribution weights can be represented as: ; In the formula, Indicates the weighted citation count. Indicates the author's contribution weight. This represents the number of times paper i has been cited. This means finding the largest integer value h such that the sum of the weighted citation counts of the first h papers is at least h. Therefore, Further measure a scholar's academic contribution from the perspective of citation count.

[0072] Specifically, the process of generating contribution deviation terms can be represented as follows: ; In the formula, Indicates the weighted citation count. Indicates the deviation from the contribution level. This represents the median weight of the contribution of all authors in the field f to which the document belongs. The standard deviation represents the weight of the contributions of all authors in the field f to which the document belongs.

[0073] In this embodiment, the process of adjusting weights in the computational domain includes: Calculate a first logarithm of the number of scholars in the field to which the literature belongs, a second logarithm of the number of scholars in the largest field, and calculate the field-adjusted weights based on the ratio of the first logarithm to the second logarithm.

[0074] Specifically, the process of generating the domain adjustment weights can be represented as follows: ; In the formula, This indicates that the domain weights are adjusted. This represents the number of scholars in field f. This indicates finding the field f with the largest number of scholars in the current field. This represents the first logarithm with base 10. It represents the second logarithm with base 10.

[0075] In this embodiment, the process of calculating the comprehensive center index includes: The domain time-series center index is generated by multiplying the weighted center index and the time-series center index. The comprehensive center index is generated by weighting the contribution deviation term and the domain time series center index.

[0076] Specifically, the process of generating comprehensive center indicators can be represented as follows: ; In the formula, Indicates the comprehensive center indicators, This indicates the weighted centrality indicator. Indicates the time-series centrality index, This indicates that the domain weights are adjusted. Indicates the domain time-series center index, Indicates the deviation from the contribution level. This represents the weighting coefficient, which is preferably 0.3.

[0077] In this embodiment, the process of iteratively constructing the initial academic collaboration network based on the comprehensive center index includes: The values ​​from the previous cycle are adjusted based on the comprehensive center index to generate the values ​​for the current cycle. The initial academic collaboration network is iteratively constructed based on the current cycle iteration value, and the iterative construction of the initial academic collaboration network is determined to stop based on the comparison between the difference between the current iteration value and the next cycle iteration value and a set threshold.

[0078] Specifically, the process of generating the values ​​for the next cycle iteration can be represented as: ; In the formula, This represents the value of the previous iteration of the currently calculated first or second node. This represents the current iteration value of the first or second node being calculated. This represents the comprehensive centrality index of the first or second node currently being calculated. Indicates the coefficient used.

[0079] Specifically, the convergence condition for stopping the iterative construction of the initial academic collaboration network can be expressed as: ; In the formula, , This represents the value of the current iteration. This indicates an academic collaboration network or initial academic collaboration network. This indicates the setting of a threshold value, preferably... .

[0080] In this embodiment, the process of iteratively constructing the initial academic collaboration network based on the current period iteration values ​​includes: The node attraction coefficient is calculated based on the current periodic iteration value of the first node and the current periodic iteration value of the second node in the initial academic cooperation network. The node attraction coefficients are passed through a Sigmoid function to generate edge generation probabilities; The initial academic collaboration network is iteratively constructed based on the edge generation probability.

[0081] Specifically, the process of calculating the edge generation probability can be expressed as: ; In the formula, This represents the node attraction coefficient between the first node u and the second node v. This represents the shortest path distance in the current network. Indicates the current network diameter. Let u and v represent the current iteration values ​​of the first node and the second node, respectively. This represents the Sigmoid function (controlling the probability within [0,1]). This represents the edge generation probability. Therefore, determining whether to construct a connection edge between two nodes based on a comparison of the edge generation probability and a threshold ensures that the constructed academic collaboration network clearly displays the collaborative relationships between scholars.

[0082] In particular, the fairness of cross-domain evaluation is improved by statistical standardization of contribution deviations, and the comprehensiveness of the comprehensive central indicators is optimized by multi-level integration. That is, by combining domain time-series weighted centrality with the weighted fusion of contribution deviations, a three-dimensional coupling of static contribution, dynamic timeliness and domain benchmark is achieved.

[0083] In this embodiment, author contribution weights are calculated based on the order of authorship, thus introducing a mechanism for linking scholar contribution differences based on authorship order into network modeling. A weighting mechanism for centrality indicators based on the publication time of the literature is introduced to generate time-series centrality indicators for the Neo4j graph database. The centrality indicators are further adjusted based on contribution deviation and time decay terms, and the academic collaboration network is iteratively constructed based on the centrality indicators. This allows the calculation of centrality indicators to be performed without additional data format conversion, directly using the author contribution weights of the edges as path weights to quickly solve for the weighted shortest path between nodes. Through contribution quantification, dynamic time-series adjustment, and multi-dimensional indicator integration, the computational efficiency of complex centrality indicators under large data volumes is significantly improved. By utilizing the Neo4j graph database architecture's weighted centrality index, weighted intermediary centrality index, and weighted eigenvector centrality index, a comprehensive characterization of scholar network roles is achieved. The weighted centrality index considers the breadth of collaboration while highlighting the value of high-contribution collaborations, avoiding the bias of quantity masking quality. The weighted intermediary centrality index quantifies the bridging role of nodes through path contribution weights, accurately identifying key intermediary scholars in cross-disciplinary collaborations. The weighted eigenvector centrality index weights each centrality index with authorship ranking contribution weights, making the weight assignments of network edges more closely reflect the actual contribution logic of academic collaborations. This provides a more scientific quantitative tool for scholar influence assessment and collaboration pattern discovery. Furthermore, through multi-dimensional time modeling of time-series centrality indicators, the dynamic analysis capabilities of the Neo4j graph database architecture's academic collaboration network are improved, providing a more research-oriented quantitative tool for research evaluation, collaboration recommendation, and policy formulation. The fairness of cross-domain evaluation is improved by statistical standardization of contribution deviations, and the comprehensiveness of the comprehensive central indicators is optimized by multi-level integration. That is, by combining domain time-series weighted centrality with the weighted fusion of contribution deviations, a three-dimensional coupling of static contribution, dynamic timeliness and domain benchmark is achieved.

[0084] Those skilled in the art will recognize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0085] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing an academic collaboration network based on big data, characterized in that, include: Construct an initial academic collaboration network with scholars as nodes and authorship relationships between scholars as edges, based on the Neo4j graph database architecture; The author contribution weights of the edges of the initial academic collaboration network are calculated based on the author authorship order of the literature, and the weighted centrality index of the initial academic collaboration network is calculated based on the author contribution weights. The weighted centrality index is adjusted based on the publication time of the literature to generate a time-series centrality index; The weighted citation count is calculated based on the author contribution weight, and the contribution deviation is calculated based on the statistical value of the contribution weight of all authors in the field to which the document belongs. The field adjustment weight is calculated based on the number of scholars in the field to which the document belongs. A comprehensive center index is calculated based on the contribution deviation term, the domain adjustment weight, and the time-series center index. The initial academic collaboration network is then iteratively constructed based on the comprehensive center index to generate the academic collaboration network.

2. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The process of calculating the author contribution weight includes: If the scholar is not the corresponding author of the document, the first contribution weight is generated based on the ratio of the number of authors of non-corresponding authors to the sum of the number of authors of the document. If a scholar is the corresponding author of a document, the ratio of the first-ranked non-corresponding author's authorship ranking to the sum of the ranking values ​​of the authors is calculated to generate the second contribution weight. The author contribution weight is generated by summing the first or second contribution weight of the first scholar and the first or second contribution weight of the second scholar who authored the same document, and then dividing by the total number of authors.

3. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The process of calculating the weighted centrality index of the initial academic collaboration network includes: A weighted degree centrality index is generated based on the sum of the author contribution weights of all edges connecting the nodes in the initial academic collaboration network. Based on the author contribution weight of the shortest path connecting two nodes in the initial academic collaboration network, a weighted intermediary center index is generated. Based on the node feature vectors of the initial academic collaboration network and the author contribution weights, a weighted feature vector center index is generated. The weighted center index is generated by weighting the weighted degree center index, the weighted intermediary center index, and the weighted feature vector center index.

4. The method for constructing an academic collaboration network based on big data according to claim 3, characterized in that, The process of generating the weighted intermediary center index includes: Obtain the first total value of the author contribution weight of the shortest path from the first node to the second node and through the third node in the initial academic collaboration network; Obtain the second total value of the author contribution weight of the shortest path from the first node to the second node of the initial academic collaboration network; A weighted intermediary center index is generated based on the ratio of the first total value to the second total value.

5. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The process of generating time series centrality indicators includes: An exponential calculation based on the difference between the current time and the publication time of the document is used to generate a publication attenuation term; An exponential calculation based on the current time and the earliest publication time of the literature stored in the initial academic collaboration network is used to generate an overall decay term; The normalized calculation of the literature publication attenuation term based on the overall attenuation term is used to generate time-series weights; The time-series centrality index is generated by weighting the weighted centrality index and the time-series weights based on multiple time slices stored in the initial academic collaboration network.

6. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The statistical values ​​include the median and standard deviation. The process for calculating the contribution deviation term includes: The contribution deviation term is generated by subtracting the median of the contribution weights of all authors in the field of the document from the weighted citation count and dividing by the standard deviation of the contribution weights of all authors in the field of the document.

7. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The process of adjusting weights in the computing domain includes: Calculate a first logarithm of the number of scholars in the field to which the literature belongs, a second logarithm of the number of scholars in the largest field, and calculate the field-adjusted weights based on the ratio of the first logarithm to the second logarithm.

8. The method for constructing an academic collaboration network based on big data according to claim 1, characterized in that, The process of calculating the comprehensive center index includes: The domain time-series center index is generated by multiplying the weighted center index and the time-series center index. The comprehensive center index is generated by weighting the contribution deviation term and the domain time series center index.

9. The method for constructing an academic collaboration network based on big data according to any one of claims 1 to 8, characterized in that, The process of iteratively constructing the initial academic collaboration network based on the comprehensive center index includes: The values ​​from the previous cycle are adjusted based on the comprehensive center index to generate the values ​​for the current cycle. The initial academic collaboration network is iteratively constructed based on the current cycle iteration value, and the iterative construction of the initial academic collaboration network is determined to stop based on the comparison between the difference between the current iteration value and the next cycle iteration value and a set threshold.

10. The method for constructing an academic collaboration network based on big data according to claim 9, characterized in that, The process of iteratively constructing the initial academic collaboration network based on the current periodic iteration values ​​includes: The node attraction coefficient is calculated based on the current periodic iteration value of the first node and the current periodic iteration value of the second node in the initial academic cooperation network. The node attraction coefficients are passed through a Sigmoid function to generate edge generation probabilities; The initial academic collaboration network is iteratively constructed based on the edge generation probability.

Images