AU2020327352B2

AU2020327352B2 - Key node identification method based on technology graph

Info

Publication number: AU2020327352B2
Application number: AU2020327352A
Authority: AU
Inventors: Bin HUA; Qiyu Lu; Ping Song; Qiqi ZHANG; Sanshan ZHAO
Original assignee: State Grid Shanghai Electric Power Co Ltd
Current assignee: State Grid Shanghai Electric Power Co Ltd
Priority date: 2020-06-18
Filing date: 2020-12-14
Publication date: 2023-01-05
Anticipated expiration: 2040-12-14
Also published as: AU2020327352A1; WO2021253758A1; CN111813951A

Abstract

The present disclosure relates to a key node identification method based on a technology graph, including: building a technology graph; performing centrality calculation on node data in the technology graph, to obtain a key node; reducing dimensions of the technical indicators of the node data by using a principal component analysis method; and analyzing a relationship between the key node and the simplified technical indicators, to obtain key nodes of different categories. Compared with the prior art, the present disclosure integrates network centrality indicators and literature measurement of scientific and technological resources, and resolves disadvantages such as undiversified and impractical indicators for identifying a key node in the technology graph. Based on relevant theories of complex network technologies, quantitative calculation is performed on relevant indicators of the technology graph, thereby helping more accurately to identify a key node, to discover a trend of technology research or a technology trend clue, and to provide decision support for technological innovation.

Description

KEY NODE IDENTIFICATION METHOD BASED ON TECHNOLOGY GRAPH TECHNICAL FIELD

[0001] The present disclosure relates to a data processing method, and in particular, to a key node identification method based on a technology graph.

BACKGROUND

[0002] Identifying key nodes, that is, key technologies and hot technologies, in a technology graph network greatly helps scientific development and technological innovation. Traditional discussion about key nodes in a network often focuses on centralization of complex networks and evaluation of node importance, and statistical properties of the network are measured by using empirical methods. Use of a single measurement indicator or method to identify a key node is not comprehensive. Each measurement indicator or method can reflect the importance of a node in the network only from a particular side. This does not conform to the actual situation. In the era of rapid development of the Internet, a simple combination of measurement indicators cannot satisfy actual requirements, and higher requirements for accuracy of identifying key nodes are put forward.

[0003] Especially now, the application of the networks are more widely and has more practical significance. The measurement indicators merely from a theoretical perspective do not meet the actual situation, and reduce the accuracy of identifying key nodes.

SUMMARY

[0004] An objective of the present disclosure is to provide a key node identification method based on a technology graph, to overcome the disadvantages in the prior art and resolve a problem of undiversified and impractical indicators for identifying a key node in a technology graph.

[0005] The objective of the present disclosure may be achieved by the following technical process:

[0006] A key node identification method based on a technology graph includes:

[0007] building a technology graph;

[0008] performing centrality calculation on node data in the technology graph, to obtain key nodes;

[0009] reducing technical indicators of the node data by a principal component analysis method, wherein the technical indicators are in multiple dimensions; and

[0010] analyzing a relationship between each of the key nodes and each of reduced technical indicators by a linear regression method, to obtain a key node representing the multiple dimensions.

[0011] The technology graph is built by extracting scientific and technological achievements from multiple websites and databases by using an entity, relationship, and attribute extraction method, and performing knowledge fusion on the extracted scientific and technological achievements.

[0012] The websites and the databases include at least one of https://www.cnki.net/, http://www.drcnet.com.cn/, a self-built resource library, research and development institution data, policy and regulation data, industry dynamics data, a patent database, and an industry standard database.

[0013] The centrality includes degree centrality, closeness centrality, and betweenness centrality.

[0014] The dimensions of the technical indicators are classified as project level, talent level and a scientific research achievements level.

[0015] Technical indicators of the project level dimension include a total quantity of projects, a fund project category, and an investment in scientific research.

[0016] Technical indicators of the talent level dimension include an average age of talents, average education of talents, and a quantity of talents.

[0017] In the scientific research achievements level dimension, scientific research achievements include papers, patents, and other achievements.

[0018] Paper-related technical indicators include a total quantity of papers, a total citation frequency, a quantity of core journal papers, a total citation frequency of core journal papers, a quantity of funded papers, a total citation frequency of funded papers, a percentage of core journal papers, a total citation frequency per paper, a citation frequency per core journal paper, a citation frequency per funded paper, and an h-indicator. Patent-related technical indicators include a total quantity of patents and a quantity of invention patents. Technical indicators related to other achievements include an achievement award, an achievement appraisal result, a quantity of standards, and editor-in-chief or associate editor publications.

[0019] Compared with the prior art, the present disclosure integrates network centrality indexes and literature measurement of scientific and technological resources, and resolves disadvantages such as undiversified and impractical indexes for identifying a key node in the technology graph. Based on relevant theories of complex network technologies, quantitative calculation is performed on relevant indexes of the technology graph, thereby helping more accurately to identify a key node, to discover a trend of technology research or a technology trend clue, and to provide decision support for technological innovation.

BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG. 1 is a flowchart of a key node identification method based on a technology graph according to this embodiment.

[0021] FIG. 2 is a technology graph built according to this embodiment.

[0022] FIG. 3 is a curve diagram of a cumulative contribution rate of each evaluation index according to this embodiment.

DETAILED DESCRIPTION

[0023] The present disclosure is described in detail below with reference to the accompanying drawings and specific embodiments. The embodiments are implemented on the premise of the technical solution of the present disclosure and provide the detailed implementations and specific operation processes, but the protection scope of the present disclosure is not limited to the following embodiment.

[0024] Embodiment

[0025] As shown in FIG. 1, a key node identification method based on a technology graph includes the following steps.

[0026] (1) Build a technology graph

[0027] Obtain metadata from https://www.cnki.net/, http://www.drcnet.com.cn/, a self-built resource library, external experts and research and development institution data, internal projects and scientific and technological achievements data, adding policy and regulation data, industry dynamics data, patent data, and industry standard data, and perform entity, relationship, and attribute extraction on the metadata, and perform entity disambiguation and co-reference resolution on the extracted information, to build a technology graph, as shown in FIG. 2.

[0028] (2) In consideration of statistical indicators of a complex network, locate a key node based on centrality indicators such as degree centrality, closeness centrality, and betweenness centrality, a node with high betweenness centrality and a high-frequency is a key technology in the field, representing hot topics of research within this period.

[0029] The degree centrality is a sum of direct connections between one node and other nodes. As a connection in the technology graph is directional, the degree centrality may be classified into in-degree centrality and out-degree centrality. The formula of degree centrality is: D (u)=

U=1 Xv(v # u), where u is any node in the technology graph, n is a quantity of nodes in the graph, and Xvu represents whether a node v is directly connected to the node u. The degree centrality is a most direct measurement indicator of node centrality in network analysis and reflects cohesion of a node. Higher degree centrality of a node indicates higher importance of the node in a network.

[0030] The closeness centrality is a reciprocal of a sum of shortest path distances from one

node to all other nodes, and reflects closeness between a node and another node in a network. n-1 The formula of closeness centrality is: Cnorm(u) = -,1 where u is any node in the

technology graph, n is a quantity of nodes in the graph, and d(u, v) is a shortest path distance

between another node v and the node u. Because a connection in the technology graph is

directional, the closeness centrality may be classified into in-closeness centrality and out

closeness centrality. The in-closeness centrality reflects integration power of a node, and the

out-closeness centrality reflects radiation power of a node.

[0031] The betweenness centrality is a quantity of shortest paths passing through a node, that

is, a quantity of times that a node serves as a bridge for a shortest path between any two nodes.

The formula of betweenness centrality is: B(u) = s, where u is any node in the p technology graph, p is a total quantity of shortest paths between a node s and a node t, and p(u)

is a quantity of shortest paths between the node s and the node t through the node u. A larger

quantity of times that a node serves as an "intermediary" indicates higher betweenness centrality

of the node, and the node serves as a "transportation hub" in the network.

[0032] (3) Performing literature measurement based on scientific and technological resources, starting from two aspects: scientific research investment and scientific research achievements.

[0033] In terms of scientific research investment, there include scientific research projects and talent teams. In terms of scientific research projects, there include a total quantity of projects, fund projects, and investment in scientific research. In terms of talent teams, there include an average age of talents, average education of talents, and a quantity of talents.

[0034] The scientific research achievements include papers, patents, standards, monographs, and achievements. Paper-related factors include a total quantity of papers, a total citation frequency, a quantity of core journal papers, a total citation frequency of core journal papers, a quantity of funded papers, a total citation frequency of funded papers, a percentage of core journal papers, a total citation frequency per paper, a citation frequency per core journal paper, a citation frequency per funded paper, and an h-indicator. Patent-related factors include a total quantity of patents and a quantity of invention patents. Achievements include an achievement award, an achievement appraisal result, a quantity of standards, and editor-in-chief or associate editor publications.

[0035] (4) Transform multiple dimensions of evaluation indicators defined in (2) and (3) into mutually independent comprehensive evaluation indicators by using principal component analysis, eliminate correlations between the evaluation indicators, and reduce a quantity of indicators for node importance evaluation.

[0036] In the present disclosure, a technology graph is built for a co-occurrence relationship of 200 technologies in scientific and technological data, and node importance is evaluated in terms of dimensions of a network topology, a project level, a talent level, and scientific research achievements. 27 evaluation indicators corresponding to each technology are separately calculated, to form a 200*27 matrix, and principal component analysis is performed on the matrix, to obtain an eigenvalue, a contribution rate, and a cumulative contribution rate. The cumulative contribution rate is shown in FIG. 3.

[0037] It can be learned from the figure that a cumulative contribution rate of the first five principal components reaches 90.79%. Therefore, selecting only the first five principal components can fully represent information contained in the 27 evaluation indicators. The evaluation matrix can be reduced to a 200*5 matrix by calculating a product of an original indicator weight value matrix and an evaluation indicator matrix that correspond to the first five principal components.

[0038] (5) The comprehensive index of evaluating the node criticality can be expressed by a linear regression, the formula is showed in (4):

[0039] Z = 0.3284*y1 + 0. 1 5 3 1 *y2 + 0.2157*y3 + 0.1196*y4 + 0.0911*y5

[0040] where yi to y5 represent the first five principal components obtained through principal component analysis in (4), the coefficients are the contribution rate of the first five principal components.

[0041] The values calculated from the above formula are sorted to obtain key nodes, and the key nodes are highlighted in the network for easy identification. In addition, this method can also be used to identify key nodes in networks, such as research fields network, authors network, and research institutes network.

[0042] The reference to any prior art in this specification is not, and should not be taken as, an acknowledgement or any form of suggestion that such prior art forms part of the common general knowledge.

[0043] It will be understood that the terms "comprise" and "include" and any of their derivatives (e.g. comprises, comprising, includes, including) as used in this specification, and the claims that follow, is to be taken to be inclusive of features to which the term refers, and is not meant to exclude the presence of any additional features unless otherwise stated or implied.

Claims

CLAIMS: 1. A key node identification method based on a technology graph, comprising:

building a technology graph;

performing centrality calculation on node data in the technology graph, to obtain key nodes;

reducing technical indicators of the node data by a principal component analysis method, wherein the technical indicators are in multiple dimensions; and

analyzing a relationship between each of the key nodes and each of reduced technical indicators by a linear regression method, to obtain a key node representing the multiple dimensions.
2. The key node identification method based on a technology graph according to claim 1, wherein the technology graph is built by extracting scientific and technological achievements from multiple websites and databases by using an entity, relationship, and attribute extraction method, and performing knowledge fusion on the extracted scientific and technological achievements.
3. The key node identification method based on a technology graph according to claim 2, wherein the websites and the databases comprise at least one of https://www.cnki.net/, http://www.drcnet.com.cn/, a self-built resource library, research and development institution data, policy and regulation data, industry dynamics data, a patent database, and an industry standard database.
4. The key node identification method based on a technology graph according to claim 1, wherein the centrality comprises degree centrality, closeness centrality, and betweenness centrality.
5. The key node identification method based on a technology graph according to claim 1, wherein the multiple dimensions of the technical indicators are classified as project level, talent level and scientific research achievements level.
6. The key node identification method based on a technology graph according to claim 5, wherein technical indicators of the project level dimension comprise a total quantity of projects, a fund project category, and an investment in scientific research.
7. The key node identification method based on a technology graph according to claim 5, wherein technical indicators of the talent level dimension comprise an average age of talents, average education of talents, and a quantity of talents.
8. The key node identification method based on a technology graph according to claim 5, wherein in the scientific research achievements level dimension, scientific research achievements comprise papers, patents, and other achievements.
9. The key node identification method based on a technology graph according to claim 8, wherein paper-related technical indicators comprise a total quantity of papers, a total citation frequency, a quantity of core journal papers, a total citation frequency of core journal papers, a quantity of funded papers, a total citation frequency of funded papers, a percentage of core journal papers, a total citation frequency per paper, a citation frequency per core journal paper, a citation frequency per funded paper, and an h-indicator; patent-related technical indicators comprise a total quantity of patents and a quantity of invention patents; and technical indicators related to other achievements comprise an achievement award, an achievement appraisal result, a quantity of standards, and editor-in-chief or associate editor publications.