CN111428056A - Method and device for constructing scientific research personnel cooperative community - Google Patents

Abstract

The invention relates to a method and device for constructing a cooperative community of scientific researchers, and belongs to the technical field of data processing. It includes the following steps: obtaining the data of scientific research projects and researchers participating in the corresponding scientific research projects, as well as the data of scientific research results and the researchers who output the corresponding scientific research results; screening the obtained data based on the influence of the researchers: deleting the influence from the cooperation data Researchers who are lower than the set value; according to the filtered cooperation data, a network of scientific researchers' cooperation relationship is generated, which includes: the number of cooperation between each scientific researcher and other researchers, and the number of cooperation refers to the number of cooperation between two researchers. The number of times that you have participated in the same scientific research project, or the number of times that the same scientific research results have been produced; using the community discovery algorithm, based on the cooperative relationship network of researchers and the influence of researchers, a collaborative community of researchers is generated. The method of the invention reduces the amount of data processing and improves the construction efficiency and information accuracy of the map.

Description

A method and device for constructing a cooperative community of scientific researchers

technical field

The invention relates to a method and device for constructing a cooperative community of scientific researchers, and belongs to the technical field of data processing.

Background technique

The scientific and technological knowledge graph cooperative community refers to a small group composed of researchers with close cooperation relationships. The relationship between them is displayed in a visual map, so as to discover the cooperation between researchers in scientific research activities. The cooperative community network in the science and technology knowledge graph has similar characteristics and community structure to the social network in real life, that is, people belong to different communities, the whole network is composed of several communities, and the nodes in each community are relatively closely connected. The connection between them is relatively sparse. The size of each node in the network and the thickness of the connection have practical significance to reveal the influence of network nodes, the cooperation relationship between partners and the closeness of cooperation. The community discovery method is to discover groups with modular structure characteristics from complex network relationships, and combine domain knowledge data to realize the exploration of the community structure of people in the domain.

Community discovery algorithms were first used in the field of social networks to discover or find social groups with the same interests and hobbies. In the field of science and technology, various fields have accumulated a large amount of valuable scientific research activity data after years of development. Community discovery based on data on researchers' collaborations yields a map that reflects community groups of researchers with close collaborations. Reflecting that researchers in the same community have certain similarities in academic research, large communities also represent a development and research direction in the corresponding technical field, providing extremely valuable technical information and scientific research materials for It provides a reference for finding research and development topics, provides support for finding partners for related research and development work, and plays an auxiliary role in the discovery of research teams in different research institutions.

Today, with the intensive development of technology, in scientific research activities, the development or research of any field or direction cannot be isolated, and the exploration and research of a certain topic in a certain field cannot be separated from the technical support of other fields. It has also led to the generation of a large number of interdisciplinary disciplines and the more and more subdivided and precise branches of technology. In this context, auxiliary technologies or supporting technologies in other fields will appear in various R&D directions in various fields in technical cooperation, which leads to a certain technical field, such as the field of tobacco technology, in the cooperation data of scientific researchers, in addition to relative to this field. There are also a large number of active technical personnel in the field, and there are also a large number of active scientific research and technical personnel in the field of edge technology or other fields, resulting in complex and huge cooperative relations and cooperative data, resulting in a large amount of calculation during community discovery, and the computational efficiency of community discovery algorithms is low. Large hardware computing resources make it inconvenient to update data in real time or frequently, resulting in information lag, and it is difficult to maintain the accuracy of the information map. At the same time, each researcher in the cooperative data is the smallest unit in the cooperative community, and it is also a node on the final information graph. The large number of nodes also directly leads to redundant graph information, poor readability, and difficulty in extracting direct and effective information; The judgment of the scientific research direction corresponding to the cooperative community should also refer to the research direction of the researchers who are nodes in the community. Therefore, the existence of a large number of researchers in other fields will affect the judgment of the scientific research direction of the community, resulting in the scientific research direction reflected by the community and related topics involved. Deviations or even errors occur, resulting in the final communication of wrong technical intelligence, which greatly affects the user experience.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method and device for constructing a cooperative community of scientific researchers, so as to solve the problem that the existing community discovery method is used in the construction of cooperative community in the technical field due to the large amount of computation, the poor readability of the generated community graph, and the problem of cooperation in the graph. There is a problem that the information in the community is biased and affects the user experience.

To achieve the above object, the scheme of the present invention includes:

A method for constructing a cooperative community of scientific researchers of the present invention includes the following steps:

1) Obtaining cooperation data, the cooperation data includes scientific research projects and scientific researchers participating in corresponding scientific research projects, as well as scientific research results and scientific research personnel who produce corresponding scientific research results; the scientific research results include papers;

2) Establish the influence model of researchers and calculate the influence of researchers;

3) Screening the cooperation data, the screening includes: deleting the scientific researcher whose influence is lower than the influence setting value from the cooperation data;

4) According to the cooperation data after screening, generate a network of scientific researcher cooperation relationship, the scientific researcher cooperation relationship network includes: the number of cooperation between each scientific researcher and other scientific researchers, and the number of cooperation times refers to the corresponding two scientific research personnel. The number of times the staff has participated in the same scientific research project or produced the same scientific research results;

5) Using the community discovery algorithm, based on the cooperative relationship network of researchers and the influence of researchers, a community map of researchers' cooperation is generated.

There are a large number of researchers and their cooperative relationships in a scientific research field. Certain supporting technologies or auxiliary technologies have appeared and used in large numbers in this field. It is valuable for their researchers to appear in the map found by the community in this field, but there are also A large number of researchers in other fields appear only a few times or even once in the cooperation data in this field, and they are all cooperation in a relatively single or several R&D directions. These collaborative data have a certain impact on community discovery and the final community discovery graph.

The processing of these data is very tricky. It is not advisable to directly delete or filter out all researchers who are not in the field based on the field. First, there are frequent scientific research activities and there are many cross-field researchers, so it is difficult to label each researcher with a specific field. Labels; Second, the presence of technical personnel in other fields in the cooperation map in this field is valuable in some cases, which is convenient for discovering the key to some unique technical problems in this field, or discovering and solving related technical problems in this field. Relying on other fields of technology and researchers, these researchers have great reference for technical information, and will bring great help to the practical role of the cooperative community map.

Faced with the above-mentioned technical difficulties in data processing, the present invention filters the scientific research activity data according to the influence of the scientific research personnel in the corresponding scientific research field before the community discovers the algorithm. The influence in this field will also be low; and researchers in other technical fields that reflect the key to the special technical problems in this field will also have a high influence in this field. Therefore, the method of the present invention can firstly reduce the number of researchers in other fields with relatively single research and development direction and few cooperation times in this field; secondly, it can also filter out the field of this field, which has little reference value as technical information. At the same time, it prevents the researchers who provide good support and assistance to various technical key issues in this field from being filtered out, although they are cross-field or marginal fields. The accuracy of the data source is guaranteed.

The method of the invention effectively reduces the data nodes with small reference value and causes interference in the later community discovery map, increases the readability of the map, and improves the user experience; in addition, the finally generated cooperative community effectively reduces the reflection of scientific research directions and related scientific research topics. deviation. At the same time, compared with the method of adjusting relevant parameters and thresholds in the community discovery algorithm, the computational complexity is reduced and the computational efficiency is improved. Provided technical support for the improvement of the update frequency of the community graph.

Further, the influence setting value N is:

Among them, N _max is the highest value of the influence of scientific researchers in the cooperation data.

Further, filtering the cooperation relationship network of scientific researchers in step 4), the filtering includes setting the number of cooperation between scientific researchers whose number of cooperation is lower than the set value of the number of cooperation as 0.

Further, the set value M of the number of times of cooperation is:

Among them, M _max is the highest frequency of collaboration between researchers in the collaboration data.

The analysis of the degree of cooperation of scientific researchers in the present invention is based on the construction of a cooperative relationship network by scientific researchers through the cooperative relationship of projects and various achievements. The quality of the analysis data will directly affect the quality of the found community structure. Therefore, the selection of scientific researchers comprehensively considers the scientific research personnel. On the one hand, it reflects the scientific research ability of scientific research workers and the ability to produce scientific research results; Provide more powerful support for the accuracy of the information reflected in the map scientific research. Therefore, there are two core research staff selection criteria for constructing the author's scientific research cooperation network, one is the lowest research staff influence, and the other is the minimum frequency of research staff cooperation. It can effectively filter out the interference data that cannot truly and accurately reflect the scientific research information of the generated community.

The invention adopts Price's Law to determine the minimum influence score and the minimum project and achievement cooperation frequency of core scientific researchers. The filtering standard is scientific and effective.

Further, in the scientific research personnel influence model, the scientific research personnel influence is determined at least by the number of scientific research outputs, and the scientific research outputs include the following categories: papers and patents.

In this field ("this field" or "the field" in this article mainly refers to the scientific research and technical field for collaborative community discovery), the output of the results is few, and the quality of the results is low. For researchers in this field, it is natural The influence is small, and it has little reference value as a technical information content; for researchers in other fields and cross-fields, even if they participate in a lot of scientific research projects and produce few results, it means that they are only suitable for Assisting with some routine problems does not reflect the special technical problems and solutions in this field. As technical information, the technologies represented in the cooperation data in this field are highly replaceable and have little reference value in the map (that is, researchers in other fields solve routine problems in this field).

For researchers in other fields or across fields, the technologies they represent solve unique problems in this field and bring about innovative effects, so they will definitely produce a considerable degree of scientific research results in this field, such as Patented or published papers (i.e. researchers in other fields address outstanding and unique problems in the field).

Therefore, the results of scientific research such as papers and patents in this field are used as the decisive factors in the influence calculation model, and the filtering effect and accuracy are better.

The following examples illustrate the difference between "scientists in other fields solve conventional problems in this field" and "researchers in other fields solve outstanding and unique problems in this field". For example, researchers in the field of chemistry have synthesized an organic compound, which can greatly improve the taste of cigarettes and reduce the production of harmful substances in the field of tobacco technology as an additive in cigarette technology. Therefore, it has become a unique innovation in the field of tobacco. Correspondingly, researchers in the field of chemistry will also produce a lot of results in the field of tobacco technology, such as papers, patents, and even national or industry standards on the dosage of this compound. Another researcher in the field of control has developed an efficient closed-loop feedback control method for furnace temperature, which can control furnace temperature more accurately and efficiently and reduce energy consumption. Researchers in the field of tobacco technology and temperature control have carried out some cooperation, and this technology is used in the process of tobacco leaf rejuvenation. These cooperation may be limited to the direction of tobacco leaf rejuvenation in the field of tobacco technology, and the number is small, while heating and temperature control It is highly substitutable for the resurgence of tobacco leaves, and other methods can be used for temperature control. Therefore, the reference value of technicians in the field of tobacco control in the cooperative community map in the field of tobacco science and technology is not large, and they remain in the map as a node. It will cause certain information interference, and the large amount of such data will increase the computational complexity of community discovery, increase the complexity of the map, and reduce the readability of the map. And because the technology related to temperature control is highly substitutable in the field of tobacco, researchers in the field of control will have few related achievements in the field of tobacco science and technology, and their achievements should all be produced in the technical field related to control. Therefore, based on the method and the influence model of the present invention, the interference data can be effectively filtered, the calculation amount can be reduced, and the readability of the map can be improved.

Further, the researcher influence model is:

Among them, P is the influence, n is the number of scientific research output of the researcher, S _i is the set score of the category to which the scientific researcher's ith scientific research output belongs, and Wi is the scientific researcher's _ith scientific research output. Out of the set weights in the category of scientific research output.

The influence calculation model for a researcher is determined based on the quantity and quality of the researcher's research output in the research field that generates the map, and at the same time for different research outputs (such as award-winning scientific and technological achievements, papers or standards) and the same type of research output Different basic scores and weights are set for different output levels (such as scientific and technological achievements that have obtained national and provincial scientific and technological awards). Using the model of the invention to calculate the influence of scientific researchers in the corresponding field is objective and accurate, and the obtained numerical values are convenient for comparison.

Further, in step 4), when generating the scientific research personnel cooperative relationship network, for the scientific research project earlier than the set time, the cooperation generated by any two scientific research personnel due to participating in this scientific research project is recorded as a time, and 0 <a<1. When calculating the number of cooperation, different mapping methods can also be used to map the number of cooperation at a specific time to a value from 0 to 1, so as to reflect the influence of different times.

In the process of building the cooperative relationship network, the degree of cooperation (number of times) between researchers is calculated, in addition to the frequency of cooperation relationship as the main analysis data, in order to ensure that the analysis result data is more timely and the information reflected by the generated community is more valuable, scientific research The degree of cooperation between personnel is based on the calculation of the author's cooperation frequency relationship value, and the time series weight parameter of the cooperation relationship is added.

Further, the community discovery algorithm is a louvain algorithm.

Further, adding a topic tag to each community in the scientific researcher cooperative community graph is, the topic tag is a set number of paper keywords with the highest word frequency among the papers published by each scientific researcher serving as a node in the corresponding community. .

Using keywords with high frequency among all keywords of all papers published by all researchers in the community as tags can reflect the research direction and research focus of this scientific research community, increase the readability of the map, and help readers identify research in each community. theme.

An apparatus for constructing a scientific researcher cooperative community of the present invention includes a processor and a memory, and the processor executes the instructions stored in the memory to realize the above-mentioned construction method for a scientific researcher cooperative community.

Description of drawings

Fig. 1 is the construction method flow chart of the scientific research personnel cooperative community of the present invention;

Figure 2 is a diagram of the knowledge entity relationship network model in the scientific research field;

Figure 3 is a schematic diagram of the influence model of researchers;

FIG. 4 is an example of a collaborative community map in the field of tobacco scientific research generated by the method of the present invention.

Detailed ways

The present invention will be described in further detail below with reference to the accompanying drawings.

Method example:

The construction method of the scientific researcher cooperative community of the present invention is described by taking the field of tobacco scientific research as an example to illustrate the method of the present invention. The method flow is shown in Figure 1, which specifically includes the following steps:

S1. Obtain the original data of scientific research cooperation. The acquisition of scientific research cooperation data mainly includes the extraction, fusion and screening of the data of scientific research projects and researchers participating in scientific research projects, scientific research results and scientific research personnel who made scientific research results. For the tobacco field, cooperative knowledge graph data sources include data in XML, RDB, EXCEL, HTML and other formats. First, formulate the core metadata of each entity, relationship, and attribute. For example, to establish a responsibility relationship (responsibleFor) between researchers and research projects, the core metadata templates of researcher entities, research project entities, and responsibility relationships are shown in Tables 1, 2, and 3 below:

Table 1 The core metadata template of the researcher entity

Table 2 Core metadata templates of scientific research project entities

serial number property name type illustrate 1 CODE long Project ID 2 NAME text project name 3 START_TIME text Project start date 4 END_TIME text project deadline 5 SOURCE text project source 6 CONTENT text Research content and significance 7 INDICATORS text Project indicators 8 PROGRAM text Project implementation plan

Table 3 assume relational core metadata template

serial number property name type illustrate 1 PERSON_CODE long Scientist ID 2 PROJECT_CODE long Research project ID 3 IS_LEADER text Is the project leader

By parsing different data sources, the data is stored according to the above template to form the original data. That is, according to the core metadata, the extraction templates of different data sources are formulated; then the original data is collected through the formulated data extraction templates.

The original data includes scientific research projects and researchers involved in corresponding scientific research projects, as well as scientific research results and researchers who have made corresponding scientific research results. The relationship between scientific research achievements, scientific research projects and scientific research personnel is shown in Figure 2. The scientific research personnel participating in or responsible for the same project form a cooperative relationship; the scientific research personnel who jointly produce the corresponding results (according to the signed scientific research personnel corresponding to the results) form a cooperative relationship. Scientific research results include papers, patents, standards, works, and achievement awards. Targeted data extraction templates are formulated for different achievement forms and project record forms, and the extracted data that can reflect the cooperative relationship of researchers forms the original data.

S2. Perform data fusion processing on the original data, and the data processing rules are shown in Table 4. Due to the variety of data extraction channels, the record formats and content fields of the original data extracted from different channels are different, and the extracted scientific research personnel and scientific research projects may also be duplicated. Integrate data from different sources according to the data fusion rules to supplement and improve the data. field to remove duplicate data. Tobacco Cooperative Knowledge Graph scientific research knowledge data fusion adopts a model-driven method, generates ETL data fusion language through configuration, and ETL data fusion language is parsed and executed by a parsing engine. Finally, the ETL data fusion language will be mapped to Spark SQL execution to complete the fusion of raw data. The original data is fused to form a graph database.

Table 4 Data fusion rules for tobacco researchers and scientific research projects

S3. Calculation of the influence of scientific researchers. Build a researcher's influence calculation model, and use the pre-calculation method to synchronously update the influence as an attribute value to the graph database. The attribute value can be updated regularly according to the data update frequency.

The calculation model of scientific research personnel influence is shown in Figure 3. The calculation of influence is based on the scientific research output of the corresponding scientific research personnel in the field of tobacco scientific research, and specifically includes a two-level index system, which is the scientific research output type and the corresponding scientific research output type. For the scientific research output level, there may also be a third-level indicator for some types of scientific research output, such as the corresponding award level in the reward (first prize, second prize, etc.). The present invention does not limit the division of scientific research output types and higher-level indicators. In this embodiment, scoring the last-level indicators of each type of output can specifically be: obtaining the last-level indicators of each type of output according to the basic score of the output type and the corresponding additional points of the second- and third-level indicators (that is, the output scores of various outputs), the last-level indicators are the second-level indicators (corresponding to patents, papers, standards, and works in Figure 3), and the outputs under the second-level indicators are scored, and there are three-level indicators. (corresponding to the reward in Figure 3), then score the output under the three-level indicator. The score of each researcher based on a research output is the output score of the output multiplied by the score weight. The score weight is related to the contribution of the researcher in the output. The contribution can be determined by the output of the output. The order of signatures is determined, or based on the original record of the output, such as a work log. The final impact score of a researcher is obtained by accumulating the scores of their various outputs. Based on this model, the formula for calculating the researcher's score is:

Among them, P is the influence score of a researcher, n is the number of scientific research outputs of the researcher, and S _i represents the score of the last level indicator corresponding to the ith scientific research output of the researcher (that is, the output of the corresponding output). Score), Wi represents the score weight (determined by the contribution size) corresponding to the _ith scientific research output of the researcher.

For example, if a technician's scientific and technological achievements have won the third prize at the provincial and ministerial level, the staff score for this output will be the output score of the third prize at the provincial and ministerial level, multiplied by the The weight of the score determined by the contribution made by the technician in the scientific and technological achievements.

S4. Data screening. In order to ensure the quantity and quality of the researcher's project cooperation data, the selection of the researcher is carried out. The researchers whose influence in the field is higher than the threshold are used as the objects for generating and displaying the cooperation map. According to Price's Law, the researchers whose influence is less than N in the field are filtered out. The calculation formula of N is as follows:

Among them, N _max represents the highest influence value among researchers in the field. After the N value is calculated, the researchers with influence greater than or equal to N in the field are regarded as the researchers in the cooperative community analysis data source.

S5. Generate researcher project partnership data. Based on the screened data of scientific researchers, by calculating the frequency of project cooperation between scientific researchers in pairs, the project cooperation relationship network of scientific researchers can be obtained (that is, the data of cooperation frequency).

Specifically, for example, researchers a, b, c, d, and e are used as node data for researchers, and their project partnership set is: {<a,b>,<a,c>,<a,d>,<a,e >,<b,c>,<b,d>,<b,e>,<c,d>,<c,e>,<d,e>}, the constructed partnership network is represented in the following matrix form:

Table 5 Tobacco research staff partnership matrix

Among them, the numbers in the matrix represent the number of collaborations between two scientific researchers in the horizontal and vertical directions. In this embodiment, if the corresponding two scientific research personnel participate in the same scientific research project once, it is considered that a cooperation occurs; If a scientific research result is produced, it is considered to produce a cooperation.

Further, further filtering can be performed on the project cooperation relationship network of researchers, and the cooperation relationship whose number of cooperation is lower than the set threshold can be deleted, so as to reduce the interference of accidental project cooperation on the final cooperation map, and further reduce the number of community discoveries. The computational effort of the algorithm.

Specifically, according to Price's Law, the threshold of the number of collaborations is positioned M, that is, the number of collaborations between two researchers whose number of collaborations is lower than M is set to 0. The formula for calculating M is as follows:

Among them, M _max represents the highest value of the number of collaborations between researchers in the cooperation data in this field. After the M value is calculated, two researchers whose number of collaborations is lower than this value are considered to have no collaboration.

For example, if the calculated value of M is 3.3, the two researchers a and c whose number of collaborations are lower than 3.3 are considered to have not cooperated. The cooperation matrix in Table 5 is adjusted as follows:

Table 6 Tobacco research staff partnership matrix (after filtering the number of collaborations)

a b c d e a 14 0 19 28 b 14 12 26 twenty two c 0 12 32 13 d 19 26 32 16 e 28 twenty two 13 16

Further, further filtering can be performed on the project cooperation relationship network of scientific researchers, and the weight of the cooperation whose cooperation time is earlier than the set time can be set to reduce the influence of early cooperation. Because the research directions and fields of researchers may change over time, early collaboration data is less meaningful and valuable to the collaborative community map. Therefore, one-level or multi-level weights can be set. For example, the weight of the cooperation relationship between two researchers in the past 2 years is 1, the weight of the cooperation relationship between the past 2-5 years is 0.8, and the weight of the cooperation relationship between more than 5 years is 0.8. is 0.5. That is, before filtering based on the cooperation time, the 4 project cooperation between the two researchers 6 years ago is regarded as 4 cooperation. After filtering according to the above weight, it is only regarded as 2 cooperation; another example is 1 cooperation 3 years ago. After only 0.8 cooperation.

S6. Community Discovery. The Louvain algorithm is used to conduct community discovery based on the finally obtained researcher collaboration network (cooperation frequency data) and researcher influence data. The Louvain algorithm is a modularity-based community discovery algorithm that can discover hierarchical community structures and maximize the discovery of the communities where researchers in the field of tobacco research are located (groups of researchers with cooperative relationships, research directions and fields similar).

Specifically, the louvain algorithm is used for the scientific research personnel cooperative relationship network formed by the scientific research personnel project cooperative relationship matrix, which specifically includes the following steps

1) Carry out the first phase of traversing the nodes in the network and transfer nodes between communities. For each node A in the network, try to add node A to the community where each neighbor node is located in turn, and calculate the modularity change ΔQ before and after joining. If the largest ΔQ>0, add node A to the community where the neighbor node with the largest ΔQ is located, otherwise the community ownership of node A is not changed;

Repeat the node transfer steps until the community ownership of all nodes does not change, the node transfer between communities is over, and the first stage is completed;

2) Reconstruct the graph in the second stage, and reconstruct the community formed after the first stage into a new node. The weight of the edges between the new nodes is the sum of the weights of the edges between the original communities, and the loop from the new node to its own. The weight of is the sum of the weights of edges between nodes in the original community;

For the reconstructed graph, iterate the first and second stages until the modularity of the entire graph no longer changes, and save the community information to which each node belongs;

According to the community information of each node, determine the final community ownership of each node, and complete the community discovery.

S7. Visualize the collaborative community of researchers in the tobacco field. In the S6 community discovery result data, the size of the nodes representing the influence of researchers and the thickness data of the edges representing the closeness of the cooperative relationship between researchers are subjected to curve normalization to ensure that the weights of data nodes and relationship edges are evenly distributed in the visual graph , the logarithm method is used to process the original data, and the process is as follows:

Processing for node size value: S _new =log ₁₀ S;

Processing for edge thickness width value: W _new =log ₁₀ W;

That is, logarithmic processing is performed on the obtained raw data of node size and edge thickness.

Each node in the visualization graph represents a researcher, the size of the node represents the influence of the researcher in the field, the larger the node, the greater the influence; the width of the connection between the nodes reflects the cooperation between the two researchers Frequency, that is, the weight value of cooperation degree. The higher the cooperation frequency, the wider the connection.

Specifically, the cooperative community visualization is implemented based on the Echarts visualization framework. By loading the cooperative community JSON data, the front-end visualization of the cooperative community graph is realized. An example of the jason data format is as follows:

Among them: nodes represents the researcher (id is the unique identifier of the researcher, label represents the name, group represents the community group number, size represents the size of the node in the community); links represents the relationship edge (source represents the unique identifier of the starting node of the edge, target represents The terminal node of the edge is uniquely identified, and the weight represents the width of the edge).

The visual map of the tobacco scientific research cooperative community finally generated according to the method of the present invention is shown in Figure 4, each node represents a scientific researcher, the size of the node represents the influence of the scientific researcher in the field of tobacco technology, and the connection between nodes The thickness of the line and the connection indicates the frequency of cooperation. The higher the frequency, the higher the degree of cooperation. The researchers with high degree of cooperation gather together to form a scientific community. The researchers in the community have a close cooperative relationship and have similar research directions and research fields. Through the cooperation relationship, it can be found that the higher the cooperation frequency between the two researchers, the closer the cooperation relationship between the two researchers.

In the cooperative community map in the field of tobacco research, the width of the relationship edge can determine the degree of close cooperation between two researchers, and the analysis of the frequency (and time series parameters) community clustering pattern of a group of related researchers' cooperation relationships can reveal the prominent relationship between researchers. The relationship link, researchers in the same community can think that there are certain similarities in academic research, which provides support for the recommendation of partners, and at the same time plays an auxiliary role in the discovery of scientific research teams in different tobacco institutions.

Identify the research topics of each community, such as extracting the keywords of the papers published by the researchers in the cooperative community, and count the word frequencies of the keywords, and select the top N (such as the top 3) with the highest word frequency as the research topics of the cooperative community. , and add research topic tags to the corresponding communities of the graph to increase the readability of the graph and help readers identify the research topics of each community.

Device Example:

A device for constructing a cooperative community of scientific researchers of the present invention includes a memory, a processor and a data interface, and the processor can obtain scientific research data in tobacco or other technical fields through the data interface, including information on projects, achievements and technical personnel, etc. And execute the program stored in the memory to implement a method for constructing a scientific researcher cooperative community of the present invention. The method of the present invention has been described clearly enough in the method embodiments, and will not be repeated here.

Claims (10)

1. a construction method of a cooperative community of scientific researchers, is characterized in that, comprises the steps: 1) Obtaining cooperation data, the cooperation data includes scientific research projects and scientific researchers participating in corresponding scientific research projects, as well as scientific research results and scientific research personnel who produce corresponding scientific research results; the scientific research results include papers; 2) Establish the influence model of researchers and calculate the influence of researchers; 3) Screening the cooperation data, the screening includes: deleting the scientific researcher whose influence is lower than the influence setting value from the cooperation data; 4) According to the cooperation data after screening, generate a network of scientific researcher cooperation relationship, the scientific researcher cooperation relationship network includes: the number of cooperation between each scientific researcher and other scientific researchers, and the number of cooperation times refers to the corresponding two scientific research personnel. The number of times the staff has participated in the same scientific research project or produced the same scientific research results; 5) Using the community discovery algorithm, based on the cooperative relationship network of researchers and the influence of researchers, a community map of researchers' cooperation is generated. 2. the construction method of scientific research personnel cooperative community according to claim 1, is characterized in that, described influence setting value N is:

Among them, N _max is the highest value of the influence of scientific researchers in the cooperation data. 3. the construction method of the cooperative community of scientific researchers according to claim 1 or 2, is characterized in that, to filter the cooperative relationship network of scientific researchers described in step 4), and described filtering comprises, the number of times of cooperation is lower than the number of times of cooperation The number of collaborations between researchers at the set value is set to 0. 4. the construction method of scientific research personnel cooperative community according to claim 3, is characterized in that, described cooperation times setting value M is:

Among them, M _max is the highest value of the number of collaborations between researchers in the collaboration data. 5. The method for constructing a cooperative community of scientific researchers according to claim 4, wherein, in the influence model of scientific researchers, the influence of scientific researchers is determined at least by the number of scientific research outputs, and the scientific research outputs include the following categories : Papers, patents. 6. The construction method of scientific research personnel cooperative community according to claim 5, is characterized in that, described scientific research personnel influence model is:

Among them, P is the influence, n is the number of scientific research output of the researcher, S _i is the set score of the category to which the scientific researcher's ith scientific research output belongs, and Wi is the scientific researcher's _ith scientific research output. Out of the set weights in the category of scientific research output. 7. The method for constructing a cooperative community of scientific researchers according to claim 6, wherein in step 4), when generating the cooperative relationship network of scientific researchers, for the scientific research project earlier than the set time, any two of them. The cooperation between researchers who participated in this scientific research project is recorded as a times, 0<a<1. 8. The method for constructing a cooperative community of scientific researchers according to claim 7, wherein the community discovery algorithm is a louvain algorithm. 9. The method for constructing a scientific researcher cooperative community according to claim 8, wherein each community in the scientific researcher cooperative community graph is added with a subject tag, and the subject tag is each node in the corresponding community as a node. In the papers published by researchers, the set number of paper keywords with the highest word frequency. 10. An apparatus for constructing a cooperative community of scientific researchers, comprising a processor and a memory, wherein the processor executes instructions stored in the memory, so as to realize the scientific research personnel according to any one of claims 1 to 9 How to build cooperative communities.

Images