Scientific researcher performance evaluation method, device and equipment and storage medium

Abstract

The invention discloses a researcher performance evaluation method, device and equipment and a storage medium, and relates to the field of human resource management, and the method comprises the steps: determining performance data through the performance information of researchers and an integral rule; determining a first target performance vector by using the performance data, and determining a second target performance vector based on the first target performance vector; determining a first clustering center and a sample point from the second target performance vector, determining a second clustering center based on the distance probability distribution between the first clustering center and the sample point, and determining the first clustering center and the second clustering center as a current clustering center; distributing the positions of the sample points by using the current clustering center, iterating the current clustering center, and adjusting the positions of the sample points in the iteration process to obtain a target clustering model; and performing performance evaluation on the to-be-evaluated performance data by using the target clustering model to obtain a corresponding evaluation result. Therefore, the method can improve the performance evaluation accuracy of the researchers.

Description

Technical Field

[0001] The present invention relates to the field of human resource management, and particularly to a method, device, equipment and storage medium for evaluating the performance of scientific research personnel. Background Art

[0002] At present, with the booming development of the knowledge economy and the rapid iteration of technological innovation, it is of great significance to scientifically and reasonably evaluate the performance of scientific research organizations and personnel for the sustainable development of enterprises.

[0003] However, there are many drawbacks in the current performance evaluation of teams or members in scientific research units. Specifically, the performance evaluation system lacks systematicness and is not detailed enough. In most cases, there is no corresponding professional system as support. This directly leads to inaccurate and one-sided evaluation results, and the workload in the evaluation process is huge. In the manual operation mode, the interference of human factors is serious, and the reliability and efficiency of the evaluation far cannot meet the actual needs.

[0004] Therefore, how to improve the accuracy of performance evaluation of scientific research personnel is a technical problem to be solved urgently at present. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method, device, equipment and storage medium for evaluating the performance of scientific research personnel, which can improve the accuracy of performance evaluation of scientific research personnel. The specific solutions are as follows:

[0006] In a first aspect, the present application provides a method for evaluating the performance of scientific research personnel, including:

[0007] Determine corresponding integral rules based on the performance types of scientific research personnel, and use the performance information of the scientific research personnel and the integral rules to determine performance data;

[0008] Use the performance data to determine a first target performance vector, reduce the dimension of the first target performance vector based on an autoencoder, and perform normalization processing on the obtained reduced-dimension vector to obtain a second target performance vector;

[0009] Determine a first clustering center and sample points from the second target performance vector, and determine a second clustering center based on the distance probability distribution between the first clustering center and the sample points through the K-Means++ algorithm, and determine the first clustering center and the second clustering center as the current clustering center;

[0010] Use the current clustering center to assign the positions of the sample points, iterate the current clustering center, and adjust the positions of the sample points during the iteration process to obtain a target clustering model;

[0011] Determine the performance data to be evaluated according to the performance appraisal requirements, and perform performance evaluation on the performance data to be evaluated by using the target clustering model, the preset appraisal range, and the preset number of clusters to obtain corresponding evaluation results.

[0012] Optionally, the determining the first target performance vector by using the performance data includes:

[0013] Determine the initial performance vector of the scientific research personnel and the weight vector of the position corresponding to the scientific research personnel;

[0014] Multiply the initial performance vector and the weight vector, and use the obtained product result and the Sigmoid activation function to determine the first target performance vector.

[0015] Optionally, the dimension reduction of the first target performance vector based on the autoencoder and the normalization processing of the obtained dimension-reduced vector to obtain the second target performance vector includes:

[0016] Configure the hidden layer of the autoencoder based on the preset dimension attenuation rule to obtain the configured autoencoder;

[0017] Reduce the dimension of the first target performance vector through the configured autoencoder based on the preset non-linear mapping function to obtain the dimension-reduced vector;

[0018] Determine the mean vector and the standard deviation vector of the dimension-reduced vector, and perform normalization processing on the dimension-reduced vector based on the mean vector, the standard deviation vector, and the preset constant to obtain the second target performance vector.

[0019] Optionally, the determining the second cluster center by the K-Means++ algorithm based on the distance probability distribution between the first cluster center and the sample points includes:

[0020] Determine the distance between the first cluster center and the sample points by the K-Means++ algorithm based on the first cluster center, the sample points, the weight vector, and the preset dimension index;

[0021] Determine the distance probability distribution based on the square value of the distance and the two-norm of the weight vector, and use the distance probability distribution to determine the second cluster center.

[0022] Optionally, the using the current cluster center to allocate the positions of the sample points, iterating the current cluster center, and adjusting the positions of the sample points during the iteration to obtain the target clustering model includes:

[0023] Allocate the positions of the sample points based on the current clustering centers, the weighted means of the job categories, the sample points, and a preset weight adjustment coefficient, and iterate on the current clustering centers;

[0024] During the iteration process, adjust the positions of the sample points until the number of iterations is equal to the preset number of iterations or the offset of the current clustering center is not greater than the preset offset threshold to end the iteration, and then end the iteration and determine the corresponding target clustering model.

[0025] Optionally, the determining the performance data to be evaluated according to the performance appraisal requirements includes:

[0026] Use natural language processing technology to determine the target researcher category and performance appraisal indicators from the performance appraisal requirements;

[0027] Determine the performance data to be evaluated from a preset performance database through a graph neural network algorithm and a time series analysis method based on the target researcher category and the performance appraisal indicators.

[0028] Optionally, the using the target clustering model, a preset appraisal range, and a preset number of clusters to perform performance appraisal on the performance data to be evaluated to obtain corresponding evaluation results includes:

[0029] Adjust the parameters of the target clustering model based on the data scale and distribution characteristics of the performance data to be evaluated to obtain an adjusted model;

[0030] Determine the performance data that meets the preset range conditions from the performance data to be evaluated based on the preset appraisal range;

[0031] Perform performance appraisal on the performance data that meets the preset range conditions based on the preset number of clusters through the adjusted model and obtain corresponding evaluation results.

[0032] In a second aspect, the present application provides a scientific research personnel performance appraisal device, including:

[0033] A data determination module, configured to determine corresponding point rules based on the performance types of scientific research personnel, and use the performance information of the scientific research personnel and the point rules to determine performance data;

[0034] A vector determination module, configured to use the performance data to determine a first target performance vector, reduce the dimension of the first target performance vector based on an autoencoder, and perform normalization processing on the obtained vector after dimension reduction to obtain a second target performance vector;

[0035] A clustering center determination module, configured to determine a first clustering center and sample points from the second target performance vectors, and determine a second clustering center based on the distance probability distribution between the first clustering center and the sample points through the K-Means++ algorithm, and determine the first clustering center and the second clustering center as the current clustering center;

[0036] A model determination module, configured to allocate the positions of the sample points by using the current clustering center, iterate the current clustering center, and adjust the positions of the sample points during the iteration process to obtain a target clustering model;

[0037] A performance evaluation module, configured to determine performance data to be evaluated according to performance evaluation requirements, and perform performance evaluation on the performance data to be evaluated by using the target clustering model, a preset evaluation range, and a preset number of clusters to obtain corresponding evaluation results.

[0038] In a third aspect, the present application provides an electronic device, including:

[0039] A memory, configured to store a computer program;

[0040] A processor, configured to execute the computer program to implement the foregoing scientific research personnel performance evaluation method.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium, configured to store a computer program; wherein, when the computer program is executed by a processor, the foregoing scientific research personnel performance evaluation method is implemented.

[0042] In this application, the corresponding point - based rules are determined based on the performance types of scientific research personnel, and the performance data is determined by using the performance information of the scientific research personnel and the point - based rules; the first target performance vector is determined by using the performance data, the first target performance vector is dimension - reduced based on an auto - encoder, and the dimension - reduced vector obtained is normalized to obtain the second target performance vector; the first clustering center and sample points are determined from the second target performance vector, and the second clustering center is determined based on the distance probability distribution between the first clustering center and the sample points through the K - Means++ algorithm, and the first clustering center and the second clustering center are determined as the current clustering centers; the positions of the sample points are assigned by using the current clustering centers, the current clustering centers are iterated, and the positions of the sample points are adjusted during the iteration process to obtain the target clustering model; the performance data to be evaluated is determined according to the performance assessment requirements, and the performance assessment of the performance data to be evaluated is carried out by using the target clustering model, the preset assessment range, and the preset number of clusters to obtain the corresponding assessment results. As can be seen from the above, in this application, the point - based rules are first determined based on the performance types of scientific research personnel, and then the performance data is obtained by combining the performance information of scientific research personnel with the point - based rules. Then, the first target performance vector is determined by using the performance data, it is dimension - reduced through an auto - encoder, and the dimension - reduced vector is normalized to obtain the second target performance vector. Then, the first clustering center and sample points are determined from the second target performance vector, the second clustering center is determined based on the distance probability distribution between the first clustering center and the sample points by using the K - Means++ algorithm, and the two are determined as the current clustering centers. After that, the positions of the sample points are assigned by using the current clustering centers, the current clustering centers are iterated, and the positions of the sample points are adjusted to obtain the target clustering model. Finally, the performance data to be evaluated is determined according to the performance assessment requirements, and it is evaluated by using the target clustering model, the preset assessment range, and the preset number of clusters to obtain the assessment results. In this way, this application can improve the accuracy of the performance assessment of scientific research personnel, thereby providing a certain scientific basis for enterprises to accurately identify talents and optimize resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to the provided drawings without creative efforts.

[0044] Figure 1 It is a flowchart of a method for evaluating the performance of scientific research personnel disclosed in this application;

[0045] Figure 2 It is a flowchart of a specific method for evaluating the performance of scientific research personnel disclosed in this application;

[0046] Figure 3 Structural schematic diagram of a performance evaluation device for scientific research personnel disclosed in this application;

[0047] Figure 4 Structural diagram of an electronic device disclosed in this application. Specific implementation manners

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] Currently, there are many drawbacks in the performance evaluation of teams or members in scientific research institutions. Specifically, the performance evaluation system lacks systematicness and is not detailed enough. In most cases, there is no corresponding professional system as support. This directly leads to inaccurate and one-sided evaluation results, and the workload in the evaluation process is huge. In the manual operation mode, the interference of human factors is serious, and the reliability and efficiency of the evaluation far cannot meet the actual needs. For this reason, this application provides a performance evaluation method, device, equipment, and storage medium for scientific research personnel, which can improve the accuracy of performance evaluation of scientific research personnel.

[0050] See Figure 1 As shown, an embodiment of the present invention discloses a performance evaluation method for scientific research personnel, including:

[0051] Step S11: Determine corresponding integral rules based on the performance types of scientific research personnel, and determine performance data by using the performance information of the scientific research personnel and the integral rules.

[0052] See Figure 1 As shown, an embodiment of the present invention discloses a performance evaluation method for scientific research personnel, including:

[0053] Step S11: Determine corresponding integral rules based on the performance types of scientific research personnel, and determine performance data by using the performance information of the scientific research personnel and the integral rules.

[0054] In this embodiment, first, the performance of scientific research personnel is classified. Their performance can cover multiple dimensions such as scientific research achievements, intellectual property rights, knowledge contributions, participation in projects and contributions, and team honors. For scientific research achievements, integral rules are set according to parameters such as the awarding entity, award level, ranking of participants, and award time. For example, a scientific research achievement that wins the first prize at the national level and ranks first is given a higher score; a scientific research achievement that wins the third prize at the provincial level and ranks lower is given a lower score. In terms of intellectual property rights, integral rules are formulated according to patent types, proprietary technology types, the level of the journals in which papers and works are published, the importance of software copyrights, and standard levels. For example, papers published in core journals and papers published in ordinary journals correspond to different scores. For knowledge contributions, integral rules are determined according to their scope of influence, importance, application effects, etc. For example, knowledge achievements that are widely applied within an enterprise and produce significant benefits are given higher scores. For the contributions of team members to projects, the overall completion of the project, participation roles, team collaboration, innovation level and achievements, and work attitude are comprehensively considered to determine the scores.

[0055] Subsequently, the performance information of scientific research personnel is collected, including employee numbers, positions, affiliated teams, integral types, integral occurrence times, etc. These performance information are matched and calculated with the corresponding integral rules to obtain performance data including personnel, the organization they belong to, positions, integral categories, integral quantities, integral times, etc., forming the basic data for personnel performance evaluation and laying a foundation for subsequent performance evaluations.

[0056] In addition, it should be noted that although this application takes personnel as the core object of expression, since a scientific research organization is essentially composed of personnel, its performance is essentially the comprehensive reflection and organic integration of the performance of personnel within the organization. Therefore, the implementation logic of this application is also applicable to the organizational performance evaluation in the scientific research field.

[0057] Step S12: Determine the first target performance vector using the performance data, reduce the dimension of the first target performance vector based on an autoencoder, and normalize the obtained vector after dimensionality reduction to obtain a second target performance vector.

[0058] In this embodiment, after obtaining the performance data of scientific research personnel, it is necessary to further determine the first target performance vector. Specifically, it is necessary to first determine the initial performance vector of the scientific research personnel and the weight vector of the position corresponding to this scientific research personnel.

[0059] Among them, the initial performance vector is a quantitative representation of the performance indicators of scientific researchers. The performance data includes the points of scientific researchers in various performance indicator fields, such as points in aspects like scientific research achievements, intellectual property rights, knowledge contributions, participation in projects and contributions, etc. Arranging these points in a certain order can form the initial performance vector of scientific researchers. The weight vector of the position reflects the degree of emphasis on each performance indicator for different positions. Different R & D positions, such as product process development, basic research, experimental operation, etc., have different focuses on performance assessment. Therefore, the performance management department will formulate corresponding performance indicator weight coefficients for each position according to management requirements. Arranging these weight coefficients in the same order as the initial performance vector can obtain the weight vector of the position.

[0060] After determining the initial performance vector and the weight vector of the position, multiply their corresponding elements to obtain a new vector. Each element of this new vector is the product of the corresponding element in the initial performance vector and the corresponding element in the position weight vector. Then, use the Sigmoid activation function to process this new vector. The Sigmoid activation function can map the input value to the range of (0, 1), making the elements of the processed vector have better interpretability and comparability. The vector obtained after being processed by the Sigmoid activation function is the first target performance vector.

[0061] Furthermore, after obtaining the first target performance vector, since its dimension may be relatively high, it will increase the complexity of subsequent calculations and there may be data redundancy. Therefore, it is necessary to perform dimensionality reduction processing on the first target performance vector. Here, an autoencoder is used to complete the vector dimensionality reduction operation. In this embodiment, it is necessary to configure the hidden layer of the autoencoder based on a preset dimensionality attenuation rule to obtain the configured autoencoder. Among them, the preset dimensionality attenuation rule means that the hidden layer dimension is configured according to the exponential attenuation rule.

[0062] After configuring the autoencoder, the first target performance vector is dimensionally reduced through the configured autoencoder based on a preset non - linear mapping function. The preset non - linear mapping function can map the high - dimensional feature vector to the low - dimensional feature vector space. In this process, the autoencoder will learn the low - dimensional representation of the input data, enabling the low - dimensional vector to retain as much important information of the high - dimensional vector as possible. And after dimensionally reducing the first target performance vector, a dimensionally reduced vector is obtained. In order to further improve the data quality, it is necessary to perform normalization processing on the dimensionally reduced vector to obtain the second target performance vector. Specifically, it is necessary to determine the mean vector and the standard deviation vector of the dimensionally reduced vector, and then perform normalization processing on the dimensionally reduced vector based on the mean vector, the standard deviation vector, and a preset constant.

[0063] Step S13: Determine the first cluster center and sample points from the second target performance vector, and based on the distance probability distribution between the first cluster center and the sample points, determine the second cluster center through the K-Means++ algorithm, and determine the first cluster center and the second cluster center as the current cluster centers.

[0064] In this embodiment, after obtaining the second target performance vector, it is necessary to select the first cluster center and sample points therefrom. The second target performance vector is the data after dimensionality reduction and normalization processing. It contains the key information of the scientific research personnel's performance and has a low dimension, which is suitable for clustering analysis. The sample points are the vectors in the second target performance vector, and they represent the performance characteristics of each scientific research personnel.

[0065] It should be noted that for the selection of the first cluster center, a random selection method can be adopted. Randomly select a vector from the sample set composed of the second target performance vector as the first cluster center. Although this random selection has a certain degree of randomness, since the cluster centers will be continuously optimized through the K-Means++ algorithm later, the initial random selection will not have too much impact on the final result.

[0066] After determining the first cluster center and sample points, next, through the K-Means++ algorithm, determine the distance between the first cluster center and the sample points based on the first cluster center, sample points, weight vector, and preset dimension index. Among them, the preset dimension index is used to specify the dimension involved in calculating the distance. And when calculating the distance, the Euclidean distance calculation method is improved by combining the position weight to calculate the distance between the first cluster center and each sample point.

[0067] Then, based on the squared values of these distances and the two-norm of the weight vector, determine the distance probability distribution. Among them, the squared value of the distance reflects the squared relationship of the difference degree between the sample point and the first cluster center, and the two-norm of the weight vector is a measure of the weight vector. By combining these two, a reasonable distance probability distribution can be obtained. Specifically, the probability that a sample point is selected as the next cluster center is proportional to the square of its shortest distance to the selected cluster center, that is, the farther the sample point is from the selected cluster center, the greater the probability of being selected as the next cluster center. The purpose of doing this is to make the newly selected cluster center able to cover the sample space as much as possible and avoid the concentration of cluster centers. Then, use this distance probability distribution to determine the second cluster center, and a sample point can be selected as the second cluster center from the sample points by random sampling according to the probability determined by the distance probability distribution.

[0068] Finally, determine the first clustering center and the second clustering center as the current clustering centers. These two clustering centers will serve as the initial conditions for subsequent iterations of the K-Means algorithm. By continuously iterating and updating the clustering centers and the belonging of sample points, effective clustering of the performance of scientific research personnel is achieved.

[0069] Step S14: Use the current clustering centers to assign the positions of the sample points, iterate the current clustering centers, and adjust the positions of the sample points during the iteration process to obtain the target clustering model.

[0070] In this embodiment, after determining the current clustering centers, first, the positions of the sample points are assigned based on the current clustering centers, the weighted mean of the job categories, the sample points, and the preset weight adjustment coefficient, and at the same time, the current clustering centers are iterated.

[0071] During the iteration process, the sample points are assigned to the category represented by the clustering center with the closest weighted distance to them. Then, based on the sample points assigned to each clustering center, the current clustering centers are iteratively updated. The position of the new clustering center is the weighted average of all sample points in this category, calculated by element-wise multiplication combined with the job weights.

[0072] Moreover, the positions of the sample points are not fixed. As the clustering centers are updated, the distances from each sample point to each clustering center also change. Therefore, it is necessary to recalculate the weighted distances from the sample points to the new clustering centers and reassign the positions of the sample points according to the new distance results. This way of dynamically adjusting the positions of the sample points enables the clustering results to be continuously optimized and more accurately reflect the actual classification of the performance of scientific research personnel.

[0073] It should be noted that the iteration process will continue until the preset termination conditions are met. There are two termination conditions: one is that the number of iterations is equal to the preset number of iterations; the other is that the offset of the current clustering centers is not greater than the preset offset threshold. The preset number of iterations is to prevent the iteration process from proceeding infinitely and ensure that the algorithm can converge within a reasonable time. The preset offset threshold is an indicator to measure the stability of the clustering centers. When the offset of the clustering centers is very small, it means that the clustering results have tended to be stable, and further iteration will not bring significant improvement. Therefore, when the termination conditions are met, the iteration ends, and the obtained clustering results determine the corresponding target clustering model.

[0074] Step S15: Determine the performance data to be evaluated according to the performance assessment requirements, and use the target clustering model, the preset assessment range, and the preset number of clusters to perform performance assessment on the performance data to be evaluated to obtain the corresponding assessment results.

[0075] In this embodiment, the performance appraisal requirements are usually presented in the form of natural language. In order to extract key information from it, natural language processing technology is needed. First, the performance appraisal requirement text is preprocessed, including removing stop words, stem extraction, part-of-speech tagging and other operations to improve the quality and processability of the text. Then, named entity recognition technology is used to identify the target scientific researcher categories in the text, such as "basic research scientific researcher", "application development scientific researcher", etc. At the same time, the keyword extraction algorithm is used to determine the performance appraisal indicators, such as "number of scientific research results", "paper publication impact factor", "project completion progress", etc.

[0076] After clarifying the target scientific researcher categories and performance appraisal indicators, the performance data to be evaluated should be screened out from the preset performance database. Here, the graph neural network algorithm and time series analysis method are used. The preset performance database stores a large amount of performance data of scientific researchers. There are complex correlations between these data, which are suitable for representation by graph structures, and graph neural networks can effectively process graph structure data and mine potential patterns and relationships in the data. The time series analysis method is used to consider the time dimension information of performance data. The performance of scientific researchers may change over time, so it is necessary to analyze the time trend and periodicity of performance data. Therefore, combined with the graph neural network algorithm and time series analysis method, and based on the target scientific researcher categories and performance appraisal indicators, the performance data to be evaluated is screened out from the preset performance database.

[0077] Furthermore, after obtaining the performance data to be evaluated, the parameters of the target clustering model need to be adjusted according to its data scale and distribution characteristics. The data scale will affect the efficiency and accuracy of clustering. If the data scale is large, it may be necessary to adjust the model's parameters such as the number of iterations and learning rate to speed up convergence; if the data scale is small, it is necessary to prevent overfitting.

[0078] In addition, the distribution characteristics of the data are also important considerations. Different performance data may present different distributions, such as normal distribution, skewed distribution, etc. Therefore, according to the distribution of the data, the distance measurement method of the clustering model, the cluster center initialization method and other parameters are adjusted to improve the adaptability of the model to the data. After the parameters are adjusted, the adjusted model is obtained. In addition, the preset assessment scope stipulates the specific conditions and boundaries of the performance evaluation. Therefore, the performance data that meets the preset range conditions is selected from the performance data to be evaluated.

[0079] Finally, the adjusted model is used to perform performance evaluation on the performance data that meets the preset range conditions based on the preset number of clusters. Among them, the preset number of clusters determines how many categories the performance data is divided into, and different numbers of clusters will result in different evaluation results. The evaluation results can be presented in various forms, such as visual charts, statistical reports, etc. Specifically, the visual chart can intuitively display the distribution of different clusters and performance characteristics, while the statistical report can list in detail the relevant indicators and data of each cluster. By analyzing the evaluation results, managers can understand the performance status of scientific research personnel and provide a basis for subsequent decisions, such as job adjustments and training plan formulation.

[0080] As can be seen from the above, in this application, the integral rule is first determined based on the performance type of scientific research personnel, and then the performance data is obtained by combining the performance information of scientific research personnel with the integral rule. Next, the first target performance vector is determined using the performance data, its dimension is reduced by an autoencoder, and the second target performance vector is obtained through normalization processing of the reduced-dimensional vector. Then, the first cluster center and sample points are determined from the second target performance vector, and the second cluster center is determined using the K-Means++ algorithm according to the distance probability distribution between the first cluster center and the sample points, and the two are set as the current cluster centers. After that, the positions of the sample points are assigned using the current cluster centers, the current cluster centers are iterated, and the positions of the sample points are adjusted to obtain the target clustering model. Finally, the performance data to be evaluated is determined according to the performance appraisal requirements, and it is evaluated using the target clustering model, the preset appraisal range, and the preset number of clusters to obtain the evaluation results. In this way, this application can improve the accuracy of performance evaluation of scientific research personnel, thereby providing a certain scientific basis for enterprises to accurately identify talents and optimize resource allocation.

[0081] The following combines Figure 2 the schematic diagram shown to specifically illustrate the technical solution of the embodiment of this application.

[0082] Specifically, first, the corresponding integral rule (i.e., the performance evaluation rule) is determined based on the performance type of scientific research personnel, and then the basic data of personnel performance evaluation (i.e., the integral information of employee classification evaluation) is collected, including the integral situation of organizations or personnel in each performance index field in the performance evaluation database, including personnel, the organization they belong to, positions, integral categories, integral quantities, integral times, matching rules, detailed situations, etc., as well as the corresponding historical performance appraisal results, to form a personnel performance evaluation database; the performance management department proposes the weight coefficients of relevant performance appraisal indicators according to different R & D positions to form the characteristic vector of the position, and its main purpose is to better reflect the performance management requirements of specific R & D positions, distinguish the focus of performance appraisal for different positions such as product process development, basic research, and experimental operations, and make the performance evaluation more accurate and targeted. In addition, the basic data of personnel performance evaluation will be stored in the employee performance integral database (i.e., the preset performance database).

[0083] According to the performance management requirements, preprocess the personnel performance data, including data cleaning, assigning different weights to personnel in different R & D positions, and data standardization or normalization. Data cleaning mainly eliminates redundant and incorrect data, processes abnormal data, and screens the data to select the personnel performance data of a certain period or a specific organization; calculate the personnel performance data with the weight feature vector of the position to obtain the performance index details of personnel in different positions; perform normalization processing on the data to scale the data to the range of (0, 1). The specific process is as follows: Define the personnel performance vector as , and the position weight vector as . Multiply the elements of the two vectors x and w, and use the sigmoid activation function to constrain it within the range of (0, 1) to obtain the performance feature vector (i.e., the first target performance vector). Among them, the formula for calculating the performance feature vector is as follows:

[0084] ;

[0085] In the formula, is the sigmoid activation function.

[0086] Then, use the autoencoder to reduce the dimension of the obtained performance feature vector based on the non-linear mapping function, so as to obtain the low-dimensional vector (i.e., the vector after dimensionality reduction). And, the autoencoder for dimensionality reduction operation adopts a symmetric network structure, and its hidden layer dimension is configured according to the exponential decay rule.

[0087] Next, perform normalization processing on the low-dimensional vector output by the autoencoder. The formula for normalization processing is as follows:

[0088] ;

[0089] In the formula, is the normalized vector (i.e., the second target performance vector), is the low-dimensional vector, and are the mean vector and standard deviation vector of all samples, and . Then, determine the randomly selected initial clustering center (i.e., the first clustering center) and sample points from the normalized vector. And the clustering center (i.e., the second clustering center) in the subsequent process is determined by the probability distribution of the distance between the initial clustering center and the sample points. Among them, the calculation formula for the probability distribution of the distance is as follows:

[0090] ;

[0091] In the formula, is the position weight, represents The L2 norm of For The distance to the nearest cluster center, express The square of the distance to its nearest cluster center, and the preset dimension index is used when calculating the distance, p is the index of the traversal sample, is the pth sample point after normalization.

[0092] Then, the positions of the sample points are assigned according to the determined cluster centers, and the cluster centers are iterated. Then, the positions of the sample points are iterated according to the cluster centers until the number of iterations is 100 or the cluster center offset is less than or equal to Stop the iteration, and determine the corresponding clustering model (that is, the improved deep embedding K-Means clustering model) according to the obtained clustering results. And determine the performance data to be evaluated from the employee performance points database according to the performance appraisal requirements, and use the clustering model, the preset appraisal scope and the preset number of clusters to perform performance appraisal on the performance data to be evaluated, or perform cluster analysis on the performance data to be evaluated, so as to obtain the corresponding evaluation results, and the evaluation results can be presented in the form of performance reports.

[0093] Accordingly, see Figure 3 As shown, the embodiment of the present application provides a scientific researcher performance evaluation device, comprising:

[0094] A data determination module 11 is used to determine a corresponding scoring rule based on the performance type of the scientific researcher, and to determine performance data using the performance information of the scientific researcher and the scoring rule;

[0095] A vector determination module 12 is used to determine a first target performance vector using the performance data, reduce the dimension of the first target performance vector based on an autoencoder, and normalize the obtained reduced-dimensional vector to obtain a second target performance vector;

[0096] A cluster center determination module 13 is used to determine a first cluster center and a sample point from the second target performance vector, and determine a second cluster center based on a distance probability distribution between the first cluster center and the sample point by using a K-Means++ algorithm, and determine the first cluster center and the second cluster center as current cluster centers;

[0097] A model determination module 14, configured to use the current cluster center to allocate the position of the sample point, iterate the current cluster center, and adjust the position of the sample point during the iteration process to obtain a target cluster model;

[0098] The performance evaluation module 15 is used to determine the performance data to be evaluated according to the performance evaluation requirements, and perform performance evaluation on the performance data to be evaluated by using the target clustering model, the preset evaluation range, and the preset number of clusters to obtain corresponding evaluation results.

[0099] As can be seen from the above, in this application, the integral rules are first determined based on the performance types of scientific research personnel, and then the performance data is obtained by combining the performance information of scientific research personnel and the integral rules. Next, the first target performance vector is determined by using the performance data, its dimension is reduced by an autoencoder, and the reduced vector is normalized to obtain the second target performance vector. Then, the first cluster center and sample points are determined from the second target performance vector, and the second cluster center is determined by using the K-Means++ algorithm according to the distance probability distribution between the first cluster center and the sample points, and the two are set as the current cluster centers. After that, the positions of the sample points are assigned by using the current cluster centers, the current cluster centers are iterated, and the positions of the sample points are adjusted to obtain the target clustering model. Finally, the performance data to be evaluated is determined according to the performance evaluation requirements, and it is evaluated by using the target clustering model, the preset evaluation range, and the preset number of clusters to obtain the evaluation results. In this way, this application can improve the accuracy of performance evaluation of scientific research personnel, thereby providing a certain scientific basis for enterprises to accurately identify talents and optimize resource allocation.

[0100] In some specific embodiments, the vector determination module 12 specifically includes:

[0101] The weight vector determination unit is used to determine the initial performance vector of the scientific research personnel and the weight vector of the position corresponding to the scientific research personnel;

[0102] The performance vector determination unit is used to multiply the initial performance vector and the weight vector, and determine the first target performance vector by using the obtained product result and the Sigmoid activation function.

[0103] In some specific embodiments, the vector determination module 12 specifically includes:

[0104] The autoencoder configuration unit is used to configure the hidden layer of the autoencoder based on a preset dimension attenuation rule to obtain a configured autoencoder;

[0105] The vector dimension reduction unit is used to reduce the dimension of the first target performance vector by using the configured autoencoder based on a preset non-linear mapping function to obtain a reduced vector;

[0106] The vector normalization unit is used to determine the mean vector and the standard deviation vector of the reduced vector, and perform normalization processing on the reduced vector based on the mean vector, the standard deviation vector, and a preset constant to obtain the second target performance vector.

[0107] In some specific embodiments, the clustering center determination module 13 specifically includes:

[0108] A distance determination unit, configured to determine the distance between the first clustering center and the sample points based on the first clustering center, the sample points, the weight vector, and a preset dimension index through the K-Means++ algorithm;

[0109] A center determination unit, configured to determine a distance probability distribution based on the square value of the distance and the two-norm of the weight vector, and determine a second clustering center by using the distance probability distribution.

[0110] In some specific embodiments, the model determination module 14 specifically includes:

[0111] A center iteration unit, configured to allocate the positions of the sample points based on the current clustering center, the weighted mean of the job categories, the sample points, and a preset weight adjustment coefficient, and iterate on the current clustering center;

[0112] A model determination unit, configured to, during the iteration process, adjust the positions of the sample points until the iteration times is equal to the preset iteration times or the offset of the current clustering center is not greater than the preset offset threshold to end the iteration, and then end the iteration and determine the corresponding target clustering model.

[0113] In some specific embodiments, the performance evaluation module 15 specifically includes:

[0114] An information determination unit, configured to determine the target researcher category and performance evaluation indicators from the performance evaluation requirements by using natural language processing technology;

[0115] A first data determination unit, configured to determine the performance data to be evaluated from a preset performance database based on the target researcher category and the performance evaluation indicators through a graph neural network algorithm and a time series analysis method.

[0116] In some specific embodiments, the performance evaluation module 15 specifically includes:

[0117] A model adjustment unit, configured to adjust the parameters of the target clustering model based on the data scale and distribution characteristics of the performance data to be evaluated to obtain an adjusted model;

[0118] A second data determination unit, configured to determine the performance data meeting the preset range conditions from the performance data to be evaluated based on a preset evaluation range;

[0119] A data evaluation unit, configured to perform performance evaluation on the performance data meeting the preset range conditions based on a preset number of clusters through the adjusted model and obtain corresponding evaluation results.

[0120] Furthermore, the embodiment of the present application also discloses an electronic device. Figure 4 It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment. The content in the figure cannot be regarded as any limitation on the scope of use of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the scientific research personnel performance evaluation method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0121] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows is any communication protocol applicable to the technical solution of the present application, and specific limitations are not imposed here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application needs, and no specific limitations are made here.

[0122] In addition, as a carrier for resource storage, the memory 22 may be a read-only memory, a random access memory, a magnetic disk, or an optical disc, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.

[0123] Among them, the operating system 221 is used to manage and control each hardware device and the computer program 222 on the electronic device 20, and it may be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the scientific research personnel performance evaluation method executed by the electronic device 20 disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs that can be used to complete other specific tasks.

[0124] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the scientific research personnel performance evaluation method disclosed above is implemented. For the specific steps of this method, reference may be made to the corresponding content disclosed in the foregoing embodiments, and details are not repeated here.

[0125] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the apparatuses disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0126] Those skilled in the art can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner 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 to exceed the scope of this application.

[0127] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0128] Finally, it should also be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0129] The technical solutions provided in this application have been introduced in detail above. Specific examples are used herein to elaborate on the principles and implementation manners of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, based on the idea of this application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A performance evaluation method for scientific research personnel, characterized in that, Including: Determine corresponding integral rules based on the performance types of scientific research personnel, and use the performance information of the scientific research personnel and the integral rules to determine performance data; Use the performance data to determine a first target performance vector, reduce the dimension of the first target performance vector based on an autoencoder, and normalize the obtained vector after dimensionality reduction to obtain a second target performance vector; Determine a first cluster center and sample points from the second target performance vector, and determine a second cluster center based on the distance probability distribution between the first cluster center and the sample points through the K-Means++ algorithm, and determine the first cluster center and the second cluster center as the current cluster centers; Use the current cluster centers to allocate the positions of the sample points, iterate the current cluster centers, and adjust the positions of the sample points during the iteration to obtain a target clustering model; Determine the performance data to be evaluated according to the performance assessment requirements, and use the target clustering model, a preset assessment range, and a preset number of clusters to perform performance assessment on the performance data to be evaluated to obtain corresponding assessment results.

2. The scientific research personnel performance evaluation method according to claim 1, wherein The using the performance data to determine a first target performance vector includes: Determine the initial performance vector of the scientific research personnel and the weight vector of the position corresponding to the scientific research personnel; Multiply the initial performance vector and the weight vector, and use the obtained product result and the Sigmoid activation function to determine the first target performance vector.

3. The scientific research personnel performance evaluation method according to claim 1, characterized in that The reducing the dimension of the first target performance vector based on an autoencoder and normalizing the obtained vector after dimensionality reduction to obtain a second target performance vector includes: Configure the hidden layer of the autoencoder based on a preset dimension attenuation rule to obtain a configured autoencoder; Reduce the dimension of the first target performance vector through the configured autoencoder based on a preset non-linear mapping function to obtain a vector after dimensionality reduction; Determine the mean vector and standard deviation vector of the vector after dimensionality reduction, and normalize the vector after dimensionality reduction based on the mean vector, the standard deviation vector, and a preset constant to obtain a second target performance vector.

4. The scientific research personnel performance evaluation method according to claim 2, wherein, The determining a second cluster center based on the distance probability distribution between the first cluster center and the sample points through the K-Means++ algorithm includes: Determine the distance between the first cluster center and the sample points through the K-Means++ algorithm based on the first cluster center, the sample points, the weight vector, and a preset dimension index; Determine the distance probability distribution based on the squared value of the distance and the two-norm of the weight vector, and use the distance probability distribution to determine the second cluster center.

5. The scientific research personnel performance evaluation method according to claim 1, wherein The using the current cluster centers to allocate the positions of the sample points, iterate the current cluster centers, and adjust the positions of the sample points during the iteration to obtain a target clustering model includes: Allocate the positions of the sample points based on the current cluster centers, the weighted mean of the position categories, the sample points, and a preset weight adjustment coefficient, and iterate the current cluster centers; During the iteration process, adjust the positions of the sample points until the number of iterations is equal to the preset number of iterations or the offset of the current clustering center is not greater than the preset offset threshold, and then end the iteration and determine the corresponding target clustering model.

6. The scientific research personnel performance evaluation method according to claim 1, wherein, The determining the performance data to be evaluated according to the performance appraisal requirements includes: Determine the target researcher category and performance appraisal indicators from the performance appraisal requirements by using natural language processing technology; Determine the performance data to be evaluated from the preset performance database through the graph neural network algorithm and time series analysis method and based on the target researcher category and the performance appraisal indicators.

7. The scientific research personnel performance evaluation method according to any one of claims 1 to 6, characterized in that The using the target clustering model, preset appraisal range, and preset number of clusters to perform performance appraisal on the performance data to be evaluated to obtain the corresponding evaluation results includes: Adjust the parameters of the target clustering model based on the data scale and distribution characteristics of the performance data to be evaluated to obtain an adjusted model; Determine the performance data that meets the preset range conditions from the performance data to be evaluated based on the preset appraisal range; Perform performance appraisal on the performance data that meets the preset range conditions based on the preset number of clusters through the adjusted model and obtain the corresponding evaluation results.

8. A performance evaluation device for scientific research personnel, characterized in that, Includes: A data determination module, configured to determine the corresponding integral rules based on the performance types of the researchers, and determine the performance data by using the performance information of the researchers and the integral rules; A vector determination module, configured to determine the first target performance vector by using the performance data, reduce the dimension of the first target performance vector based on the autoencoder, and perform normalization processing on the obtained vector after dimension reduction to obtain the second target performance vector; A clustering center determination module, configured to determine the first clustering center and sample points from the second target performance vector, and determine the second clustering center based on the distance probability distribution between the first clustering center and the sample points through the K-Means++ algorithm, and determine the first clustering center and the second clustering center as the current clustering center; A model determination module, configured to allocate the positions of the sample points by using the current clustering center, iterate the current clustering center, and adjust the positions of the sample points during the iteration process to obtain the target clustering model; A performance appraisal module, configured to determine the performance data to be evaluated according to the performance appraisal requirements, and perform performance appraisal on the performance data to be evaluated by using the target clustering model, preset appraisal range, and preset number of clusters to obtain the corresponding evaluation results.

9. An electronic device, characterized in that, Includes: A memory, configured to store a computer program; A processor, configured to execute the computer program to implement the researcher performance appraisal method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, For storing a computer program; wherein, when the computer program is executed by the processor, it implements the researcher performance appraisal method according to any one of claims 1 to 7.

Images