Enterprise talent evaluation management method and system based on data mining

Abstract

The invention discloses an enterprise talent evaluation management method and system based on data mining. The method comprises the following steps: establishing a weight distribution model of posts on a five-dimensional evaluation system; inputting the feature vector required by the post to be evaluated and a candidate basic requirement prepared according to the five-dimensional evaluation system into the weight distribution model, and obtaining the weight of each dimension of the post to be evaluated on the five-dimensional evaluation system; collecting candidate data of a post to be evaluated, and performing scoring and weighted fusion to obtain a candidate comprehensive ability vector; the product sum of each component and the corresponding weight is the matching degree of the candidate and the post to be evaluated; and generating a recommendation list according to the matching degree sequence. The method has the advantages that the self-adaptive algorithm, semantic analysis and time sequence prediction technology are combined, different post and real-time requirements of an enterprise can be met, structured and unstructured data are integrated, growth trend prediction is combined, the accuracy of a talent evaluation result is remarkably improved, and support is provided for long-term human resource planning of the enterprise.

Description

Technical Field

[0001] The present invention relates to the technical field of human resource management, and is particularly applicable to an enterprise talent evaluation management method and system based on data mining. Background Art

[0002] In the existing human resource management system of enterprises, talent evaluation is an important part of recruitment, promotion, salary adjustment and performance management. However, the traditional talent evaluation method mainly relies on manual scoring, and the weight of each scoring item is fixed, that is, the traditional talent scoring system is fixed and unchanging. It cannot adjust the weight of each scoring item according to different positions or real-time needs of the enterprise, and it lacks dynamic adaptation capabilities, which often leads to inaccurate talent evaluation results.

[0003] At the same time, traditional talent evaluation is mostly based on the analysis of structured data such as education, professional title, and years of work experience, but less information mining is done on structured data such as performance reviews and project descriptions. In fact, each person's performance review and each person's description of the project they participated in contain a large amount of evaluation information about the talent's ability. Traditional talent evaluation lacks the use of unstructured data, resulting in the inability to comprehensively evaluate the talent's ability.

[0004] In addition, for enterprises, the evaluation of talents is not about their past abilities, but about predicting their future development potential through their past abilities. However, the current traditional talent evaluation system is mostly static evaluation, which can only reflect the current performance of talents, but cannot predict the future development trend of talents, and cannot achieve talent selection planning that supports the long-term development of enterprises through talent evaluation.

[0005] Therefore, there is an urgent need for a new evaluation method and system that can dynamically adjust the scoring weights of each scoring item, deeply mine multi-source data and combine growth trend predictions to meet the company's needs for accuracy, comprehensiveness and foresight in talent assessment. Summary of the invention

[0006] The present invention aims to provide a method and system for enterprise talent evaluation management based on data mining, aiming to solve the problems in current enterprise talent evaluation management that the weights of various scoring items cannot be adjusted according to positions or real-time needs, the lack of utilization of unstructured data, and the inability to predict the future development trend of talents.

[0007] To achieve the above object, the present invention adopts the following technical solutions: The enterprise talent evaluation management method and system based on data mining of the present invention comprises the following steps: S1, establish a five-dimensional talent evaluation system; S2, establish a weight distribution model for positions in the five-dimensional evaluation system; S3, input the characteristic vector of the job requirements to be evaluated and the basic requirements of the candidates formulated according to the 5-dimensional evaluation system into the weight allocation model to obtain the weight of each dimension of the job to be evaluated in the 5-dimensional evaluation system; S4, collect candidate data for the positions to be evaluated according to the five-dimensional evaluation system, including structured data and unstructured data; S5, score the structured data and unstructured data according to the five-dimensional evaluation system, and then perform weighted fusion to obtain the candidate's comprehensive evaluation vector; S6, using principal component analysis to reduce the candidate's comprehensive evaluation vector into three principal components to obtain the candidate's comprehensive ability vector; S7, the sum of the products of each component in the candidate's comprehensive ability vector and the corresponding weight is the matching degree between the candidate and the position to be evaluated; S8, generating a candidate recommendation list by sorting according to the matching degree.

[0008] Furthermore, the five-dimensional evaluation system described in step S1 includes five dimensions: educational background, project ability, performance level, professional skills and achievement honors. Each dimension includes several secondary evaluation items and scoring rules.

[0009] Furthermore, step S2 specifically includes: S2.1, use NLP algorithm to analyze job descriptions and generate feature vectors of job requirements through clustering; S2.2, collect historical talent evaluation data for various positions according to the five-dimensional evaluation system, including annual performance, competency, development potential, project achievements, project information, ability characteristics, educational background, project capabilities, performance level, professional skills and achievement honors; S2.3, based on the feature vectors of each job requirement and the historical talent evaluation data for each job, a job weight allocation model is established using the random forest algorithm.

[0010] Furthermore, the structured data refers to the educational background, project capabilities, performance level, professional skills and achievement honors that directly reflect the five-dimensional evaluation system; the unstructured data refers to the data that needs to be included in the five-dimensional evaluation system after word segmentation, denoising and feature extraction using natural language processing technology.

[0011] Furthermore, it also includes collecting historical talent evaluation data of several same people in the same position in time series, and using LSTM network training to obtain a talent future development trend prediction model to predict the future growth of candidates.

[0012] The enterprise talent evaluation management system based on data mining described in the present invention includes an adaptive weight allocation module, a multi-source data mining module, a growth trend analysis module, and an intelligent recommendation module; The adaptive weight allocation module, based on the 5-dimensional evaluation system, establishes a weight allocation model for a position through a random forest algorithm according to the position requirement feature vector and the historical talent evaluation data of the position; obtains the weight of each dimension of the position to be evaluated in the 5-dimensional evaluation system according to the weight allocation model, the feature vector of the position to be evaluated and the basic requirements of the candidates formulated according to the 5-dimensional evaluation system; The multi-source data mining module is used to collect structured and unstructured data of candidates for the positions to be evaluated by using natural language processing technology through word segmentation, denoising and feature extraction according to the five-dimensional evaluation system; The growth trend analysis module collects several historical talent evaluation data sets of the same person in the same position in time series, and uses LSTM network training to obtain a talent future development trend prediction model to predict the future growth of the candidate; The intelligent recommendation module scores structured data and unstructured data according to a five-dimensional evaluation system, performs weighted fusion, and uses principal component analysis to reduce the dimension to three principal components to obtain a comprehensive ability vector of the candidate; the sum of the products of each component and the corresponding weight is calculated as the matching degree between the candidate and the position to be evaluated; and a candidate recommendation list is generated by sorting the matching degrees.

[0013] Furthermore, it also includes a visualization module for visually presenting the development trends of candidates in each dimension of the five-dimensional evaluation system through charts.

[0014] The advantages of the present invention are that it combines adaptive algorithms, semantic analysis and time series prediction technology, can meet the different positions and real-time needs of an enterprise, integrate structured and unstructured data, and combine with growth trend prediction, which significantly improves the accuracy of talent assessment results and provides support for the enterprise's long-term human resource planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flowchart of the enterprise talent evaluation management method based on data mining.

[0016] Figure 2 The block diagram of the enterprise talent evaluation management system based on data mining described in the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1 The enterprise talent evaluation management method based on data mining described in the present invention is as follows: Figure 1 As shown, the following steps are included: S1. Establish a five-dimensional talent evaluation system. The five-dimensional evaluation system includes five dimensions: educational background, project ability, performance level, professional skills and achievement honors. Each dimension includes several secondary evaluation items and scoring rules.

[0019] S2, establish a weight distribution model for positions in the five-dimensional evaluation system. Specifically, it includes the following steps: S2.1, use NLP algorithm to analyze job descriptions and generate feature vectors of job requirements through clustering. First, clean the job description text, remove stop words and irrelevant information, then segment the text and generate word vector representation. Then use K-means clustering algorithm to cluster feature words into capability categories and generate feature vectors of job requirements.

[0020] S2.2, collect historical talent evaluation data for various positions according to the five-dimensional evaluation system, including annual performance, competency, development potential, project results, project information, ability characteristics, educational background, project capabilities, performance level, professional skills and achievement honors.

[0021] In this step, the talent evaluation data needs to be classified by position, normalized, and labeled according to the five dimensions of the five-dimensional evaluation system: educational background, project capabilities, performance level, professional skills, and achievement honors.

[0022] S2.3, based on the feature vectors of each job requirement and the historical talent evaluation data of each job, a job weight allocation model is established through the random forest algorithm. The weight allocation model can output the contribution weights of the five dimensions in the five-dimensional evaluation system to job adaptation according to the feature vectors of job requirements.

[0023] S3, input the characteristic vector of the job requirements to be evaluated and the basic requirements of the candidates formulated according to the five-dimensional evaluation system into the weight allocation model to obtain the weight of each dimension of the job to be evaluated in the five-dimensional evaluation system.

[0024] S4, collect candidate data for the positions to be evaluated according to the five-dimensional evaluation system, including structured data and unstructured data. The structured data refers to the educational background, project capabilities, performance level, professional skills and achievements that directly reflect the five-dimensional evaluation system; the unstructured data refers to the data that needs to be included in the five-dimensional evaluation system after word segmentation, denoising and feature extraction using natural language processing technology.

[0025] The BERT model is used to perform semantic segmentation and vectorization on unstructured data, extract key features that represent the candidate's ability, and then label each key feature according to the five dimensions of the five-dimensional evaluation system: educational background, project ability, performance level, professional skills, and achievement honors. The introduction of unstructured data here makes talent evaluation more comprehensive and objective, and provides multi-dimensional support for corporate talent screening.

[0026] S5, score the structured data and unstructured data separately according to the 5-dimensional evaluation system, and then weight and merge them to obtain the candidate's comprehensive evaluation vector. The 5 dimensions of the 5-dimensional evaluation system include several secondary evaluation items and scoring rules. According to the scoring rules, score the structured data and unstructured data separately, and then perform weighted calculations. In each dimension, the score ratio of structured data and unstructured data can be customized according to the information source, credibility, accuracy and the company's own preferences, and then obtain the candidate's comprehensive evaluation vector.

[0027] S6, use principal component analysis to reduce the candidate's comprehensive evaluation vector into three principal components to obtain the candidate's comprehensive ability vector.

[0028] S7, the sum of the products of each component in the candidate's comprehensive ability vector and the corresponding weight is the matching degree between the candidate and the position to be evaluated.

[0029] S8, generating a candidate recommendation list by sorting according to the matching degree.

[0030] The enterprise talent evaluation management method based on data mining described in the present invention also includes collecting a number of historical talent evaluation data of the same person in the same position in time series, and using LSTM network training to obtain a talent future development trend prediction model for predicting the future growth of candidates.

[0031] like Figure 2 As shown, the enterprise talent evaluation and management system based on data mining described in the present invention adopts the enterprise talent evaluation and management method based on data mining described in the present invention, including an adaptive weight allocation module, a multi-source data mining module, a growth trend analysis module, and an intelligent recommendation module.

[0032] The adaptive weight allocation module, based on the 5-dimensional evaluation system, establishes a weight allocation model for a position through a random forest algorithm according to the position requirement feature vector and the historical talent evaluation data of the position; obtains the weight of each dimension of the position to be evaluated in the 5-dimensional evaluation system according to the weight allocation model, the feature vector of the position to be evaluated and the basic requirements of the candidates formulated according to the 5-dimensional evaluation system; The multi-source data mining module is used to collect structured and unstructured data of candidates for the positions to be evaluated by using natural language processing technology through word segmentation, denoising and feature extraction according to the five-dimensional evaluation system; The growth trend analysis module collects several historical talent evaluation data sets of the same person in the same position in time series, and uses LSTM network training to obtain a talent future development trend prediction model to predict the future growth of the candidate; The intelligent recommendation module scores structured data and unstructured data according to a five-dimensional evaluation system, performs weighted fusion, and uses principal component analysis to reduce the dimension to three principal components to obtain a comprehensive ability vector of the candidate; the sum of the products of each component and the corresponding weight is calculated as the matching degree between the candidate and the position to be evaluated; and a candidate recommendation list is generated by sorting the matching degrees.

[0033] It also includes a visualization module, which is used to intuitively present the development trends of candidates in each dimension of the 5-dimensional evaluation system through charts.

[0034] Example 2 After actual questionnaire research and based on the needs of corporate talent assessment, the specific scoring criteria for the 5-dimensional evaluation system are determined as follows: •Educational background dimension: • The highest academic qualification is a full-time master's degree with a score of 90 points.

[0035] • The highest academic qualification is a full-time undergraduate degree with a score of 60 points.

[0036] •Project capability dimension: • Participating in 1 project as a project leader will be awarded 5 points.

[0037] • Participating in 1 project as a project participant will be awarded 2 points.

[0038] •Performance level dimension: • Each “Excellent” is worth 10 points.

[0039] • Each “Good” score is worth 7 points.

[0040] • Each time you get a “Competent”, you will get 5 points.

[0041] • For each “Incompetent” you get, your score is -5 points.

[0042] • Professional skills dimension: • Assistant Engineer will receive 10 marks.

[0043] • Engineers are awarded 20 points.

[0044] • Senior Engineer scores 30 points.

[0045] • Professor-level senior engineers score 50 points.

[0046] • Obtain 1 national certificate, with a score of 10 points.

[0047] • Obtain 1 provincial or ministerial certificate, with a score of 8 points.

[0048] • Obtain 1 General Certificate with a score of 5 points.

[0049] • Achievement Honor Dimension: • Each contribution as the first author is worth 10 points.

[0050] • As a participant, each item is worth 2 points.

[0051] • Obtain national awards, each worth 10 points.

[0052] • Obtain provincial or ministerial awards, each worth 8 points.

[0053] • Obtain ordinary rewards, each worth 5 points.

[0054] For a software development engineer position, the core requirements are as follows: •Educational background: Bachelor degree or above (Master degree or above is preferred).

[0055] •Project capabilities: Have independent development capabilities, with priority given to project leader experience.

[0056] •Performance Level: Performance record must be “Good” and above.

[0057] •Professional skills: Must master programming languages ​​(such as Java, Python) and possess a technical title of intermediate engineer or above.

[0058] •Achievement Honors: Priority will be given to those who have published papers, patents, or won awards in national competitions.

[0059] The structured information of the obtained candidate information is as follows: •Educational background: Master's degree, major in computer science and technology; •Project capability: Served as project leader for 3 projects and participated in 5 projects; •Performance level: 2 times "excellent", 3 times "good"; • Professional skills: Engineer title, 1 national qualification certificate, 1 provincial certificate; •Achievement honors: 1 national-level paper (first author), 1 award-winning participation in a national-level competition.

[0060] The candidate's unstructured information is segmented using natural language processing (NLP) technology to extract skill-related keywords such as "Python", "Java", "data analysis", etc.; denoising to remove irrelevant content such as "hobbies", "no relevant work experience", etc.; feature extraction to extract skill types such as programming languages, tool experience, algorithm capabilities, etc.; as well as quantifying the depth and breadth of skills, such as a "proficient" score of 8 points and a "mastery" score of 5 points.

[0061] The unstructured information in the obtained candidate information is as follows: •Programming languages: Python, Java (high proficiency).

[0062] •Tool experience: Familiar with TensorFlow, using it for 2 years.

[0063] •Algorithmic ability: Master the commonly used algorithms for machine learning.

[0064] •Specific ability information: Efficiently complete multi-task development and have strong innovation capabilities.

[0065] Then, according to the specific scoring criteria of the determined 5-dimensional evaluation system, which are not all listed in this embodiment, the scores of the non-structured data and the structured data in each dimension of the 5-dimensional evaluation system are obtained respectively.

[0066] In this embodiment, the weight of non-structured data is 30%, and the weight of structured data is 70%. The scores of each dimension of non-structured data and structured data are weighted and fused to obtain a comprehensive evaluation vector of the candidate.

[0067] For example, the candidate's comprehensive evaluation vector V=[85,35,49,66,48], use principal component analysis to reduce the candidate's comprehensive evaluation vector to 3 principal components, and obtain the candidate's comprehensive ability vector V'=[0.75,0.62,0.51].

[0068] According to the weight distribution model of the position in the 5-dimensional evaluation system, for the software development engineer position in this embodiment, the weight vectors of the five dimensions in the 5-dimensional evaluation system are W=[0.25, 0.2, 0.15, 0.3, 0.1] The sum of the products of each component in the candidate's comprehensive ability vector and the corresponding weight is the matching degree between the candidate and the position to be evaluated. Candidates are recommended according to the matching degree.

Claims

1. A method for enterprise talent evaluation management based on data mining, characterized in that: The following steps are involved: S1, establish a five-dimensional talent evaluation system; S2, establish a weight distribution model for positions in the five-dimensional evaluation system; S3, input the characteristic vector of the job requirements to be evaluated and the basic requirements of the candidates formulated according to the 5-dimensional evaluation system into the weight allocation model to obtain the weight of each dimension of the job to be evaluated in the 5-dimensional evaluation system; S4, collect candidate data for the positions to be evaluated according to the five-dimensional evaluation system, including structured data and unstructured data; S5, score the structured data and unstructured data according to the five-dimensional evaluation system, and then perform weighted fusion to obtain the candidate's comprehensive evaluation vector; S6, using principal component analysis to reduce the candidate's comprehensive evaluation vector into three principal components to obtain the candidate's comprehensive ability vector; S7, the sum of the products of each component in the candidate's comprehensive ability vector and the corresponding weight is the matching degree between the candidate and the position to be evaluated; S8, generating a candidate recommendation list by sorting according to the matching degree.

2. The enterprise talent evaluation management method based on data mining according to claim 1, characterized in that: The five-dimensional evaluation system described in step S1 includes five dimensions: educational background, project ability, performance level, professional skills and achievement honors. Each dimension includes several secondary evaluation items and scoring rules.

3. The enterprise talent evaluation management method based on data mining according to claim 1 is characterized by: Step S2 specifically includes: S2.1, use NLP algorithm to analyze job descriptions and generate feature vectors of job requirements through clustering; S2.2, collect historical talent evaluation data for various positions according to the five-dimensional evaluation system, including annual performance, competency, development potential, project achievements, project information, ability characteristics, educational background, project capabilities, performance level, professional skills and achievement honors; S2.3, based on the feature vectors of each job requirement and the historical talent evaluation data for each job, a job weight allocation model is established using the random forest algorithm.

4. The enterprise talent evaluation management method based on data mining according to claim 1, characterized in that: The structured data refers to the educational background, project capabilities, performance level, professional skills and achievement honors that directly reflect the five-dimensional evaluation system; the unstructured data refers to the data that needs to be included in the five-dimensional evaluation system after word segmentation, denoising and feature extraction using natural language processing technology.

5. The enterprise talent evaluation management method based on data mining according to claim 1, characterized in that: It also includes collecting historical talent evaluation data of several people in the same position in time series, and using LSTM network training to obtain a talent future development trend prediction model to predict the future growth of candidates.

6. An enterprise talent evaluation management system based on data mining, characterized by: Including adaptive weight allocation module, multi-source data mining module, growth trend analysis module, and intelligent recommendation module; The adaptive weight allocation module, based on the 5-dimensional evaluation system, establishes a weight allocation model for a position through a random forest algorithm according to the position requirement feature vector and the historical talent evaluation data of the position; obtains the weight of each dimension of the position to be evaluated in the 5-dimensional evaluation system according to the weight allocation model, the feature vector of the position to be evaluated and the basic requirements of the candidates formulated according to the 5-dimensional evaluation system; The multi-source data mining module is used to collect structured and unstructured data of candidates for the positions to be evaluated by using natural language processing technology through word segmentation, denoising and feature extraction according to the five-dimensional evaluation system; The growth trend analysis module collects several historical talent evaluation data sets of the same person in the same position in time series, and uses LSTM network training to obtain a talent future development trend prediction model to predict the future growth of the candidate; The intelligent recommendation module scores structured data and unstructured data according to a five-dimensional evaluation system, performs weighted fusion, and uses principal component analysis to reduce the dimension to three principal components to obtain a comprehensive ability vector of the candidate; the sum of the products of each component and the corresponding weight is calculated as the matching degree between the candidate and the position to be evaluated; and a candidate recommendation list is generated by sorting the matching degrees.

7. The enterprise talent evaluation management system based on data mining according to claim 6, characterized in that: It also includes a visualization module, which is used to intuitively present the development trends of candidates in each dimension of the 5-dimensional evaluation system through charts.

Images