Human resource management system based on digital intelligent technology

Abstract

The invention discloses a human resource management system based on a digital intelligent technology, and aims to optimize talent recruitment and management through technologies such as big data, artificial intelligence and cloud computing, and the system comprises the following steps: the system realizes personalized recruitment recommendation by constructing a fine-grained portrait of a job seeker, and improves the recruitment recommendation efficiency; deep learning and integrated learning technologies are adopted to improve the precision of post matching, the system combines big data analysis, predicts a talent flow trend, optimizes a recruitment process and generates employee occupational development suggestions, the system also adopts a national secret encryption technology to guarantee data security, and a user identity is authenticated through a two-factor verification technology, so that the user experience is improved. And the safety and the compliance of the system are ensured. According to the invention, automation and intelligentization of human resource management can be realized, and scientificity and accuracy of talent management are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of human resource management, and in particular to a human resource management system based on digital intelligence technology. Background Art

[0002] With the rapid development of information technology, traditional human resource management methods have gradually exposed problems such as inefficiency and delayed decision-making. Especially in the process of talent recruitment and employee management, the efficiency and accuracy of manual resume screening and job matching are limited. With the continuous development of big data and artificial intelligence technology, traditional methods are difficult to meet the needs of modern enterprises for efficient and accurate recruitment, resulting in lengthy recruitment processes and high risks of talent turnover.

[0003] In the existing technology, traditional human resource management systems mainly rely on manual operations and simple rule-based matching methods. These methods are unable to cope with large amounts of job seeker information and job requirements, and have the following defects: 1. Inefficient matching accuracy: Traditional systems mostly rely on manual screening or simple algorithm matching, which cannot accurately identify the potential matching relationship between job seekers and positions, resulting in inefficient recruitment and easy loss of excellent talents.

[0004] 2. Lengthy recruitment process: Traditional methods rely on manual processes and static rules, which cannot be flexibly adjusted according to real-time data and recruitment needs. The recruitment cycle is too long and suitable candidates are missed.

[0005] 3. Unable to dynamically adapt to changes in talent flow: Existing systems usually lack real-time predictions of employee flow trends and changes in the talent market, resulting in slow response by companies in personnel planning and adjustments, making it difficult to achieve accurate human resource management.

[0006] 4. Poor data security: Due to data privacy and security issues, traditional systems have insufficient encryption protection measures for human resources data, making it vulnerable to attacks or leaks, affecting the compliance and security of the enterprise.

[0007] Therefore, how to provide a human resource management system based on digital intelligence technology is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0008] One purpose of the present invention is to propose a human resource management system based on digital intelligence technology. The present invention optimizes the whole process of talent recruitment, management and training by combining big data, artificial intelligence, cloud computing, deep learning and security encryption technology. The present invention describes in detail the algorithms of intelligent matching, personalized recommendation, recruitment process prediction and loss risk prediction, and adopts national secret encryption and two-factor verification technology to ensure data security. It has the advantages of high recruitment efficiency, accurate job matching, strong data security and strong decision-making support capabilities.

[0009] A human resource management system based on digital intelligence technology according to an embodiment of the present invention includes the following steps: S1, human resource data collection module, used to collect human resource data from job seekers, enterprises and industry data sources, and format the collected human resource data; S2, talent portrait construction module, uses classification and clustering algorithms to analyze formatted human resource data, extract multi-dimensional labels of job seekers, and generate fine-grained talent portraits; S3, intelligent matching and recommendation module, based on fine-grained talent portraits, uses machine learning, ensemble learning and knowledge enhancement technologies to calculate the suitability of job seekers and positions, and generates personalized talent recommendation results; S4, microservice management module, is used to split each functional module through microservice architecture based on personalized talent recommendation results, and use lightweight communication protocol to realize data interaction between modules; S5, intelligent decision-making analysis module, based on personalized talent recommendation results, uses deep learning algorithms and business intelligence analysis tools to analyze recruitment processes, talent flow trends and job adaptation, and generates recruitment optimization strategies, employee career development suggestions and loss risk predictions; S6, computing resource management module, which provides elastic scaling capabilities of computing, storage and network resources based on cloud computing, and combines edge computing technology to process employee behavior data collected by enterprise IoT devices in real time, and integrates and shares data with enterprise management systems; S7, security control module, uses national secret encryption technology to store and transmit encrypt human resources data, combines two-factor verification technology to authenticate users, and provides permission management, log auditing and data security compliance testing.

[0010] Optionally, the S2 specifically includes: S21, data preprocessing module, used to standardize, remove duplicates and convert features of formatted human resource data and extract job seeker text data, wherein the job seeker text data is vectorized and based on Method to calculate the feature weight vector of job seeker text data , where each feature weight vector The calculation is as follows: ; in, Represents job seekers In text data The weight on Representing text data In job seekers The frequency of occurrence in text data, Indicates that it contains text data The number of job seekers, Represents the total number of all job applicants; S22, feature extraction module, connected to the data preprocessing module, is used to extract skills, education background, work experience and industry suitability features from the standardized human resources data, calculate feature importance using a decision tree algorithm based on information gain, and construct a feature vector set , where the importance of each feature The calculation is as follows: ; in, Representation characteristics For the target variable The information gain of Represents the target variable The entropy of Indicates that given a feature The conditional entropy when S23, cluster analysis module, connected with feature extraction module, is used to group job seeker text data based on clustering algorithm to form different categories of talent groups, and adopt weighted Clustering algorithm, which calculates the similarity between job applicants based on weighted Euclidean distance: ; in, Represents job seekers and job seekers The similarity between and Respectively represent the job seekers in characteristics The value of the dimension, Representation characteristics The weight factor of Indicates the total number of feature dimensions; S24, a multi-dimensional label generation module, connected to the cluster analysis module, is used to generate the job seeker's skill label, job matching label, industry suitability label and career development potential label based on the grouping result of the job seeker's text data based on the clustering algorithm, and adopt The normalization method calculates the probability of a job seeker being suitable for each job category: ; in, Represents job seekers In job category The probability of adaptation under Represents job seekers In job category The feature scores on Represents the total number of all job categories, Represents job seekers In job category Feature scores on ; S25, talent portrait optimization module, connected with the multi-dimensional label generation module, is used to optimize the talent portrait based on the job matching label, and dynamically adjust the job seeker's matching probability in combination with the Bayesian update model: ; in, Represents job seekers In job category The updated adaptation probability under is the initial adaptation probability, Indicates the job seeker's preference for job category The confidence level of relevant behavioral data, B represents the behavioral data of job seekers, , Indicates the total number of all job categories; S26, the talent portrait output module communicates with the talent portrait optimization module to generate fine-grained talent portraits based on the optimized multi-dimensional labels.

[0011] Optionally, the S23 specifically includes: S231, a feature matrix construction module, used to receive a feature vector set provided by the feature extraction module, and perform standardization on the feature vectors to construct a job seeker feature matrix ,in Represents job seekers In the feature dimension The normalized value on is the total number of feature dimensions; S232, a weighted similarity calculation module, connected to the feature matrix construction module, for calculating the similarity between job seekers based on the weighted Euclidean distance ; S233, initial cluster center selection module, connected with weighted similarity calculation module, for optimizing The algorithm selects the initial cluster center and randomly selects the first cluster center , and then according to the selection probability Select the remaining cluster centers: ; in, is the cluster center As the selection probability of the new cluster center, is the cluster center The minimum distance to the currently selected cluster center, is the total number of cluster centers; S234, weighted clustering iteration module, connected to the initial cluster center selection module, for dynamic weighted clustering The algorithm iteratively optimizes the job seeker text data. In each iteration, it calculates the weighted distance from the data point in the job seeker text data to each cluster center and redistributes the categories according to the minimum distance so that each cluster center Update to the weighted mean of its cluster members: ; in, represents the updated cluster center, Indicates that the current A collection of job seekers of the class, For job seekers The characteristic vector of Represents job seekers The weighting factor of is the normalized weight sum of the clustering categories; S235, the convergence judgment module is connected to the weighted clustering iteration module to judge whether the clustering process has converged and calculate the clustering center of the previous round The cluster center after this round of update The change between : ; in, represents the change of all cluster centers, and Represent the cluster centers of the current round and the previous round respectively, is the total number of clusters, if Less than the set convergence threshold , the clustering iteration is terminated; S236, a clustering result output module, connected to the convergence judgment module, is used to output the final job seeker category grouping and provide the grouping result to the multi-dimensional label generation module.

[0012] Optionally, the S3 specifically includes: S31, data input module, used to receive job seeker profile data and enterprise recruitment demand data generated by fine-grained talent portrait, and use the job seeker feature matrix And build a job characteristics matrix ,in and Represents the job seekers and positions in the characteristic dimensions The value on Indicates the total number of feature dimensions; S32, a suitability calculation module, connected to the data input module, for calculating the suitability between the job seeker and the position based on a machine learning method , where weighted cosine similarity is used to calculate the degree of fitness: ; in, Represents job seekers With position The adaptability, and Respectively represent job seekers and positions In the feature dimension The value on Represents feature dimension The weight factor controls the influence of features on the matching calculation. Indicates the total number of feature dimensions; S33, integrated learning optimization module, connected to the fitness calculation module, is used to optimize the fitness between job seekers and positions using an integrated learning method, and calculate the final position recommendation score based on a weighted voting mechanism of multiple classifiers ; S34, the knowledge enhancement module, is connected to the integrated learning optimization module and is used to optimize the matching process between job seekers and positions based on the knowledge graph enhancement technology. and job knowledge graph , calculate the relationship path score between job seekers and positions ; S35, personalized recommendation result generation module, connected with the knowledge enhancement module, is used to generate personalized recommendation results based on the suitability of job seekers and positions. , Final job recommendation score and relationship path score Calculate the job seeker's final job recommendation index : ; in, Represents job seekers At work The final job recommendation index on To adjust the hyperparameters that influence the weight of each score; S36, a recommendation result output module, connected to the personalized recommendation result generation module, is used to generate a recommendation result based on the job seeker's final job recommendation index. Generate personalized talent recommendation results.

[0013] Optionally, the S5 specifically includes: S51, data processing module, used to receive personalized talent recommendation results, obtain recruitment process data, talent flow data and job adaptation data, and build a recruitment feature matrix , flow characteristic matrix and job adaptation feature matrix ; S52, recruitment process modeling module, connected with the data processing module, is used to Perform recruitment process prediction and use gated recurrent units to calculate recruitment time estimates : ; in, represents the estimated value of the predicted recruitment time, express The weight matrix of the model, is the bias term, is a nonlinear activation function; S53, talent flow trend modeling module, connected with the recruitment process modeling module, is used to Calculate the talent flow trend and calculate the attrition rate by combining variational autoencoder with Markov decision process : ; in, Indicates at time The rate of internal brain drain, represents the set of all possible employee states, For employee status The initial probability of represents the set of possible flow behaviors, Indicates employee status Mobile behavior The probability of Indicates employee status Take mobile action The reward value after that; S54, job adaptation optimization module, connected with the talent flow trend modeling module, is used to optimize the job adaptation feature matrix based on the job adaptation feature matrix. Calculating job fit adjustment scores ; S55, intelligent decision output module, connected with the job suitability optimization module, is used to estimate the recruitment time based on the , Talent flow trend modeling, attrition rate and job fit adjustment scores , generate recruitment optimization strategies, employee career development suggestions and turnover risk predictions.

[0014] Optionally, the S7 specifically includes: S71, data encryption module, used for storing and transmitting encrypted human resources data, using national secret algorithm to encrypt human resources data, the national secret algorithm includes Asymmetric encryption, Hash Algorithms and Symmetric encryption algorithm; S72, user identity authentication module, connected to the data encryption module, used to authenticate the user based on two-factor authentication, using a one-time password and based on The digital signature is combined with the user's identity confirmation; S73, the authority management module is connected to the user identity authentication module and is used to allocate user authority based on role-based access control technology and set the access control matrix , where the access rights calculation formula is as follows: ; in, Indicates user About Resources Access rights; S74, the log audit module is connected to the authority management module, and is used to record and store user operation logs. The log data is stored in a hash chain structure to prevent log tampering: ; in, For the The hash value of the log. Represents log data, Indicates data concatenation operation. is a hash function; S75, data security compliance detection module, connected with the log audit module, is used to detect system compliance based on data integrity verification and access control strategy, using Data integrity verification method for trees, The root hash value of the tree is calculated as follows: ; in, for The root hash value of the tree, and are the hash values ​​of the left subtree and the right subtree respectively.

[0015] The beneficial effects of the present invention are: (1) The present invention optimizes the construction of talent portraits and intelligent matching algorithms by combining machine learning, integrated learning, deep learning and knowledge enhancement technologies, realizes personalized talent recommendation, and improves the accuracy and efficiency of recruitment matching. Based on long short-term memory networks, variational autoencoders and Markov decision processes, the talent flow trend is predicted, so that the system can dynamically adapt to market changes and effectively improve the scientificity and flexibility of recruitment management.

[0016] (2) The present invention uses the proximal strategy optimization algorithm in reinforcement learning to optimize the job matching strategy, and combines cloud computing and edge computing technology to achieve real-time adjustment of job requirements. The intelligent decision-making analysis module predicts the recruitment process cycle, combined with the job suitability optimization algorithm, improves the automation level of enterprise recruitment management, reduces manual intervention and lengthy cycles in the recruitment process, and improves the recruitment success rate.

[0017] (3) The present invention combines national secret encryption technology, two-factor authentication, role-based access control and hash chain log storage technology to ensure the storage and transmission security of human resources data, and achieve precise control of user rights and security compliance testing. The Merkle tree verification method is used to ensure data integrity, improve the data security and compliance of the system, and effectively reduce the risk of data leakage and illegal access.

[0018] (4) The present invention is based on multi-dimensional data analysis to achieve recruitment process optimization, talent development prediction and employee turnover risk prediction, and provides recruitment optimization strategies and job suitability adjustment suggestions through an intelligent decision support system, so that enterprises can allocate human resources more scientifically and efficiently, reduce talent loss, and improve the accuracy of talent management of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is the overall architecture diagram of a human resource management system based on digital intelligence technology proposed in the present invention. DETAILED DESCRIPTION

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0021] refer to Figure 1 , a human resource management system based on digital intelligence technology, comprising the following steps: S1. Human resource data collection module, used to collect human resource data from job seekers, enterprises and industry data sources, and format the collected human resource data. The formatting process includes data cleaning, standardization, deduplication and structured conversion; This implementation method obtains human resource data from job seekers, enterprises and industry data sources, and formats the data to ensure the accuracy and consistency of subsequent portrait construction, matching analysis and decision support, thereby improving the system's data processing capabilities and data circulation efficiency.

[0022] S2, talent portrait construction module, uses classification and clustering algorithms to analyze formatted human resource data, extract multi-dimensional labels of job seekers, and generate fine-grained talent portraits; S3, intelligent matching and recommendation module, based on fine-grained talent portraits, uses machine learning, ensemble learning and knowledge enhancement technologies to calculate the suitability of job seekers and positions, and generates personalized talent recommendation results; S4, microservice management module, is used to split each functional module through microservice architecture based on personalized talent recommendation results, and use lightweight communication protocol to realize data interaction between modules; This implementation adopts a microservice architecture, combined with a lightweight communication protocol, to split the human resources management system into independent functional modules, including job posting, resume delivery, interview arrangement, etc. Service registration and discovery, load balancing, and gateways ensure efficient communication and dynamic expansion of each module, and use Docker for containerized deployment to improve the elasticity and maintainability of the system; integrate machine learning recommendation models, optimize job matching based on job seeker behavior data, enhance the accuracy and intelligence of the system, and improve recruitment efficiency and talent matching accuracy.

[0023] S5, intelligent decision-making analysis module, based on personalized talent recommendation results, uses deep learning algorithms and business intelligence analysis tools to analyze recruitment processes, talent flow trends and job adaptation, and generates recruitment optimization strategies, employee career development suggestions and loss risk predictions; S6, computing resource management module, which provides elastic scaling capabilities of computing, storage and network resources based on cloud computing, and combines edge computing technology to process employee behavior data collected by enterprise IoT devices in real time, and integrates and shares data with enterprise management systems; This implementation method realizes device integration and shared intercommunication. Device integration realizes unified access of multiple heterogeneous devices and is compatible with multiple types of transmission protocols. Through edge computing, the back-end pressure is reduced, and real-time collection and processing of sensor data are realized to ensure the real-time nature of data; shared intercommunication seamlessly connects platform data with the enterprise's internal ERP and CRM systems to achieve unified management of business data. It also provides rich API interfaces and flexible configuration management functions to facilitate users to customize the system according to their needs.

[0024] S7, security control module, uses national secret encryption technology to store and transmit encrypt human resources data, combines two-factor verification technology to authenticate users, and provides permission management, log auditing and data security compliance testing.

[0025] In this implementation, S2 specifically includes: S21, data preprocessing module, used to standardize, remove duplicates and convert features of formatted human resource data and extract job seeker text data, which includes basic information, educational background, work experience, job search intention, etc., wherein the job seeker text data is word vectorized and based on Method to calculate the feature weight vector of job seeker text data , where each feature weight vector The calculation is as follows: ; in, Represents job seekers In text data The weight on Representing text data In job seekers The frequency of occurrence in text data, Indicates that it contains text data The number of job seekers, Represents the total number of all job applicants; S22, feature extraction module, connected to the data preprocessing module, is used to extract skills, education background, work experience and industry suitability features from the standardized human resources data, calculate feature importance using a decision tree algorithm based on information gain, and construct a feature vector set , where the importance of each feature The calculation is as follows: ; in, Representation characteristics For the target variable The information gain of Represents the target variable The entropy of Indicates that given a feature The conditional entropy when S23, cluster analysis module, connected with feature extraction module, is used to group job seeker text data based on clustering algorithm to form different categories of talent groups, and adopt weighted Clustering algorithm, which calculates the similarity between job applicants based on weighted Euclidean distance: ; in, Represents job seekers and job seekers The similarity between and Respectively represent the job seekers in characteristics The value of the dimension, Representation characteristics The weight factor of Indicates the total number of feature dimensions; S24, a multi-dimensional label generation module, connected to the cluster analysis module, is used to generate a job applicant's job matching label based on the grouping result of the job applicant's text data based on the clustering algorithm, wherein the job matching label refers to a key label that measures the degree of adaptation between the job applicant and different positions, and adopts The normalization method calculates the probability of a job seeker being suitable for each job category: ; in, Represents job seekers In job category The probability of adaptation under Represents job seekers In job category The feature scores on Represents the total number of all job categories, Represents job seekers In job category Feature scores on ; This implementation method is based on the grouping results of job seekers' text data, comprehensively considering skills, experience, industry adaptability and job-seeking intentions, and generates job matching labels. The system parses the job seekers' text data, extracts core information such as skills, work experience, industry background and job-seeking intentions, and constructs job seekers' feature vectors. Job seekers are divided into different categories through cluster analysis to determine the matching characteristics of each category in specific positions. The system combines job requirement information to calculate the matching degree between job seekers and positions, including dimensions such as skill matching, industry adaptation and job competency potential, and normalizes the matching results to make the matching degrees of different dimensions comparable under a unified standard. The system dynamically adjusts the matching weights according to the matching results and job seekers' category characteristics, generates accurate job matching labels, and provides a reliable basis for talent recommendation, ensuring that the matching results meet both job requirements and job seekers' career development goals.

[0026] S25, talent portrait optimization module, connected with the multi-dimensional label generation module, is used to optimize the talent portrait based on the job matching label, and dynamically adjust the job seeker's matching probability in combination with the Bayesian update model: ; in, Represents job seekers In job category The updated adaptation probability under is the initial adaptation probability, Indicates the job seeker's preference for job category The confidence level of relevant behavioral data, B represents the behavioral data of job seekers, , Indicates the total number of all job categories; S26, the talent portrait output module communicates with the talent portrait optimization module to generate fine-grained talent portraits based on the optimized multi-dimensional labels.

[0027] This implementation method accurately constructs the talent portrait of job seekers by combining text feature quantization, feature extraction based on information gain, weighted cluster analysis and Bayesian update model. It uses weighted Euclidean distance to calculate the similarity between job seekers, optimizes talent grouping, and dynamically adjusts job suitability based on behavioral data to improve the accuracy and stability of talent recommendations, realize personalized, multi-dimensional matching optimization, and provide more intelligent and efficient support for human resource management.

[0028] In this implementation manner, the S23 specifically includes: S231, a feature matrix construction module, used to receive a feature vector set provided by the feature extraction module, and perform standardization on the feature vectors to construct a job seeker feature matrix ,in Represents job seekers In the feature dimension The normalized value on is the total number of feature dimensions; S232, a weighted similarity calculation module, connected to the feature matrix construction module, for calculating the similarity between job seekers based on the weighted Euclidean distance ; S233, initial cluster center selection module, connected with weighted similarity calculation module, for optimizing The algorithm selects the initial cluster center and randomly selects the first cluster center , and then according to the selection probability Select the remaining cluster centers: ; in, is the cluster center As the selection probability of the new cluster center, is the cluster center The minimum distance to the currently selected cluster center, is the total number of cluster centers; S234, weighted clustering iteration module, connected to the initial cluster center selection module, for dynamic weighted clustering The algorithm iteratively optimizes the job seeker text data. In each iteration, it calculates the weighted distance from the data point in the job seeker text data to each cluster center and redistributes the categories according to the minimum distance so that each cluster center Update to the weighted mean of its cluster members: ; in, represents the updated cluster center, Indicates that the current A collection of job seekers of the class, For job seekers The characteristic vector of Represents job seekers The weighting factor of is the normalized weight sum of the clustering categories; S235, the convergence judgment module is connected to the weighted clustering iteration module to judge whether the clustering process has converged and calculate the clustering center of the previous round The cluster center after this round of update The change between : ; in, represents the change of all cluster centers, and Represent the cluster centers of the current round and the previous round respectively, is the total number of clusters, if Less than the set convergence threshold , the clustering iteration is terminated; S236, a clustering result output module, connected to the convergence judgment module, is used to output the final job seeker category grouping and provide the grouping result to the multi-dimensional label generation module.

[0029] The present invention improves clustering accuracy and stability by constructing a feature matrix and calculating the similarity of job seekers based on weighted Euclidean distance, combined with an optimized initial cluster center selection algorithm. Dynamic weighted clustering is used to iteratively optimize the classification of job seekers, making the category allocation more accurate, and the stability of the clustering results is ensured by the convergence judgment module, thereby improving the accuracy of talent portraits and providing a more reliable basis for subsequent matching and recommendation.

[0030] In this implementation, S3 specifically includes: S31, data input module, used to receive job seeker profile data and enterprise recruitment demand data generated by fine-grained talent portrait, and use the job seeker feature matrix And build a job characteristics matrix ,in and Represents the job seekers and positions in the characteristic dimensions The value on Indicates the total number of feature dimensions; S32, a suitability calculation module, connected to the data input module, for calculating the suitability between the job seeker and the position based on a machine learning method , where weighted cosine similarity is used to calculate the degree of fitness: ; in, Represents job seekers With position The adaptability, and Respectively represent job seekers and positions In the feature dimension The value on Represents feature dimension The weight factor controls the influence of features on the matching calculation. Indicates the total number of feature dimensions; S33, integrated learning optimization module, connected to the fitness calculation module, is used to optimize the fitness between job seekers and positions using an integrated learning method, and calculate the final position recommendation score based on a weighted voting mechanism of multiple classifiers ; S34, the knowledge enhancement module, is connected to the integrated learning optimization module and is used to optimize the matching process between job seekers and positions based on the knowledge graph enhancement technology. and job knowledge graph , calculate the relationship path score between job seekers and positions ; S35, personalized recommendation result generation module, connected with the knowledge enhancement module, is used to generate personalized recommendation results based on the suitability of job seekers and positions. , Final job recommendation score and relationship path score Calculate the job seeker's final job recommendation index : ; in, Represents job seekers At work The final job recommendation index on To adjust the hyperparameters that influence the weight of each score; S36, a recommendation result output module, connected to the personalized recommendation result generation module, is used to generate a recommendation result based on the job seeker's final job recommendation index. Generate personalized talent recommendation results.

[0031] This implementation combines weighted cosine similarity calculation, integrated learning optimization and knowledge enhancement technology to comprehensively improve the matching accuracy between job seekers and positions. The relationship path score between job seekers and positions is constructed through knowledge graphs, and the recommendation index is dynamically adjusted using personalized recommendation algorithms to improve the accuracy and adaptability of talent recommendations, providing companies with more efficient and intelligent talent recruitment solutions.

[0032] In this implementation manner, S5 specifically includes: S51, data processing module, used to receive personalized talent recommendation results, obtain recruitment process data, talent flow data and job adaptation data, and build a recruitment feature matrix , flow characteristic matrix and job adaptation feature matrix ; S52, recruitment process modeling module, connected with the data processing module, is used to Perform recruitment process prediction and use gated recurrent units to calculate recruitment time estimates : ; in, represents the estimated value of the predicted recruitment time, express The weight matrix of the model, is the bias term, is a nonlinear activation function; S53, talent flow trend modeling module, connected with the recruitment process modeling module, is used to Calculate the talent flow trend and calculate the attrition rate by combining variational autoencoder with Markov decision process : ; in, Indicates at time The rate of internal brain drain, represents the set of all possible employee states, For employee status The initial probability of represents the set of possible flow behaviors, Indicates employee status Mobile behavior The probability of Indicates employee status Take mobile action The reward value after that; S54, job adaptation optimization module, connected with the talent flow trend modeling module, is used to optimize the job adaptation feature matrix based on the job adaptation feature matrix. Calculating job fit adjustment scores ; S55, intelligent decision output module, connected with the job suitability optimization module, is used to estimate the recruitment time based on the , Talent flow trend modeling, attrition rate and job fit adjustment scores , generate recruitment optimization strategies, employee career development suggestions and turnover risk predictions.

[0033] This implementation combines recruitment process modeling, talent flow trend prediction and job suitability optimization to achieve accurate talent recruitment and loss risk prediction. Recruitment time estimation is optimized through gated recurrent units, talent flow trends are predicted using variational autoencoders and Markov decision processes, and job matching strategies are dynamically optimized in combination with deep reinforcement learning, thereby improving recruitment efficiency, reducing talent loss rates, and providing companies with accurate recruitment optimization strategies and career development suggestions.

[0034] In this implementation manner, the S7 specifically includes: S71, data encryption module, used for storing and transmitting encrypted human resources data, using national secret algorithm to encrypt human resources data, the national secret encryption algorithm includes Asymmetric encryption, Hash Algorithms and Symmetric encryption algorithm; S72, user identity authentication module, connected to the data encryption module, used to authenticate the user based on two-factor authentication, using a one-time password and based on The digital signature is combined with the user's identity confirmation; S73, the authority management module is connected to the user identity authentication module and is used to allocate user authority based on role-based access control technology and set the access control matrix , where the access rights calculation formula is as follows: ; in, Indicates user About Resources Access rights; S74, the log audit module is connected to the authority management module, and is used to record and store user operation logs. The log data is stored in a hash chain structure to prevent log tampering: ; in, For the The hash value of the log. Represents log data, Indicates data concatenation operation. is a hash function; S75, data security compliance detection module, connected with the log audit module, is used to detect system compliance based on data integrity verification and access control strategy, using Data integrity verification method for trees, The root hash value of the tree is calculated as follows: ; in, for The root hash value of the tree, and are the hash values ​​of the left subtree and the right subtree respectively.

[0035] The present invention adopts national secret encryption technology combined with two-factor authentication to achieve secure storage and transmission of human resources data and ensure the reliability of user identity authentication. It accurately manages permissions through role-based access control, combines hash chain log storage to prevent data tampering, and uses Merkle tree to verify data integrity, ensuring that the system meets security compliance requirements and improves data security and the accuracy of access management.

[0036] Embodiment 1: In order to verify the feasibility of the present invention in implementation, the present invention is applied to the human resource management system of a large technology company with 5,000 employees. The company needs to recruit about 1,000 new employees each year and manage the job adjustment, career development planning and performance appraisal of about 2,000 employees. The traditional talent management method has problems such as low recruitment efficiency, low matching accuracy, and insufficient prediction ability of talent loss, which seriously affects the talent management level of the company. The company decided to implement the human resource management system of the present invention to improve recruitment efficiency, optimize talent matching, reduce the risk of talent loss, and enhance data security.

[0037] The enterprise first deploys the human resources management system of the present invention and accesses data sources such as historical recruitment data, job seeker information, job requirements and employee performance. After the system standardizes and cleans the historical data, it uses the intelligent analysis module to optimize recruitment, recommend talents, and predict the risk of loss, and uses the security control module to ensure data security. The specific application process is as follows: In the optimization of recruitment process, the system receives the job demand information provided by the enterprise, and deeply analyzes the data of job seekers to build a fine-grained talent portrait. Using machine learning and ensemble learning methods, the system calculates the matching degree between job seekers and positions, and dynamically adjusts the recommendation strategy based on reinforcement learning to ensure that the recommendation results are more in line with the needs of the enterprise.

[0038] In terms of predicting talent flow trends, the system uses a combination of variational autoencoders and Markov decision processes to predict the probability of employee turnover in the next year and provide targeted talent retention recommendations. The job suitability optimization module dynamically adjusts the job recommendation strategy based on a deep reinforcement learning algorithm to improve the accuracy of talent management.

[0039] In terms of data security, the system uses the national SM2, SM3, and SM4 encryption technologies, combined with two-factor authentication and role-based access control mechanisms, to achieve user authentication, authority management, and data storage encryption. Logs are stored in a hash chain structure to ensure that data cannot be tampered with, and the Merkle tree verification method is used for data integrity testing.

[0040] In order to evaluate the actual effect of the system, the enterprise conducted a six-month comparative test after implementing the system of the present invention, and compared the effect with the traditional manual recruitment and management method. The experimental data is shown in the following table.

[0041] Table 1 Comparison of recruitment matching efficiency ; Table 2 Comparison of talent flow trend forecasts ; Table 3 Comparison of data security and system access ; Experimental results show that the system of the present invention is significantly superior to traditional methods in terms of recruitment matching accuracy, talent loss prediction accuracy and data security. In terms of recruitment efficiency, the system of the present invention shortens the job recruitment cycle by 48.6%, reduces the job applicant screening time by 83.3%, and improves the accuracy of candidate matching. In terms of talent flow prediction, the system's prediction accuracy is improved by 25.2% compared with traditional methods, and the talent loss rate of key positions is reduced by 5.1%. In terms of data security, the system effectively reduces the risk of data leakage through two-factor authentication and SM4 encryption technology, and ensures the security of user access through dynamic RBAC permission management.

[0042] In summary, the human resource management system of the present invention successfully solves the problems of inefficient matching, inaccurate loss prediction and insufficient data security in traditional recruitment and talent management methods, and provides enterprises with an efficient, safe and accurate human resource management solution.

[0043] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A human resource management system based on digital intelligence technology, characterized in that: The steps include: S1, human resource data collection module, used to collect human resource data from job seekers, enterprises and industry data sources, and format the collected human resource data; S2, talent portrait construction module, uses classification and clustering algorithms to analyze formatted human resource data, extract multi-dimensional labels of job seekers, and generate fine-grained talent portraits; S3, intelligent matching and recommendation module, based on fine-grained talent portraits, uses machine learning, ensemble learning and knowledge enhancement technologies to calculate the suitability of job seekers and positions, and generates personalized talent recommendation results; S4, microservice management module, is used to split each functional module through microservice architecture based on personalized talent recommendation results, and use lightweight communication protocol to realize data interaction between modules; S5, intelligent decision-making analysis module, based on personalized talent recommendation results, uses deep learning algorithms and business intelligence analysis tools to analyze recruitment processes, talent flow trends and job adaptation, and generates recruitment optimization strategies, employee career development suggestions and loss risk predictions; S6, computing resource management module, which provides elastic scaling capabilities of computing, storage and network resources based on cloud computing, and combines edge computing technology to process employee behavior data collected by enterprise IoT devices in real time, and integrates and shares data with enterprise management systems; S7, security control module, uses national secret encryption technology to store and transmit encrypt human resources data, combines two-factor verification technology to authenticate users, and provides permission management, log auditing and data security compliance testing.

2. According to claim 1, a human resource management system based on digital intelligence technology is characterized in that: The S2 specifically includes: S21, data preprocessing module, used to standardize, remove duplicates and convert features of formatted human resource data and extract job seeker text data, wherein the job seeker text data is vectorized and based on Method to calculate the feature weight vector of job seeker text data , where each feature weight vector The calculation is as follows: ； in, Represents job seekers In text data The weight on Representing text data In job seekers The frequency of occurrence in text data, Indicates that it contains text data The number of job seekers, Represents the total number of all job applicants; S22, feature extraction module, connected to the data preprocessing module, is used to extract skills, education background, work experience and industry suitability features from the standardized human resources data, calculate feature importance using a decision tree algorithm based on information gain, and construct a feature vector set , where the importance of each feature The calculation is as follows: ； in, Representation characteristics For the target variable The information gain of Represents the target variable The entropy of Indicates that given a feature The conditional entropy when S23, cluster analysis module, connected with feature extraction module, is used to group job seeker text data based on clustering algorithm to form different categories of talent groups, and adopt weighted Clustering algorithm, which calculates the similarity between job applicants based on weighted Euclidean distance: ； in, Represents job seekers and job seekers The similarity between and Respectively represent the job seekers in characteristics The value of the dimension, Representation characteristics The weight factor of Indicates the total number of feature dimensions; S24, a multi-dimensional label generation module, connected to the cluster analysis module, is used to generate the job seeker's skill label, job matching label, industry suitability label and career development potential label based on the grouping result of the job seeker's text data based on the clustering algorithm, and adopt The normalization method calculates the probability of a job seeker being suitable for each job category: ； in, Represents job seekers In job category The probability of adaptation under Represents job seekers In job category The feature scores on Represents the total number of all job categories, Represents job seekers In job category Feature scores on ; S25, talent portrait optimization module, connected with the multi-dimensional label generation module, is used to optimize the talent portrait based on the job matching label, and dynamically adjust the job seeker's matching probability in combination with the Bayesian update model: ； in, Represents job seekers In job category The update adaptation probability under is the initial adaptation probability, Indicates the job seeker's preference for job category The confidence level of relevant behavioral data, B represents the behavioral data of job seekers, , Indicates the total number of all job categories; S26, the talent portrait output module communicates with the talent portrait optimization module to generate fine-grained talent portraits based on the optimized multi-dimensional labels.

3. A human resource management system based on digital intelligence technology according to claim 2, characterized in that: The S23 specifically includes: S231, a feature matrix construction module, used to receive a feature vector set provided by the feature extraction module, and perform standardization on the feature vectors to construct a job seeker feature matrix ,in Represents job seekers In the feature dimension The normalized value on is the total number of feature dimensions; S232, a weighted similarity calculation module, connected to the feature matrix construction module, for calculating the similarity between job seekers based on the weighted Euclidean distance ; S233, initial cluster center selection module, connected with weighted similarity calculation module, for optimizing The algorithm selects the initial cluster center and randomly selects the first cluster center , and then according to the selection probability Select the remaining cluster centers: ； in, is the cluster center As the selection probability of the new cluster center, is the cluster center The minimum distance to the currently selected cluster center, is the total number of cluster centers; S234, weighted clustering iteration module, connected to the initial cluster center selection module, for dynamic weighted clustering The algorithm iteratively optimizes the job seeker text data. In each iteration, it calculates the weighted distance from the data point in the job seeker text data to each cluster center and redistributes the categories according to the minimum distance so that each cluster center Update to the weighted mean of its cluster members: ； in, represents the updated cluster center, Indicates that the current A collection of job seekers of the class, For job seekers The characteristic vector of Represents job seekers The weighting factor of is the normalized weight sum of the clustering categories; S235, the convergence judgment module is connected to the weighted clustering iteration module to judge whether the clustering process has converged and calculate the clustering center of the previous round The cluster center after this round of update The change between : ； in, represents the change of all cluster centers, and Represent the cluster centers of the current round and the previous round respectively, is the total number of clusters, if Less than the set convergence threshold , the clustering iteration is terminated; S236, a clustering result output module, connected to the convergence judgment module, is used to output the final job seeker category grouping and provide the grouping result to the multi-dimensional label generation module.

4. A human resource management system based on digital intelligence technology according to claim 3, characterized in that: The S3 specifically includes: S31, data input module, used to receive job seeker profile data and enterprise recruitment demand data generated by fine-grained talent portrait, and use the job seeker feature matrix And build a job characteristics matrix ,in and Represents the job seekers and positions in the characteristic dimensions The value on Indicates the total number of feature dimensions; S32, a suitability calculation module, connected to the data input module, for calculating the suitability between the job seeker and the position based on a machine learning method , where weighted cosine similarity is used to calculate the degree of fitness: ； in, Represents job seekers With position The adaptability, and Respectively represent job seekers and positions In the feature dimension The value on Represents feature dimension The weight factor controls the influence of features on the matching calculation. Indicates the total number of feature dimensions; S33, integrated learning optimization module, connected to the fitness calculation module, is used to optimize the fitness between job seekers and positions using an integrated learning method, and calculate the final position recommendation score based on a weighted voting mechanism of multiple classifiers ; S34, the knowledge enhancement module, is connected to the integrated learning optimization module and is used to optimize the matching process between job seekers and positions based on the knowledge graph enhancement technology. and job knowledge graph , calculate the relationship path score between job seekers and positions ; S35, personalized recommendation result generation module, connected with the knowledge enhancement module, is used to generate personalized recommendation results based on the suitability of job seekers and positions. , Final job recommendation score and relationship path score Calculate the job seeker's final job recommendation index : ； in, Represents job seekers At work The final job recommendation index on To adjust the hyperparameters that influence the weight of each score; S36, a recommendation result output module, connected to the personalized recommendation result generation module, is used to generate a recommendation result based on the job seeker's final job recommendation index. Generate personalized talent recommendation results.

5. A human resource management system based on digital intelligence technology according to claim 4, characterized in that: The S5 specifically includes: S51, data processing module, used to receive personalized talent recommendation results, obtain recruitment process data, talent flow data and job adaptation data, and build a recruitment feature matrix , flow characteristic matrix and job adaptation feature matrix ; S52, recruitment process modeling module, connected with the data processing module, is used to Perform recruitment process prediction and use gated recurrent units to calculate recruitment time estimates : ； in, represents the estimated value of the predicted recruitment time, express The weight matrix of the model, is the bias term, is a nonlinear activation function; S53, talent flow trend modeling module, connected with the recruitment process modeling module, is used to Calculate the talent flow trend and calculate the attrition rate by combining variational autoencoder with Markov decision process : ； in, Indicates at time The rate of internal brain drain, represents the set of all possible employee states, For employee status The initial probability of represents the set of possible flow behaviors, Indicates employee status Mobile behavior The probability of Indicates employee status Take mobile action The reward value after that; S54, job adaptation optimization module, connected with the talent flow trend modeling module, is used to optimize the job adaptation feature matrix based on the job adaptation feature matrix. Calculating job fit adjustment scores ; S55, intelligent decision output module, connected with the job suitability optimization module, is used to estimate the recruitment time based on the , Talent flow trend modeling, attrition rate and job fit adjustment scores , generate recruitment optimization strategies, employee career development suggestions and turnover risk predictions.

6. A human resource management system based on digital intelligence technology according to claim 5, characterized in that: The S7 specifically includes: S71, data encryption module, used for storing and transmitting encrypted human resources data, using national secret algorithm to encrypt human resources data, the national secret algorithm includes Asymmetric encryption, Hash Algorithms and Symmetric encryption algorithm; S72, user identity authentication module, connected to the data encryption module, used to authenticate the user based on two-factor authentication, using a one-time password and based on The digital signature is combined with the user's identity confirmation; S73, the authority management module is connected to the user identity authentication module and is used to allocate user authority based on role-based access control technology and set the access control matrix , where the access rights calculation formula is as follows: ； in, Indicates user About Resources Access rights; S74, the log audit module is connected to the authority management module, and is used to record and store user operation logs. The log data is stored in a hash chain structure to prevent log tampering: ； in, For the The hash value of the log. Represents log data, Indicates data concatenation operation. is a hash function; S75, data security compliance detection module, connected with the log audit module, is used to detect system compliance based on data integrity verification and access control strategy, using Data integrity verification method for trees, The root hash value of the tree is calculated as follows: ； in, for The root hash value of the tree, and are the hash values ​​of the left subtree and the right subtree respectively.

Images