Artificial intelligence-based talent information management optimization method and system

Abstract

The invention discloses a talent information management optimization method and system based on artificial intelligence. The method comprises the steps of talent information collection and preprocessing, talent feature modeling, post demand modeling, intelligent matching and recommendation, talent pool dynamic maintenance and visual analysis and decision support. The invention relates to the field of human resource intelligent management, in particular to a talent information management optimization method and system based on artificial intelligence. According to the scheme, the multi-dimensional talent information is collected, the talent feature vector model is constructed, and the relationship features among the talent information are constructed through the graph structure, so that the problems of insufficient feature distinction degree and lack of relationship modeling are solved; spectral time features in talent information are extracted, different features are fused, talent feature modeling is more comprehensive, and talent characteristics are described from multiple angles; and the talent feature vectors are updated by adopting a mode of regularly obtaining the latest dynamic states of the candidates, the occupational trajectory change is found in time, and the matching degree of the candidates and the posts is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of human resource intelligent management, in particular to a talent information management optimization method and system based on artificial intelligence. BACKGROUND

[0002] Traditional talent information management systems mainly rely on manual resume screening, interview evaluation and other methods, which have problems such as low efficiency, strong subjectivity and low matching degree. With the rapid development of big data and artificial intelligence technology, although some talent management systems have introduced keyword search and simple intelligent scoring functions, traditional talent feature extraction methods still have obvious limitations. These traditional methods often focus only on a few basic attributes or single-dimensional data of talents, making it difficult to fully and accurately depict the multi-dimensional characteristics of talents. At the same time, general talent information management methods cannot timely capture the changing trend of the career trajectory of candidates, resulting in reduced adaptability of talents to positions, and thus weakening the overall efficiency of talent management.

[0003] Therefore, there is an urgent need for a method and system that can combine multi-source heterogeneous data, use AI models for intelligent analysis, and achieve efficient, accurate and personalized talent management. SUMMARY

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present application provides a talent information management optimization method and system based on artificial intelligence. In view of the problem that traditional talent feature extraction methods often focus only on a few basic attributes or single-dimensional data of talents, making it difficult to fully and accurately depict the multi-dimensional characteristics of talents, the present application collects multi-dimensional talent information, constructs a talent feature vector model, constructs the relationship features between talent information through graph structure, improves the quality of feature representation, and solves the problems of insufficient feature discrimination and lack of relationship modeling. At the same time, the spectral time features in the talent information are extracted, the multi-dimensional time series data is analyzed from the frequency spectrum angle, and different features are fused to make the talent feature modeling more comprehensive and multi-angle depiction of talent traits. In view of the problem that general talent information management methods cannot timely capture the changing trend of the career trajectory of candidates, resulting in reduced adaptability of talents to positions, the present application uses the method of periodically obtaining the latest dynamics of candidates to update the talent feature vector, timely discovers the changes in the career trajectory, and improves the matching degree of candidates to positions.

[0005] The present application provides a talent information management optimization method based on artificial intelligence, which comprises the following steps:

[0006] Step S1: Talent information collection and preprocessing, collecting candidate-related information from recruitment platforms, social media and enterprise databases, and preprocessing the candidate-related information to obtain talent data;

[0007] Step S2: talent characteristic modeling, constructing a talent characteristic vector model, extracting features from talent data, capturing the candidate's career trajectory trend in combination with time series, and generating a unified dimension talent characteristic vector;

[0008] Step S3: job requirement modeling, analyzing the job description released by the enterprise, extracting core requirements, constructing a job requirement vector, and calculating the similarity with the talent characteristic vector;

[0009] Step S4: intelligent matching and recommendation, talent to job matching based on the similarity results of the job requirement vector and the talent characteristic vector, outputting the matching score and the recommended ranking list;

[0010] Step S5: dynamic maintenance of talent pool, regularly obtaining the latest dynamics of candidates, and automatically updating the talent characteristic vector;

[0011] Step S6: visual analysis and decision support, providing visual reports of talent and job matching scores and recommended ranking lists, supporting human resources for multi-dimensional screening and comparative analysis.

[0012] Further, in step S2, talent characteristic modeling specifically includes the following steps:

[0013] Step S21: data enhancement, converting talent data into enhanced data using minimum-maximum normalization and Z-score normalization;

[0014] Step S22: encoder construction, using a one-dimensional convolutional neural network to construct an encoder, inputting enhanced data, outputting a feature vector, and using a single multilayer perceptron for nonlinear projection to generate an enhanced feature vector; the one-dimensional convolutional neural network includes a convolutional layer, a max-pooling layer, a ReLU activation function, batch normalization, and Dropout;

[0015] Step S23: similarity calculation and contrast loss function, calculating the similarity and contrast loss function of the enhanced feature vector, optimizing the enhanced feature vector, and obtaining the optimized enhanced feature vector, the formula is as follows: ; ;

[0016] In the formula, and represent two enhanced feature vectors, is the transpose of , is the similarity calculation, represents the norm calculation, represents the contrast loss, is the exponential function, is a hyperparameter, is a training batch size, denotes a current sample, and is a training sample index, is a power function;

[0017] Step S24: Spatio-temporal feature extraction, constructing a graph structure according to the correlation of talent information data, and extracting time features and space features to obtain spatio-temporal relationship features of talent information;

[0018] Step S25: Spectral feature extraction, extracting spectral time features for multi-dimensional time series data in talent information data, and outputting weighted spectral time features;

[0019] Step S26: Feature fusion, fusing the optimized enhanced feature vector, spatio-temporal relationship features of talent information, and weighted spectral time features to obtain a final talent feature vector;

[0020] Further, in step S24, spatio-temporal feature extraction includes the following steps:

[0021] Step S241: Graph structure learning, constructing a graph structure according to the correlation of talent data, and optimizing the graph structure using a loss function, the formula used is as follows: ;

[0022] In the formula, denotes the similarity between nodes, and denotes a node feature vector, and denotes a node index in the graph structure, is the transpose of the weight vector, is a ReLU activation function, denotes the total number of nodes in the graph structure;

[0023] Step S242: Time and space feature extraction, using time and space attention mechanisms to extract time features and space features;

[0024] Step S243: Relationship feature extraction, using Chebyshev graph convolution and time convolution operations in the graph structure to process the time features to obtain spatio-temporal relationship features of talent information;

[0025] Further, in step S25, spectral feature extraction includes the following steps:

[0026] Step S251: Short-time Fourier transform, using discrete Fourier transform to process multi-dimensional time series data to generate a time-frequency graph;

[0027] Step S252: spectrum feature extraction, using a VGG-16 network to extract the spectrum feature of the time-frequency graph, obtaining a spectrum feature;

[0028] Step S253: time step feature, converting the spectrum feature into a time step feature;

[0029] Step S254: spectral time feature, using a gated recurrent unit combined with an attention mechanism to process the time step feature, capturing long-term dependencies in the time step feature, obtaining a weighted spectral time feature.

[0030] The application provides a talent information management optimization system based on artificial intelligence, which comprises a talent information collection module, a talent feature modeling module, a job requirement modeling module, an intelligent matching and recommendation module, a talent database dynamic maintenance module and a visual analysis and decision module.

[0031] The talent information collection module collects candidate information from a recruitment platform, social media and an enterprise database, and pre-processes the candidate information to obtain talent data.

[0032] The talent feature modeling module constructs a talent feature vector model, extracts features from the talent data, captures the career trajectory change trend of the candidate in combination with a time sequence, and generates a talent feature vector in a unified dimension.

[0033] The job requirement modeling module analyzes the job description published by the enterprise, extracts core requirements, constructs a job requirement vector, and performs similarity calculation with the talent feature vector.

[0034] The intelligent matching and recommendation module matches talents to jobs based on the similarity result of the job requirement vector and the talent feature vector, and outputs a matching score and a recommended ranking list.

[0035] The talent database dynamic maintenance module regularly obtains the latest dynamics of the candidate, and automatically updates the talent feature vector.

[0036] The visual analysis and decision module provides a visual report of the matching score and the recommended ranking list of talents and jobs, and supports multi-dimensional screening and comparative analysis of human resources.

[0037] The application has the following beneficial effects by adopting the above scheme:

[0038] (1) In view of the problem that the traditional talent feature extraction method often focuses only on a few basic attributes or single-dimensional data of talents, and it is difficult to comprehensively and accurately depict the multi-dimensional features of talents, the scheme collects multi-dimensional talent information, constructs a talent feature vector model, constructs the relationship features between talent information in the graph structure, extracts the space-time features in the talent information in the graph structure by using the time convolution and time attention mechanism and the space attention mechanism, improves the feature representation quality, solves the problems of insufficient feature discrimination and lack of relationship modeling, and extracts the spectral time features in the talent information, analyzes the multi-dimensional time series data from the frequency spectrum, and fuses the extracted different features, so that the talent feature modeling is more comprehensive, and the talent characteristics are depicted from multiple angles;

[0039] (2) In view of the problem that the general talent information management method cannot timely capture the change trend of the career track of the candidate, and the adaptability of the talent and the post is reduced, the scheme adopts the mode of periodically obtaining the latest dynamics of the candidate, updates the talent feature vector, timely discovers the change of the career track, and based on the updated talent feature vector, the system can more accurately evaluate the matching degree of the candidate and the post, so as to provide more forward-looking and accurate decision support for talent selection and post configuration, effectively improve the efficiency and effect of talent management, and ensure that the talent resources are most reasonably utilized. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 A flowchart of a talent information management optimization method based on artificial intelligence is provided for the present application;

[0041] Figure 2 A schematic diagram of a talent information management optimization system based on artificial intelligence is provided for the present application;

[0042] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0044] Embodiment one, refer to Figure 1 The present application provides a talent information management optimization method based on artificial intelligence, which comprises the following steps:

[0045] Step S1: talent information collection and preprocessing, collecting candidate-related information from recruitment platforms, social media and enterprise databases, and preprocessing candidate-related information to obtain talent data, candidate-related information including personal basic information profile, education background, work experience, skill label, project experience, job intention, hobby specialty and award situation;

[0046] Step S2: talent feature modeling, constructing a talent feature vector model, extracting features from talent data, capturing candidate career trajectory trends in combination with time series, and generating a unified dimension talent feature vector;

[0047] Step S3: job requirement modeling, using a pre-trained BERT model to analyze the job description published by the enterprise, extracting core requirements including education information, key skills and experience requirements, constructing a job requirement vector, and calculating the similarity with the talent feature vector;

[0048] Step S4: intelligent matching and recommendation, matching talents to jobs based on the similarity results of the job requirement vector and the talent feature vector, outputting matching scores and recommendation ranking lists, generating matching scores according to similarity values, and generating recommendation ranking lists from high to low, and recommending the most matched talents to the corresponding positions;

[0049] Step S5: dynamic maintenance of talent pool, regularly obtaining the latest dynamics of candidates, and automatically updating the talent feature vector;

[0050] Step S6: visual analysis and decision support, providing visual reports of talent and job matching scores and recommendation ranking lists, converting complex matching scores and recommendation ranking lists into easy-to-understand charts and graphs, supporting human resources for multi-dimensional screening and comparative analysis, including screening specific positions, viewing the matching score distribution of different candidates, and comparing the adaptability of multiple candidates in different positions.

[0051] Embodiment two, based on the above embodiment, in step S2, talent feature modeling, specifically including the following steps:

[0052] Step S21: data enhancement, using min-max normalization and Z-score normalization to convert talent data into enhanced data;

[0053] Step S22: encoder construction, using a one-dimensional convolutional neural network to construct an encoder, inputting enhanced data, outputting a feature vector, and using a single multilayer perceptron for nonlinear projection to generate an enhanced feature vector; the one-dimensional convolutional neural network includes a convolutional layer, a max pooling layer, a ReLU activation function, batch normalization and Dropout;

[0054] Step S23: Similarity calculation and contrast loss function, calculate the similarity and contrast loss function of the enhanced feature vector, optimize the enhanced feature vector, and obtain the optimized enhanced feature vector, the formula is as follows: ; ;

[0055] In the formula, and represent two enhanced feature vectors, is the transpose of , is the similarity calculation, represents the norm calculation, represents the contrast loss, is the exponential function, is the hyperparameter, is the training batch size, represents the current sample, and are the training sample indexes, is the power function;

[0056] Step S24: Spatio-temporal feature extraction, construct a graph structure according to the relevance of talent information data, and extract time features and spatial features to obtain spatio-temporal relationship features of talent information;

[0057] Step S25: Frequency spectrum feature extraction, extract spectral time features for multi-dimensional time series data in talent information data, and output weighted spectral time features;

[0058] Step S26: Feature fusion, fuse the optimized enhanced feature vector, spatio-temporal relationship features of talent information, and weighted spectral time features to obtain the final talent feature vector.

[0059] Embodiment three, based on the above embodiment, in step S24, the spatio-temporal feature extraction includes the following steps:

[0060] Step S241: Graph structure learning, construct a graph structure according to the relevance inside the talent data, and optimize the graph structure using a loss function, the formula is as follows: ;

[0061] In the formula, represents the similarity between nodes, and represent node feature vectors, and represent node indexes in the graph structure, is the transpose of the weight vector, is the ReLU activation function, denotes the total number of nodes in the graph structure; ;

[0062] wherein, denotes a loss function, denotes the sum over all node pairs , is the square of the Euclidean distance between the feature vectors of the nodes, is a regularization parameter, denotes the square of the Frobenius norm of all nodes, is the adjacency matrix of all nodes;

[0063] Step S242: Time and space feature extraction, using time and space attention mechanism to extract time and space features;

[0064] Step S243: Relationship feature extraction, using Chebyshev graph convolution and time convolution operation in the graph structure to process the time features, obtaining the spatio-temporal relationship features of talent information.

[0065] Embodiment four, based on the above embodiment, in step S25, the spectral feature extraction includes the following steps:

[0066] Step S251: Short-time Fourier transform, using discrete Fourier transform to process the multi-dimensional time series data, generating a time-frequency graph;

[0067] Step S252: Spectral feature extraction, using VGG-16 network to extract the spectral features of the time-frequency graph, obtaining the spectral features; VGG-16 network has a depth of 16 layers, extracts local features through convolution layers and ReLU activation functions, uses a maximum pooling layer for dimension reduction and feature selection, and finally outputs the spectral features;

[0068] Step S253: Time step feature, the spectral features are segmented and converted into five time step features each containing five features;

[0069] Step S254: Spectro-temporal feature, using a gated recurrent unit combined with an attention mechanism to process the time step features, capturing long-term dependencies in the time step features, obtaining weighted spectro-temporal features.

[0070] Embodiment five, based on the above embodiment, in step S254, the time step features are processed using a gated recurrent unit combined with an attention mechanism, specifically including the following steps:

[0071] Step S2541: The gated recurrent unit combined with the attention mechanism controls the flow of information through the update gate and the reset gate;

[0072] Step S2542: capturing long-term dependencies in the sequence by updating the hidden state;

[0073] Step S2543: calculating the attention score of each time step and normalizing the attention score into attention weight using the Softmax function, the formula is as follows:

[0074] In the formula, is the normalized attention weight, is the attention score, denotes the time step, is the total number of time steps, denotes the current attention score;

[0075] Step S2544: applying the attention weight to the sequence of the spectral-temporal features to obtain the weighted spectral-temporal features.

[0076] By performing the operation, the traditional talent feature extraction method often focuses only on a few basic attributes or single-dimensional data of talents, and it is difficult to comprehensively and accurately depict the multi-dimensional characteristics of talents. The present scheme collects multi-dimensional talent information, constructs a talent feature vector model, constructs the relationship characteristics between talent information through a graph structure, extracts the space-time characteristics in the talent information in the graph structure by using the time convolution and time attention mechanism and the space attention mechanism, improves the quality of feature representation, solves the problems of insufficient feature discrimination and lack of relationship modeling. At the same time, the spectral-temporal characteristics in the talent information are extracted, the multi-dimensional time series data is analyzed from the frequency spectrum, and the different extracted features are fused, so that the talent feature modeling is more comprehensive and the talent characteristics are described from multiple angles.

[0077] In the sixth embodiment, referring to Figure 2 The present application provides a talent information management optimization system based on artificial intelligence, which comprises a talent information collection module, a talent feature modeling module, a job demand modeling module, an intelligent matching and recommendation module, a talent database dynamic maintenance module and a visual analysis and decision-making module.

[0078] The talent information collection module collects candidate-related information from recruitment platforms, social media and enterprise databases, and pre-processes the candidate-related information to obtain talent data.

[0079] The talent feature modeling module constructs a talent feature vector model, extracts features from the talent data, captures the candidate career trajectory trend in combination with the time series, and generates a unified dimension talent feature vector.

[0080] ​The post demand modeling module analyzes the job description published by the enterprise, extracts the core requirements, constructs a post demand vector, and performs similarity calculation with a talent characteristic vector;

[0081] The intelligent matching and recommendation module matches talents to posts based on the similarity result of the post demand vector and the talent characteristic vector, and outputs a matching score and a recommended ranking list;

[0082] The talent pool dynamic maintenance module regularly obtains the latest dynamics of candidates and automatically updates the talent characteristic vector;

[0083] The visual analysis and decision module provides a visual report of the matching score and the recommended ranking list of talents and posts, and supports multi-dimensional screening and comparative analysis of human resources.

[0084] By performing the operation, for the general talent information management method cannot timely capture the change trend of the career track of the candidate, resulting in the problem of reducing the adaptability of talents and posts, the present scheme adopts the mode of regularly obtaining the latest dynamics of the candidate, updating the talent characteristic vector, timely discovering the change of the career track, based on the updated talent characteristic vector, the system can more accurately evaluate the matching degree of the candidate and the post, thereby providing more forward-looking and accurate decision support for talent selection and post configuration, effectively improving the efficiency and effect of talent management, and ensuring that the talent resources are most reasonably utilized.

[0085] It should be noted that in this paper, relationship 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 such actual relationship or order between the entities or operations. Moreover, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0086] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

[0087] The above describes the present application and its embodiments, which are not limited, and the drawings only show one of the embodiments of the present application, and the actual structure is not limited thereto. In general, if a person skilled in the art is inspired thereby, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution are not creative, and should belong to the protection scope of the present application.

Claims

1. A talent information management optimization method based on artificial intelligence, characterized by: The method comprises the following steps: Step S1: Talent information collection and preprocessing: collecting candidate-related information from recruitment platforms, social media, and corporate databases, and preprocessing the candidate-related information to obtain talent data; Step S2: Talent feature modeling: building a talent feature vector model, extracting features from talent data, combining time series to capture the changing trends of candidates’ career trajectories, and generating a talent feature vector with unified dimensions. Step S3: Job requirement modeling: parse the job descriptions posted by the company, extract the core requirements, construct a job requirement vector, and calculate the similarity with the talent feature vector; Step S4: Intelligent matching and recommendation, matching talents to positions based on the similarity results between the job requirement vector and the talent feature vector, and outputting the matching score and recommendation ranking list; Step S5: Dynamically maintain the talent pool, regularly obtain the latest candidate updates, and automatically update the talent feature vector; Step S6: Visual analysis and decision support, providing a visual report of the matching scores between talents and positions and the recommended ranking list, supporting human resources to conduct multi-dimensional screening and comparative analysis.

2. The talent information management optimization method based on artificial intelligence according to claim 1, characterized in that: In step S2, talent characteristics modeling is performed, specifically including the following steps: Step S21: Data enhancement, using min-max normalization and Z-score normalization to convert talent data into enhanced data; Step S22: Encoder construction, using a one-dimensional convolutional neural network to construct an encoder, input the enhanced data, output a feature vector, and use a single multi-layer perceptron for nonlinear projection to generate an enhanced feature vector; Step S23: similarity calculation and contrast loss function, calculating the similarity and contrast loss function of the enhanced feature vector, optimizing the enhanced feature vector, and obtaining the optimized enhanced feature vector; Step S24: Extracting spatiotemporal features: constructing a graph structure based on the correlation of talent information data, and extracting temporal and spatial features to obtain spatiotemporal relationship features of talent information; Step S25: Spectral feature extraction, extracting spectral-time features from the multidimensional time series data in the talent information data, and outputting weighted spectral-time features; Step S26: Feature fusion, fusing the optimized enhanced feature vector, the spatiotemporal relationship features of the talent information, and the weighted spectral-temporal features to obtain the final talent feature vector.

3. The method for optimizing talent information management based on artificial intelligence according to claim 2, characterized in that: In step S24, spatiotemporal feature extraction includes the following steps: Step S241: Graph structure learning: constructing a graph structure based on the internal correlation of talent data and optimizing the graph structure using a loss function. The formula used is as follows: ； Where, Represents the similarity between nodes, and represents the node feature vector, and Represents the node index in the graph structure, is the transpose of the weight vector, is the ReLU activation function, Represents the total number of nodes in the graph structure; Step S242: Temporal and spatial feature extraction, using the temporal and spatial attention mechanism to extract temporal features and spatial features; Step S243: Relational feature extraction, using Chebyshev graph convolution and time convolution operations in the graph structure to process the time features and obtain the spatiotemporal relationship features of the talent information.

4. An artificial intelligence-based talent information management optimization system, used to implement the artificial intelligence-based talent information management optimization method as described in claim 1, characterized in that: The system includes talent information collection module, talent characteristics modeling module, job demand modeling module, intelligent matching and recommendation module, talent pool dynamic maintenance module and visual analysis and decision-making module; The talent information collection module collects candidate-related information from recruitment platforms, social media, and corporate databases, and pre-processes the candidate-related information to obtain talent data; The talent feature modeling module constructs a talent feature vector model, extracts features from talent data, combines time series to capture the changing trends of candidates' career trajectories, and generates a talent feature vector with unified dimensions. The job requirement modeling module parses the job descriptions published by the enterprise, extracts the core requirements, constructs the job requirement vector, and calculates the similarity with the talent feature vector; The intelligent matching and recommendation module matches talents to positions based on the similarity results between the job requirement vector and the talent feature vector, and outputs a matching score and a recommended ranking list; The talent pool dynamic maintenance module regularly obtains the latest candidate dynamics and automatically updates the talent feature vector; The visual analysis and decision-making module provides a visual report of the matching scores between talents and positions and a recommended ranking list, supporting human resources to conduct multi-dimensional screening and comparative analysis.

Images