A data simulation method based on data visualization

Abstract

The application discloses a data simulation method based on data visualization, and relates to the technical field of data analysis, comprising: constructing a distributed stream processing framework, performing parallel preprocessing on a target data set to be simulated to generate a time sequence characteristic stream; converting the time sequence characteristic stream into a multi-dimensional visualization model through an adaptive mapping layer, and realizing dynamic rendering in a visualization interface; triggering a simulation engine to execute real-time simulation based on the multi-dimensional visualization model, and jointly optimizing the visualization parameters and the simulation parameters through a feedback optimization mechanism. Through the introduction of a dynamic calculation method of the job level correlation coefficient and the space-time overlap degree, the design of the time-varying adjustment factor and the hierarchical attention mechanism, and the construction of a novel data processing and visualization framework, the application realizes the accurate identification of the talent development path across different levels, the timeliness optimization of data preprocessing, and the effective sharing of the best practices among departments.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to a data simulation method based on data visualization. BACKGROUND

[0002] In the field of human resource management, data visualization and simulation technology is a key means to improve talent management decision-making efficiency. Traditional talent evaluation methods mainly rely on static index system and experience judgment, which is difficult to effectively capture the dynamic characteristics and potential laws of talent development. In recent years, with the development of distributed computing and deep learning technology, talent portrait construction and development path prediction based on large-scale time series data have gradually become a research hotspot. The current technical solution usually adopts a combination of feature engineering and dimensionality reduction visualization, which maps high-dimensional talent data to a low-dimensional visualization space, and assists talent development decision-making through simulation. However, this method still has many technical bottlenecks in data preprocessing, feature extraction and visualization mapping.

[0003] The existing technology has the following limitations when processing talent development data: first, the traditional data slicing and similarity calculation method ignores the influence of organizational levels on behavior patterns, resulting in inaccurate cross-level talent development path identification; second, the standardized processing adopts a fixed normalization method, which fails to reflect the time-dependent decay characteristics of human data, resulting in unreasonable historical data weight allocation; third, the data processing strategies of different departments are relatively independent, lacking an effective experience sharing mechanism, which limits the horizontal promotion of best practices; finally, the visualization mapping process does not fully consider the hierarchical structure characteristics of the organization, making the visualization results inconsistent with the actual organizational structure, and making it difficult to intuitively display talent distribution patterns.

[0004] Therefore, how to construct a technical framework that can effectively integrate data preprocessing, visualization mapping and simulation optimization to achieve accurate analysis and visualization of talent development data is a technical problem that needs to be solved. SUMMARY

[0005] In view of the above problems, the present application is proposed.

[0006] Therefore, the present application provides a data simulation method based on data visualization, which can solve the problems mentioned in the background art.

[0007] To solve the above technical problems, the present application provides the following technical solution: a data simulation method based on data visualization, comprising: constructing a distributed stream processing framework, parallel preprocessing the target data set to be simulated to generate a time series feature stream;

[0008] The time sequence feature flow is converted into a multi-dimensional visualization model through an adaptive mapping layer, and dynamic rendering is realized in a visualization interface; the adaptive mapping layer comprises a feature selection unit and a dimension mapping unit;

[0009] Based on the multi-dimensional visualization model, a simulation engine is triggered to perform real-time simulation, and a feedback optimization mechanism is used to jointly optimize the visualization parameters and the simulation parameters.

[0010] As a preferred scheme of the data simulation method based on data visualization, the target data set to be simulated is preprocessed in parallel to generate a time sequence feature flow, comprising the following steps:

[0011] The target data set is divided into multiple data slices according to the data correlation degree, and a data dependency graph is constructed based on the time sequence feature similarity of the data slices, wherein the nodes of the data dependency graph represent the data slices, and the edges represent the correlation strength between the slices.

[0012] The data slices are assigned to computing nodes, and the computing nodes perform standardization processing on the data slices to generate preprocessed data.

[0013] Each computing node selects a data cleaning strategy and a feature extraction method based on the time sequence feature distribution characteristics of the data slice using a cooperative learning mechanism to generate a local feature flow.

[0014] The local feature flows of each computing node are summarized, and a hierarchical attention mechanism is used to capture the multi-scale dependency relationship between the features to generate a global time sequence feature flow.

[0015] As a preferred scheme of the data simulation method based on data visualization, the time sequence feature similarity is calculated by behavior feature similarity and job level correlation coefficient.

[0016] ;

[0017] ;

[0018] wherein, denotes the time sequence behavior similarity of the employee data slice i and j , is the adaptive weight of the k th behavior feature, and denote the i th behavior feature value of the slice j and k , is a feature scale parameter, is a job level correlation coefficient, and To examine the time window, and This refers to the job level value. For job grade differences;

[0019] When the similarity between different data shards exceeds a dynamic threshold, a connection edge is established in the data dependency graph, and the weight of the connection edge is dynamically adjusted according to the difference in job level and the degree of time overlap.

[0020] The standardization process uses a time-varying adjustment factor to normalize the eigenvalues:

[0021] ;

[0022] The time-varying adjustment factor is dynamically calculated based on time decay characteristics and characteristic variability:

[0023] ;

[0024] in, These are the standardized eigenvalues. These are the original eigenvalues. and Let k be the mean and standard deviation of feature k. As a time-varying adjustment factor, For the current time, For reference time points, For time decay parameters, Let k be the coefficient of variation of feature k. The maximum coefficient of variation, These are the balancing parameters.

[0025] As a preferred embodiment of the data visualization-based data simulation method described in this invention, the collaborative learning mechanism updates the processing strategy weights based on the temporal similarity and departmental relevance among data fragments.

[0026] ;

[0027] ;

[0028] in, and They are nodes i The weights of the processing strategies before and after the update For learning rate, For nodes i The set of adjacent nodes, For departmental relevance, and Departmental codes, For departmental related parameters;

[0029] When the computing node detects a new data pattern, the processing strategy is propagated to adjacent nodes according to the department correlation degree;

[0030] The hierarchical attention mechanism establishes feature correlation through the interaction of the ability dimension and the level dimension:

[0031] ;

[0032] The feature weight is adjusted based on the ability-level matching matrix:

[0033] ;

[0034] Wherein, the attention output of the ability dimension c and the level dimension h , , , the query, key value and value matrix respectively, the attention dimension, the ability-level matching matrix, the level feature, the ability feature, the number of features, the professional field coding matrix.

[0035] As a preferred scheme of the data simulation and emulation method based on data visualization, wherein: the time sequence feature flow is converted into a multi-dimensional visualization model through an adaptive mapping layer, and dynamic rendering is realized in a visualization interface, including the following steps:

[0036] The feature selection unit performs importance evaluation on the time sequence feature flow, and filters a key feature subset based on the contribution of the feature to the talent development path;

[0037] The dimension mapping unit performs dimension reduction transformation on the key feature subset to generate a visualization feature matrix, and converts the visualization feature matrix into a multi-dimensional visualization model to perform dynamic rendering.

[0038] As a preferred scheme of the data simulation and emulation method based on data visualization, wherein: the importance evaluation is based on the information gain and time sequence stability of the feature to calculate the feature weight, and when the feature weight exceeds a dynamic screening threshold, it is included in the key feature subset:

[0039] ;

[0040] ;

[0041] wherein, is a comprehensive importance score of feature s, is an information gain score, is a balance factor, is a time series stability score, denotes the value of the feature at time point t, is a time window size;

[0042] The dimension reduction transformation maps the high-dimensional feature space to the visualization space by preserving the hierarchical relationship and functional correlation between features, and the dynamic rendering adjusts the visualization effect according to the time series changes of talent characteristics:

[0043] ;

[0044] ;

[0045] wherein, is a mapping loss function, and is the mapped low-dimensional coordinate, is a job-related coefficient, and is a job level, is the highest job level, is a job impact factor.

[0046] As a preferred solution of the data simulation and simulation method based on data visualization according to the present application, wherein: based on the multi-dimensional visualization model, a simulation engine is triggered to perform real-time simulation, and a feedback optimization mechanism is used to jointly optimize the visualization parameters and simulation parameters, including the following steps:

[0047] Based on the multi-dimensional visualization model, talent distribution characteristics and development trajectory characteristics are extracted, and simulation scene parameters are generated, including talent flow parameters and ability development parameters;

[0048] The simulation engine performs talent development simulation according to the simulation scene parameters to generate talent distribution prediction results and ability development prediction results;

[0049] Through the feedback optimization mechanism, the deviation between the prediction results and the actual observation results is compared, and the visualization parameters and simulation parameters are adjusted.

[0050] To further solve the above technical problems, the present application provides the following technical solutions: a data simulation and simulation system based on data visualization, comprising:

[0051] An organization management unit is configured to construct a distributed stream processing framework, perform parallel preprocessing on a target data set to be simulated, and generate a time-series feature stream; the organization management unit comprises:

[0052] An organization chart sub-module is configured to support dynamic setting and display of an organization chart based on the time-series feature stream; the organization chart sub-module optimizes a visual effect of a superior-inferior relationship through an adaptive mapping layer; an organization level sub-module is configured to support dynamic definition and display of an organization level based on the time-series feature stream; the organization level sub-module updates state information of the organization level in real time through the distributed stream processing framework; an organization archive sub-module is configured to support organization archiving and historical query based on the time-series feature stream; the organization archive sub-module realizes multi-dimensional visual display of archived data through the adaptive mapping layer.

[0053] An employee management unit is configured to trigger a simulation engine to perform real-time simulation based on the multi-dimensional visual model, and jointly optimize visual parameters and simulation parameters through a feedback optimization mechanism; the employee management unit comprises: a roll sheet sub-module configured to support dynamic management of an employee roll sheet, a custom roll sheet, and a historical roll sheet based on the time-series feature stream; a transaction management sub-module configured to support employee transaction management, including pre-employment management, transfer management, probation period management, and resignation management, based on the time-series feature stream.

[0054] An attendance management unit is configured to support attendance query and statistics based on the time-series feature stream; the attendance management unit comprises: a scheme management sub-module configured to support attendance scheme management, including holiday management, shift management, and shift scheduling rule management, based on the time-series feature stream; a query and statistics sub-module configured to support attendance query and statistics, including employee shift scheduling query, daily attendance report, and monthly attendance report, based on the time-series feature stream.

[0055] A salary management unit is configured to support starting salary, salary adjustment, and salary suspension operations based on the time-series feature stream; the salary management unit comprises: a starting salary adjustment sub-module configured to support starting salary, salary adjustment, and salary suspension operations based on the time-series feature stream; an accounting and approval sub-module configured to support salary accounting and approval processes based on the time-series feature stream.

[0056] The performance evaluation unit is used for supporting evaluation dimension and unit index maintenance based on the time sequence characteristic flow; the performance evaluation unit comprises: a normal distribution submodule used for supporting normal distribution scheme management based on the time sequence characteristic flow; and a result maintenance submodule used for supporting evaluation progress query and result maintenance based on the time sequence characteristic flow. A computer device comprises a memory and a processor, and the memory stores a computer program; when the processor executes the computer program, the steps of the data simulation and emulation method based on data visualization are implemented.

[0057] A computer readable storage medium stores a computer program; when the computer program is executed by a processor, the steps of the data simulation and emulation method based on data visualization are implemented.

[0058] The present application has the following advantages: the present application introduces the job level difference and time overlap degree into the data dependency graph, so that the similarity calculation better reflects the organizational level characteristics, improves the accuracy of talent development path identification, and discovers the cross-level ability transfer rule that is difficult to be perceived by the traditional method; the time-varying adjustment factor is used for standardization processing, the data timeliness and feature stability are dynamically balanced, the historical data reference value is reasonably decayed, and the interference of outdated experience on current decision is effectively avoided; the feature comprehensive importance evaluation mechanism is introduced, the distinguishing ability and time sequence stability of the feature are organically combined, and the stable feature that is easily ignored by the traditional method but has unique value for predicting the long-term development potential of employees is discovered; through the synergistic effect of the department correlation degree and the ability-level matching matrix, the data processing experience is selectively shared while the department individuality is maintained, the horizontal transmission of best practices is promoted, and blind copying of inappropriate experience is avoided; the mapping mechanism based on the job level correlation makes the visualization result better reflect the organizational level structure, helps managers intuitively understand the talent echelon construction status, and the design of the dynamic screening threshold makes the feature selection process adaptively adjusted according to the characteristics of different job level sequences, and the practicability and adaptability of the scheme are improved. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0060] Figure 1 The present application proposes a data simulation and emulation method based on data visualization, and the overall flowchart is shown in the figure.

[0061] Figure 2A computer device diagram in a data simulation method based on data visualization. DETAILED DESCRIPTION

[0062] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present application.

[0063] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced without the specific details that are set forth in the description, and it is understood that the present application is not limited to the embodiments described herein and can be practiced with or without the other apparatuses, systems, components, compositions, etc. described.

[0064] Embodiment 1, Reference Figure 1 For an embodiment of the present application, a data simulation method based on data visualization is provided.

[0065] S1: Construct a distributed stream processing framework, and perform parallel preprocessing on a target data set to be simulated to generate a time series feature stream.

[0066] Specifically, the distributed stream processing framework includes a data cleaning module and a feature extraction module.

[0067] S1.1: Divide the target data set into multiple data shards according to the data correlation degree, and construct a data dependency graph based on the time series feature similarity of the data shards.

[0068] Specifically, the nodes of the data dependency graph represent the data shards, and the edges represent the correlation strength between the shards. Specifically, the time series feature similarity is calculated through the behavior feature similarity and the job level correlation coefficient. When the similarity between different data shards exceeds a dynamic threshold, a connection edge is established in the data dependency graph, and the weight of the connection edge is dynamically adjusted according to the job level difference and the time overlap degree.

[0069] The calculation formula of the time series feature similarity is:

[0070] ;

[0071] ;

[0072] wherein, denotes the time series behavior similarity of the employee data shards i and j , and is the kan adaptive weight of a behavior feature, and represent the first behavior feature value of a slice i and j k , respectively. is a feature scale parameter, is a job level correlation coefficient, and is a time window for observation, and is a job level value, is a job level difference parameter.

[0073] It should be noted that in the human resource management scenario, the dynamic threshold is an adaptive similarity determination standard, which is calculated by statistical analysis of historical talent development trajectory data. The dynamic threshold adopts a sliding time window mechanism, and based on the behavior feature distribution and job level migration mode of the employees in the window, the employees are divided into multiple similarity clusters by the K-means clustering method, and the weighted average of the minimum distance between clusters is taken as the benchmark threshold, and the benchmark threshold is dynamically adjusted according to the organizational structure level and department attribute. The preferred benchmark threshold range is 0.6-0.8, higher level employees tend to choose a higher threshold to ensure more accurate similarity matching, while lower level employees use a relatively loose threshold to obtain more development reference samples. The dynamic threshold in the present application not only considers the hierarchical characteristics of the organization, but also maintains the flexibility of talent development path recognition, effectively balancing the accuracy and generalization of similarity determination. In addition, the traditional employee behavior similarity calculation often only focuses on the behavior data itself, ignoring the influence of the organizational level on the behavior mode. The introduction of the job level correlation coefficient not only considers the time overlap degree, but also describes the influence of the organizational level through the job level difference item, which is helpful to discover the talent development path and ability migration law across levels.

[0074] S1.2: Assign a computing node to the data slice, and the computing node performs standardization processing on the data slice to generate preprocessed data.

[0075] Specifically, the standardization processing adopts a time-varying adjustment factor to normalize the feature value, and the time-varying adjustment factor is dynamically calculated based on the time decay characteristic and the feature variation degree, and is used to balance the timeliness and stability of the data.

[0076] The standardization processing adopts the following adaptive normalization formula:

[0077] ;

[0078] ;

[0079] wherein, ​is a standardized feature value, is an original feature value, and is the mean and standard deviation of feature k, is a time-varying adjustment factor, is the current time, is a reference time point, is a time decay parameter, is the coefficient of variation of feature k, is the maximum coefficient of variation, is a balance parameter.

[0080] It should be noted that in human resource data processing, the traditional standardization method does not consider the timeliness and volatility of data, while personnel performance data has obvious time decay characteristics. The introduction of time-varying adjustment factor not only considers the timeliness of data, but also reflects the stability of different indicators through the coefficient of variation, which helps to more accurately evaluate the continuous performance of employees.

[0081] S1.3: Based on the time sequence feature distribution characteristics of data shards, each computing node adopts a collaborative learning mechanism to select data cleaning strategies and feature extraction methods, and generates local feature streams.

[0082] Specifically, the collaborative learning mechanism updates the processing strategy weight according to the time sequence feature similarity and department relevance between data shards:

[0083] ;

[0084] ;

[0085] wherein, and are the processing strategy weights of node i after and before updating, is the learning rate, is a set of nodes adjacent to node i , is the department relevance, and are department codes, is a department association parameter.

[0086] When the computing node detects a new data pattern, the processing strategy is propagated to the adjacent nodes according to the degree of department association.

[0087] It should be noted that in human resource management, the data processing strategies of different departments often differ. The introduction of department relevance This makes the data processing experience of similar departments can be shared preferentially, while maintaining cross-departmental learning, which helps to maintain the department characteristics while promoting the horizontal spread of best practices.

[0088] S1.4: Aggregate local feature streams of each computing node, capture multi-scale dependency between features through hierarchical attention mechanism, and generate global time series feature stream.

[0089] Specifically, the hierarchical attention mechanism establishes feature association through the interaction of ability dimension and level dimension:

[0090] ;

[0091] Adjust the feature weight based on the ability-level matching matrix:

[0092] ;

[0093] where, represents the attention output of the ability dimension c and the level dimension h , , , query, key value and value matrix respectively, attention dimension, ability-level matching matrix, level feature, ability feature, feature quantity, professional field coding matrix.

[0094] The ability-level matching matrix is dynamically updated with professional field features. In talent development and ability assessment, traditional feature extraction methods often separate the ability dimension and the level dimension. By introducing the ability-level matching matrix , by capturing the differentiated requirements of different levels for each ability, it can automatically adjust the importance weight of the feature, which helps to build a more realistic talent portrait and development path for the organization.

[0095] It should be noted that in the field of human resource data visualization and simulation, there are three main technical problems in the traditional method: first, only the behavior data itself is considered in data slicing and similarity calculation, ignoring the influence of organizational level characteristics on behavior patterns, resulting in inaccurate identification of cross-level talent development path; second, the fixed normalization method is adopted in standardization processing, which cannot reflect the time decay characteristics of human data, resulting in unreasonable weight allocation of historical data; finally, the data processing strategies of different departments are independent, lacking effective experience sharing mechanism, limiting the horizontal promotion of best practices. The application introduces the dynamic calculation method of job level correlation coefficient and space-time overlap, establishes a data dependency graph based on organizational level, and realizes the accurate identification of cross-level talent development path; by designing the time-varying adjustment factor, the time decay characteristics and feature variation degree are introduced into the standardization processing process, improving the adaptability of data preprocessing; by constructing hierarchical attention mechanism and department correlation calculation model, the differential sharing of data processing experience between different departments is realized, and the effective dissemination of best practices based on maintaining department characteristics is promoted.

[0096] Preferably, the S1 step of the application has the following technical effects: first, by integrating job level difference and time overlap factor in the data dependency graph, the similarity calculation better reflects the organizational level characteristics, not only improves the accuracy of talent development path identification, but also discovers the cross-level ability transfer rule that the traditional method cannot detect; second, the standardization processing method based on time-varying adjustment factor dynamically balances the data timeliness and feature stability, so that the reference value of historical data is reasonably decayed, avoiding the interference of outdated experience on current decision; third, through the synergistic effect of department correlation and ability-level matching matrix, the selective sharing of data processing experience is realized while maintaining the department individuality, which not only promotes the horizontal dissemination of best practices, but also avoids the blind replication of inappropriate experience.

[0097] S2: convert the time series feature stream through the adaptive mapping layer into a multi-dimensional visualization model, and realize dynamic rendering in the visualization interface.

[0098] Specifically, the adaptive mapping layer includes a feature selection unit and a dimension mapping unit.

[0099] S2.1: perform importance evaluation on the time series feature stream through the feature selection unit, and select a key feature subset based on the contribution of the feature to the talent development path.

[0100] Specifically, the importance evaluation is based on the information gain and time series stability of the feature to calculate the feature weight, and when the feature weight exceeds the dynamic screening threshold, the feature is included in the key feature subset:

[0101] ;

[0102] ;

[0103] wherein, is the comprehensive importance score of feature s, is the information gain score, is the balance factor, is the time series stability score, denotes the value of the feature at time point t, is the time window size.

[0104] Preferably, the dynamic screening threshold in the present application is a feature screening standard value based on the dynamic adjustment of historical data distribution and current business demand, which is obtained by statistical analysis of the feature importance distribution of historical excellent talent samples. Specifically, the median of the sample feature importance score is taken as the reference value, and adaptive adjustment is made according to the data distribution characteristics of different job sequences. For example, for the talent evaluation of management sequence, the threshold tends to select a higher reference value due to the relatively stable feature index system, while for the technical sequence, the threshold is relatively flexible considering the rapid update of skill indicators. In practice, the preferred range of dynamic screening threshold is usually set to 0.8-1.2 times of the reference value, and will be dynamically updated according to the department size, organizational level and evaluation period, so as to ensure the representativeness of the screened features while also taking into account the individual needs of different business scenarios. Through this dynamic threshold mechanism, feature selection bias caused by fixed threshold is avoided, and the pertinence and timeliness of feature screening are improved.

[0105] It should be noted that the present application combines information gain and time series stability through feature importance calculation method, so that feature selection considers both discriminability and stability, which helps to screen out more suitable feature indicators for talent evaluation.

[0106] S2.2: The dimension mapping unit performs dimension reduction transformation on the key feature subset to generate a visual feature matrix, and converts the visual feature matrix into a multi-dimensional visual model to perform dynamic rendering;

[0107] Specifically, dimension reduction transformation maps high-dimensional feature space to visual space by preserving the hierarchical relationship and functional correlation between features, and dynamic rendering adjusts the visualization effect according to the time series change of talent features:

[0108] ;

[0109] ;

[0110] wherein, is the mapping loss function, and are the mapped low-dimensional coordinates, is a job level correlation coefficient, and is a job level, is a highest job level, is a job level impact factor.

[0111] Compared with the traditional dimension reduction method, the loss function adopted by the present application only retains the key job level correlation constraint, so that the visualization result can more clearly show the hierarchical structure of the organization, while reducing the computational complexity.

[0112] The traditional data mapping and visualization method fails to effectively combine information gain and time stability in the feature selection process, resulting in selected features with high discrimination but poor time stability, which cannot accurately reflect the continuous development characteristics of talents. Moreover, the visualization mapping process ignores the hierarchical structure characteristics of the organization, making the visualization result inconsistent with the actual organizational structure, and it is difficult to intuitively show the talent distribution law. Therefore, the present application introduces a feature comprehensive importance evaluation mechanism to organically combine the discrimination ability and time stability of the features, and uses a mapping loss function based on job level correlation, so that the dimension reduction result can maintain the organizational hierarchical structure. Such settings have the following effects: first, through the combined evaluation of feature importance and time stability, some stability features that are easily overlooked by traditional methods are found. These features may not have high instantaneous discrimination, but they have unique value in predicting the long-term development potential of employees; second, the mapping mechanism based on job level correlation enables the visualization result to better reflect the organizational hierarchical structure, helping managers intuitively understand the talent echelon construction situation; finally, the design of the dynamic threshold selection makes the feature selection process adaptively adjust according to the characteristics of different job level sequences, improving the practicality and adaptability of the scheme.

[0113] S3: Trigger the simulation engine to perform real-time simulation based on the multi-dimensional visualization model, and perform joint optimization of the visualization parameters and simulation parameters through the feedback optimization mechanism.

[0114] S3.1: Extract talent distribution features and development trajectory features based on the multi-dimensional visualization model to generate simulation scene parameters.

[0115] Specifically, the simulation scene parameters include talent flow parameters and ability development parameters. The talent distribution features reflect the current talent structure state of the organization, and the development trajectory features describe the historical talent development path, and the parameter generation process is based on the historical data distribution features.

[0116] S3.2: The simulation engine performs talent development simulation based on the simulation scene parameters to generate talent distribution prediction results and ability development prediction results.

[0117] Specifically, the talent development simulation adopts the Monte Carlo simulation method to generate the prediction result distribution through multiple sampling calculations.

[0118] S3.3: Adjust the visualization parameters and simulation parameters by feedback optimization mechanism to the deviation between the prediction results and the actual observation results.

[0119] Specifically, when the deviation between the prediction results and the actual observation results exceeds the expected error range, the parameter values are updated based on the gradient descent method, and the updated parameters are fed back to the visualization model and the simulation engine.

[0120] In summary, the present application integrates job level difference and time overlap factor in data dependency graph, making similarity calculation better reflect organizational level characteristics, improving the accuracy of talent development path identification, and discovering cross-level ability transfer rules that traditional methods cannot detect; using the standardization processing method of time-varying adjustment factor, through dynamic balance of data timeliness and feature stability, the reference value of historical data is reasonably decayed, effectively avoiding the interference of outdated experience on current decision; introducing feature comprehensive importance evaluation mechanism, combining the distinguishing ability of features with time stability, discovering stable features that traditional methods easily ignore but have unique value for predicting long-term development potential of employees; through the synergistic effect of department correlation and ability-level matching matrix, the selectivity of data processing experience is shared while maintaining department individuality, which not only promotes the horizontal spread of best practices, but also avoids blind copying of inappropriate experience; the mapping mechanism based on job level relevance makes the visualization results better reflect the organizational level structure, helping managers intuitively understand the talent echelon construction situation, and the design of dynamic screening threshold makes the feature selection process adaptively adjust according to the characteristics of different job level sequences, improving the practicality and adaptability of the scheme.

[0121] Embodiment 2, refer to Figure 2 For an embodiment of the present application, a data simulation and simulation system based on data visualization is provided, comprising: an organization management unit for constructing a distributed stream processing framework, performing parallel preprocessing on a target data set to be simulated, and generating a time series feature stream; the organization management unit comprises:

[0122] An organizational chart sub-module for supporting dynamic setting and display of an organizational chart based on the time series feature stream; the organizational chart sub-module optimizes the visualization effect of the superior-inferior relationship through an adaptive mapping layer;

[0123] An organization level sub-module for supporting dynamic definition and display of an organization level based on the time series feature stream; the organization level sub-module updates the state information of the organization level in real time through the distributed stream processing framework;

[0124] An organization archive sub-module for supporting organization archiving and historical query based on the time series feature stream; the organization archive sub-module realizes multi-dimensional visualization display of archived data through the adaptive mapping layer;

[0125] An employee management unit is configured to trigger a simulation engine to perform real-time simulation based on the multi-dimensional visualization model and to jointly optimize the visualization parameters and the simulation parameters through a feedback optimization mechanism. The employee management unit includes:

[0126] A name list sub-module is configured to support dynamic management of employee name lists, custom name lists and historical name lists based on the time sequence characteristic flow.

[0127] A transaction management sub-module is configured to support employee transaction management, including pre-employment management, transfer management, probation period management and resignation management, based on the time sequence characteristic flow.

[0128] An attendance management unit is configured to support attendance query and statistics based on the time sequence characteristic flow. The attendance management unit includes:

[0129] A scheme management sub-module is configured to support attendance scheme management, including holiday management, shift management and shift scheduling rule management, based on the time sequence characteristic flow.

[0130] A query and statistics sub-module is configured to support attendance query and statistics, including employee shift scheduling query, daily attendance report and monthly attendance report, based on the time sequence characteristic flow.

[0131] A salary management unit is configured to support starting salary, salary adjustment and salary suspension operations based on the time sequence characteristic flow. The salary management unit includes:

[0132] A starting salary and salary adjustment sub-module is configured to support starting salary, salary adjustment and salary suspension operations based on the time sequence characteristic flow.

[0133] An accounting and approval sub-module is configured to support salary accounting and approval processes based on the time sequence characteristic flow.

[0134] A performance evaluation unit is configured to support evaluation dimension and unit index maintenance based on the time sequence characteristic flow. The performance evaluation unit includes:

[0135] A normal distribution sub-module is configured to support normal distribution scheme management based on the time sequence characteristic flow.

[0136] A result maintenance sub-module is configured to support evaluation progress query and result maintenance based on the time sequence characteristic flow.

[0137] It should be noted that the system parallelly preprocesses employee behavior data through a distributed stream processing framework, generates a time series feature stream, and combines an adaptive mapping layer and a feedback optimization mechanism to realize dynamic management and accurate prediction of organizational structure, employee affairs, attendance management, salary accounting, performance evaluation, and other aspects. In practical applications, the system first comprehensively sorts and dynamically sets the organizational structure of the enterprise through the organizational management unit. The organizational structure chart submodule utilizes the time series feature stream to support dynamic updating and display of the organizational structure chart, and simultaneously optimizes the visualization effect of the superior-inferior relationship through the adaptive mapping layer, enabling managers to intuitively understand the changing trend of the organizational structure; the organizational level submodule updates the state information of the organizational level in real time based on the time series feature stream, supporting dynamic definition and display of the organizational state of different levels; the organizational archive submodule displays the historical evolution process of the organization through multi-dimensional visualization of historical data, helping managers query and analyze the historical evolution process of the organization.

[0138] In terms of employee management, the employee management unit realizes comprehensive management of the employee roster and affairs. The roster submodule supports dynamic management of the employee roster, custom roster, and historical roster based on the time series feature stream, ensuring the accuracy and timeliness of employee information; the affair management submodule covers the entire process management from pre-employment to resignation, including key links such as transfer, probation period, and resignation, predicts the behavior trajectory and development potential of employees through the time series feature stream, and provides decision support for management. For example, in pre-employment management, the system can predict the employee's entry time and job suitability based on historical behavior data, and display the prediction results through a visual interface, thereby helping enterprises make personnel arrangements in advance.

[0139] The attendance management unit realizes fine management of the enterprise's attendance system through the scheme management submodule and the query and statistics submodule. The scheme management submodule supports holiday management, shift management, and shift scheduling rule management, and can develop personalized attendance schemes according to the needs of different regions and departments; the query and statistics submodule supports employee shift scheduling query, daily attendance report, and monthly attendance report, and visually displays key indicators such as employee attendance rate and tardiness rate through multi-dimensional visualization models, providing data support for management. For example, when the abnormal rate of attendance in a certain department significantly increases, the system will automatically trigger the early warning mechanism and remind the relevant personnel through the visual interface to take timely measures.

[0140] The salary management unit realizes efficient management of the salary system through a starting salary and salary adjustment submodule and an accounting and approval submodule. The starting salary and salary adjustment submodule supports starting salary, salary adjustment, and salary suspension operations based on time sequence characteristics flow, can automatically generate a new salary plan according to changes in the job level and post adjustment of employees, and display the salary change trend through a visual interface; the accounting and approval submodule is responsible for the salary accounting and approval process, quickly processes large-scale salary data through a distributed flow processing framework, and uses a feedback optimization mechanism to improve the accuracy of accounting and the efficiency of approval. For example, when an employee is promoted or transferred, the system will automatically adjust his / her salary plan and display the adjusted salary composition through a visual interface.

[0141] The performance evaluation unit realizes scientific management of the performance evaluation system through a normal distribution submodule and a result maintenance submodule. The normal distribution submodule supports normal distribution scheme management based on time sequence characteristics flow, and can flexibly adjust the evaluation standard according to the actual situation of the enterprise; the result maintenance submodule supports evaluation progress query and result maintenance, and visually displays the evaluation results and development trend of employees through a multi-dimensional visual model. For example, in the annual evaluation, the system can automatically generate a normal distribution scheme that meets the needs of the enterprise according to the time sequence characteristics flow, and display the evaluation results through a visual interface to help management understand the performance of employees.

[0142] In summary, through the collaborative work of multiple functional units, the system successfully solves the limitations of traditional HR management systems in data preprocessing, feature extraction, and visual mapping, significantly improves the accuracy and decision-making efficiency of human resource management, and provides strong technical support for the talent development and organization construction of enterprises.

[0143] Embodiment 3, with reference to Figure 2 For an embodiment of the present application, which is different from the previous embodiment, the function, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make a computer device (which can be a personal computer, server, or network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various program code storage media.

[0144] The logic and / or steps represented in the flow diagrams or otherwise described herein, for example, can be considered as a sequence of executable instructions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this specification, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.

[0145] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that can be later executed by a computer.

[0146] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: a discrete logic circuit having logic gates for implementing logic functions upon data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), or the like.

[0147] It should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application but not limit the present application, and although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A data simulation method based on data visualization, characterized in that, include: A distributed stream processing framework is constructed to perform parallel preprocessing on the target dataset to be simulated, generating a time-series feature stream, including: The target dataset is divided into multiple data pieces according to the degree of data correlation. A data dependency graph is constructed based on the temporal feature similarity of the data pieces. The nodes of the data dependency graph represent data pieces, and the edges represent the correlation strength between the pieces. A computing node is allocated to the data shard, and the computing node performs standardization processing on the data shard to generate preprocessed data; Based on the temporal distribution characteristics of the data shards, each computing node uses a collaborative learning mechanism to select a data cleaning strategy and feature extraction method to generate a local feature stream; The local feature flows of each computing node are aggregated, and the multi-scale dependencies between features are captured through a hierarchical attention mechanism to generate a global temporal feature flow. The time-series feature stream is converted into a multi-dimensional visualization model through an adaptive mapping layer, and dynamically rendered in the visualization interface, including: The importance of the time-series feature stream is evaluated by the feature selection unit, and a subset of key features is selected based on the contribution of the features to the talent development path. The dimension mapping unit performs dimensionality reduction transformation on the key feature subset to generate a visualization feature matrix, and converts the visualization feature matrix into a multi-dimensional visualization model for dynamic rendering; The adaptive mapping layer includes a feature selection unit and a dimension mapping unit; The simulation engine is triggered to perform real-time simulation based on the multi-dimensional visualization model, and the visualization parameters and simulation parameters are jointly optimized through a feedback optimization mechanism. The joint optimization of visualization and simulation parameters through feedback optimization mechanism specifically involves comparing the deviation between the prediction results and the actual observation results. When the deviation between the prediction results and the actual observation results exceeds the expected error range, the parameter values ​​are updated based on the gradient descent method, and the updated parameters are fed back to the visualization model and simulation engine.

2. The data simulation method based on data visualization as described in claim 1, characterized in that: The temporal feature similarity is calculated using behavioral feature similarity and job level correlation coefficient: Among them, R ij ω represents the temporal behavioral similarity between employee data shards i and j. k x is the adaptive weight for the k-th behavioral feature. ik and x jk δ represents the k-th behavior feature value of slices i and j respectively. k φ is the characteristic scale parameter. ij T is the job level correlation coefficient. i and T j To examine the time window, p i and p j θ represents the job level value. ij For job grade differences; When the similarity between different data shards exceeds a dynamic threshold, a connection edge is established in the data dependency graph, and the weight of the connection edge is dynamically adjusted according to the difference in job level and the degree of time overlap. The standardization process uses a time-varying adjustment factor to normalize the eigenvalues: The time-varying adjustment factor is dynamically calculated based on time decay characteristics and characteristic variability: Among them, Z ik The standardized eigenvalues, x ik μ is the original eigenvalue. k and σ k Let β be the mean and standard deviation of feature k. k (t) is the time-varying adjustment factor, where t is the current time. r Let τ be the time decay parameter, and v be the reference time point. k v is the coefficient of variation of behavioral characteristic k. max λ is the maximum coefficient of variation, and λ is the equilibrium parameter.

3. The data simulation method based on data visualization as described in claim 2, characterized in that: The collaborative learning mechanism updates the processing strategy weights based on the temporal feature similarity and departmental relevance between data fragments: in, and Let be the processing policy weights of node i before and after the update, respectively; η be the learning rate; N(i) be the set of nodes adjacent to node i; and D be the learning rate. ij For departmental relevance, d i and d j The department code is ρ, and the department association parameter is ρ. When the computing node detects a new data pattern, it propagates the processing strategy to neighboring nodes according to the degree of departmental correlation. The hierarchical attention mechanism establishes feature associations through the interaction of capability and hierarchy dimensions: Adjusting feature weights based on the capability-hierarchy matching matrix: Among them, A c,h Q represents the attention output of the capability dimension c and the hierarchy dimension h. c,h K c,h V c,h These represent query, key-value, and value matrices, respectively, d dim For the attention dimension, M c,h For the capability-hierarchy matching matrix, h i For hierarchical features, c i Here, n represents the number of features, and E represents the capability feature. c,h Encoding matrix for professional fields.

4. The data simulation method based on data visualization as described in claim 3, characterized in that: The importance assessment calculates feature weights based on feature information gain and temporal stability. When a feature weight exceeds a dynamic screening threshold, it is included in the subset of key features. I s =α·G s +(1-α)·S s ; Among them, I s G is the comprehensive importance score of feature s. s S is the information gain fraction, α is the balance factor, and S is the information gain fraction. s For time series stability score, y t This represents the value of the feature at time point t, where N is the size of the time window; The dimensionality reduction transformation maps the high-dimensional feature space to the visualization space by preserving the hierarchical relationship and functional relevance between features. The dynamic rendering adjusts the visualization effect according to the temporal changes of talent features. L m =∑ i,j (1+P ij )·||U i -IN j || 2 ; Among them, L m For the mapping loss function, U i and U j P represents the mapped low-dimensional coordinates. ij g is the job level correlation coefficient. i and g j For job level, g max κ represents the highest job level, and κ is the job level influence factor.

5. The data simulation method based on data visualization as described in claim 4, characterized in that: The simulation engine is triggered to perform real-time simulation based on the multi-dimensional visualization model, including the following steps: Based on the multidimensional visualization model, talent distribution characteristics and development trajectory characteristics are extracted to generate simulation scenario parameters, which include talent mobility parameters and capability development parameters. The simulation engine performs talent development simulation based on the simulation scenario parameters, generating talent distribution prediction results and capability development prediction results.

6. A data simulation system based on data visualization, based on the data simulation method based on data visualization according to any one of claims 1 to 5, characterized in that: include, The organization and management unit is used to build a distributed stream processing framework, perform parallel preprocessing on the target dataset to be simulated, and generate a time-series feature stream. The organizational management unit includes: The organizational chart submodule is used to support the dynamic setting and display of organizational charts based on the time-series feature stream; the organizational chart submodule optimizes the visualization effect of superior-subordinate relationships through an adaptive mapping layer; The organization-level submodule is used to support the dynamic definition and display of organization levels based on the time-series feature stream; the organization-level submodule updates the organization-level status information in real time through a distributed stream processing framework. The organization archiving submodule is used to support organization archiving and historical querying based on the time-series feature stream; the organization archiving submodule realizes multi-dimensional visualization of archived data through an adaptive mapping layer; The employee management unit is used to trigger the simulation engine to execute real-time simulation based on the multi-dimensional visualization model, and to jointly optimize the visualization parameters and simulation parameters through a feedback optimization mechanism; the employee management unit includes: The roster submodule is used to support dynamic management of employee rosters, custom rosters, and historical rosters based on the time-series feature stream. The transaction management submodule is used to support employee transaction management based on the time-series characteristic stream, including pre-employment management, transfer management, probation period management, and resignation management. An attendance management unit is configured to support attendance query and statistics based on the time-series feature stream; the attendance management unit includes: The scheme management submodule is used to support attendance scheme management based on the time-series feature stream, including holiday management, shift management, and scheduling rule management. The query and statistics submodule is used to support attendance query and statistics based on the time-series feature stream, including employee shift schedule query, daily attendance report, and monthly attendance report; The payroll management unit is used to support start-up salary, salary adjustment, and salary suspension operations based on the time-series feature stream; the payroll management unit includes: The starting salary and salary adjustment submodule is used to support starting salary, salary adjustment, and salary suspension operations based on the time-series feature stream. The accounting and approval submodule is used to support salary accounting and approval processes based on the time-series feature flow. The performance appraisal unit is used to support the maintenance of appraisal dimensions and unit indicators based on the time-series feature stream; the performance appraisal unit includes: The normal distribution submodule is used to support normal distribution scheme management based on the time-series feature stream; The result maintenance submodule is used to support assessment progress query and result maintenance based on the time-series feature stream.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the data simulation method based on data visualization as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the data simulation method based on data visualization as described in any one of claims 1 to 5.

Images