Enterprise data intelligent analysis system based on deep learning

Abstract

The invention discloses an enterprise data intelligent analysis system based on deep learning, which relates to the technical field of data analysis, and realizes real-time acquisition and transmission of data through a data acquisition and processing module, division of department areas and integration of a staff performance system, a training system and an attendance system through multiple APIs (Application Program Interface); the effective data analysis module analyzes performance data and calculates a performance index Jxzs, the loss prediction analysis module predicts employee loss in combination with historical demission data and a performance evaluation result, the recruitment demand analysis module predicts recruitment demands in advance and calculates a recruitment demand index Zpzs, and the learning model optimization module optimizes a model and updates a data set. The real-time performance and the accuracy of the model are ensured; the system evaluates each index by presetting a plurality of thresholds, generates a corresponding evaluation result and proposes improvement measures, thereby effectively reducing the employee loss risk, and improving the recruitment efficiency and employee satisfaction.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to an enterprise data intelligent analysis system based on deep learning. Background Technology

[0002] Enterprise data intelligence analytics systems integrate big data, artificial intelligence, and machine learning technologies to provide real-time data processing and in-depth analysis, helping businesses identify trends, optimize decision-making, and improve efficiency. The system includes modules for data acquisition, storage, analysis, and visualization, and is widely used in market analysis, customer management, supply chain optimization, and other fields, significantly improving enterprise competitiveness and operational effectiveness.

[0003] In invention patent application number 202410220067.X, an enterprise data analysis and evaluation system is disclosed, belonging to the field of enterprise management and data analysis technology. It includes: a data acquisition module for collecting internal enterprise data; a data integration module for classifying and integrating the collected internal enterprise data; a data transmission module for transmitting the integrated enterprise data, the data transmission module being connected to the data integration module; a data storage module for storing internal enterprise data, the data storage module being connected to the data transmission module; a data analysis module for analyzing the internal enterprise data to discover patterns and trends in the data; and a data evaluation module for evaluating the analysis results to determine the enterprise's operational status and development trends. This invention, by integrating data acquisition, data processing, data analysis, and data evaluation functions, can comprehensively, accurately, and quickly analyze and evaluate enterprise data, providing strong support for enterprise decision-making.

[0004] However, when the enterprise data analysis and evaluation system mentioned in this invention is applied to enterprise human resource management, the data is from diverse sources, including employee performance, training records, and attendance data. Inaccuracies or inconsistencies can easily occur during the data collection process, affecting the reliability of the analysis results. Furthermore, it cannot accurately predict employee performance, turnover rate, and recruitment needs. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a deep learning-based enterprise data intelligent analysis system, solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based enterprise data intelligent analysis system, comprising a data acquisition and processing module, a model building and deployment module, a performance data analysis module, a churn prediction and analysis module, a recruitment demand analysis module, and a learning model optimization module;

[0007] The data acquisition and processing module is used to collect internal human resources-related data; it integrates employee performance system, training system and attendance system through API data interface, and collects and transmits data in real time using stream processing tools.

[0008] The model building and deployment module is used to build and deploy deep learning models; it selects convolutional neural networks, trains and evaluates them, and finally deploys them.

[0009] The performance data analysis module is used to collect and analyze performance data; to analyze the performance data using a deep learning model, obtain the performance index Jxzs and evaluate it; and to generate performance evaluation results.

[0010] The turnover prediction and analysis module is used to predict employee turnover based on performance evaluation results. By collecting historical turnover data and combining it with performance evaluation results, and then using a deep learning model for analysis, high-risk employees are identified, turnover prediction results are generated, and employee retention strategies are formulated.

[0011] The recruitment demand analysis module is used to predict recruitment demand in advance based on the turnover prediction report. It extracts historical recruitment data and historical employee turnover data, and performs cleaning and feature extraction to construct a recruitment demand data set. At the same time, it calculates and evaluates the recruitment demand index Zpzs, and conducts recruitment based on the evaluation results.

[0012] The learning model optimization module is used to iterate the learning model on a fixed-period dataset, including iterating on the dataset with fixed-period updates.

[0013] Preferably, the data acquisition and processing module divides the enterprise into regions based on the number of departments and labels them accordingly, specifically as QY1, QY2, QY3, ..., QY n n represents the number of segmented regions; and performance data for each department's region is collected separately.

[0014] Simultaneously, multiple API interfaces are used to connect to the employee performance system, training system, and attendance system of each department / region. Statistical analysis is performed on the information collected from these systems to obtain sets of employee performance-related statistical information, training-related statistical information, and attendance record statistical analysis information. Secondly, retry and data persistence mechanisms are established during data transmission to prevent data loss. Finally, a parallel processing strategy is adopted to process multiple data streams in parallel. The data is then cleaned, preprocessed, and stored.

[0015] Preferably, the model building and deployment module is used to select a convolutional neural network as the core algorithm based on the enterprise's human resources data, and divide the data collected by the data acquisition and processing module into training set, validation set and test set according to the proportion, which are used for model training, model tuning and final model evaluation respectively; the learning model is deployed through containerization technology and put into use to monitor and evaluate the enterprise's performance data, turnover data and recruitment demand data in real time.

[0016] Preferably, the performance data analysis module includes a performance analysis unit and a performance evaluation unit; the performance analysis unit performs dimensionless processing on all performance data of each department, constructs a performance data set, and calculates the performance index Jxzs using the following formula:

[0017]

[0018] In the formula, Gxl represents work efficiency, Rwl represents task completion rate, Pjt represents average delivery time difference, Mcl represents monthly output, Gzl represents work quality, Tsl represents complaint rate, Jsl represents promotion rate, Lrl represents profit margin, and Cql represents attendance rate.

[0019] Preferably, the performance evaluation unit uses a preset performance threshold Q. 1 and performance threshold Q 2 The performance index Jxzs is evaluated; and the performance threshold Q is... 1 >Performance threshold Q 2 The specific assessment content is as follows:

[0020] If the performance index Jxzs > the performance threshold Q 1 The first performance result is generated, indicating that the department's performance is satisfactory, including satisfactory work efficiency Gxl, work quality Gzl, and attendance rate Cql.

[0021] If the performance threshold Q 1 ≥Performance Index Jxzs≥Performance Threshold Q 2 The second performance result is generated. If there are non-compliance items in the department, including the department's work efficiency Gxl, work quality Gzl and attendance rate Cql, the non-compliance items are marked and sent to the relevant departments for reminders. The item with the lowest value is analyzed and improvement measures are implemented for the department.

[0022] If the performance threshold Q 2 > The performance index Jxzs generates a third performance result, departmental early warning items, including departmental work efficiency Gxl, work quality Gzl, and attendance rate Cql, and simultaneously sends notifications to relevant departments and management departments.

[0023] Preferably, the churn prediction and analysis module includes a churn analysis unit and a churn assessment and prediction unit; the churn analysis unit makes predictions based on the performance evaluation results. When the performance evaluation results include the third performance result, it performs the prediction of personnel churn in the department. By analyzing the historical data of departing personnel in the department and the performance index Jxzs of the third performance result, and after processing all the data into dimensionless form, a churn prediction data set is constructed.

[0024] In this process, natural language processing technology was used to extract keywords from the historical data of departing employees of the department, and the frequency of keyword occurrence was obtained to form employee satisfaction Ymyd. The keywords included the number of times work environment appeared Znl, the number of times career development appeared Jxp, the number of times compensation and benefits appeared Jxf, and the number of times workload appeared Jst.

[0025] By extracting the churn prediction data set, the pre-churn index Ylzs is calculated. The specific calculation formula is as follows:

[0026]

[0027] In the formula, Mlcz represents the monthly difference in the number of employees, and Dsl represents the number of employees in the current month.

[0028] Preferably, the churn assessment and prediction unit uses a preset churn threshold W. 1 and churn threshold W 2 And the churn threshold W 1 > Churn threshold W 2 The results were compared and evaluated with the pre-churn index Ylzs. The specific evaluation content is as follows:

[0029] If the pre-churn index Ylzs > churn threshold W 1 At this point, a third turnover result is generated, indicating that the department has an abnormal turnover risk. In addition to investigating the specific reasons for the current employees' dissatisfaction, recruitment demand forecasting should also be carried out.

[0030] If the churn threshold W 1 ≥Pre-churn index Ylzs≥Churning threshold W 2 At this point, a second churn result is generated, indicating that the department's churn risk is abnormal, but has not exceeded the minimum churn threshold W. 2 ; Take staff retention measures, including conducting in-depth investigations to understand the specific reasons for employee dissatisfaction and making adjustments based on the reasons to retain staff;

[0031] If the churn threshold W 2 >Pre-churn index Ylzs: At this point, the first churn result is generated, indicating that the churn risk of this department is normal; maintain the existing management strategy, and continue to pay attention to employee satisfaction and performance, and maintain the current work environment and corporate culture.

[0032] Preferably, the recruitment demand analysis module includes a recruitment analysis unit and a demand assessment unit;

[0033] The recruitment analysis unit collects historical recruitment data, performs preprocessing and dimensionless processing, and finally constructs a recruitment data set. When the turnover prediction result includes a third turnover result, the pre-turnover index Ylzs in the third turnover result is simultaneously fitted with the relevant recruitment data in the recruitment data set, and the recruitment demand index Zpzs is calculated using the following formula:

[0034]

[0035] In the formula, Njz represents the average number of recruitments per year, Njb represents the average recruitment cost per year, Nls represents the average number of employees leaving per year, Nzp represents the average number of employees recruited per year, and Plsn represents the average number of years it takes for employees to leave.

[0036] Preferably, the demand assessment unit uses a preset recruitment threshold E. 1 and recruitment threshold E 2 And the recruitment threshold E 1 > Recruitment Threshold E 2 The recruitment demand index Zpzs was compared and evaluated, and the following evaluation results were generated:

[0037] If the recruitment demand index Zpzs > recruitment threshold E 1 At this point, the first recruitment result is generated, indicating that the current recruitment demand is abnormal and exceeds the preset recruitment threshold E. 1 Increase recruitment budget, and increase recruitment channels and frequency;

[0038] If the recruitment threshold E 1 ≥ Recruitment Demand Index Zpzs ≥ Recruitment Threshold E 2 At this point, a second recruitment result is generated, indicating that the current recruitment demand is abnormal, but not below the recruitment threshold E. 2 At this point, pay attention to recruitment effectiveness and staff turnover, and adjust recruitment strategies accordingly.

[0039] If the recruitment threshold E 2 > Recruitment demand index Zpzs. At this point, a third recruitment result is generated, indicating whether the current recruitment demand is normal or there is no recruitment demand. The existing recruitment strategy and pace will be maintained, with a focus on employee satisfaction and internal mobility.

[0040] Preferably, the learning model optimization module optimizes the existing learning model through preset optimization algorithms and strategies to improve the model's prediction accuracy and adaptability. At the same time, it updates the dataset regularly and finally combines historical data and real-time data to retrain and evaluate the model through deep learning algorithms.

[0041] Based on the evaluation results, the model is further optimized, including parameter fine-tuning and architecture optimization. Finally, the model's running status and prediction results are monitored in real time.

[0042] This invention provides an enterprise data intelligent analysis system based on deep learning. It has the following beneficial effects:

[0043] (1) The data acquisition and processing module solves the problem of inaccurate or inconsistent data caused by the diverse sources of human resources data, thereby improving the reliability of the analysis results. Specifically, the data acquisition and processing module integrates the employee performance system, training system, and attendance system through regional division and multiple API interfaces, realizing real-time data acquisition and transmission. At the same time, by establishing retry and data persistence mechanisms, data loss is prevented. Parallel processing strategies are adopted to process multiple data streams in parallel, and data is cleaned and preprocessed, improving data quality and consistency.

[0044] (2) Simultaneously, by predicting employee performance, turnover rate, and recruitment needs, the system helps companies formulate recruitment and human resource management strategies. The model building and deployment module selects a convolutional neural network for model training and deployment; the performance data analysis module analyzes performance data and calculates the performance index Jxzs; the turnover prediction analysis module combines historical turnover data and performance evaluation results to predict employee turnover; the recruitment demand analysis module predicts recruitment demand in advance and calculates the recruitment demand index Zpzs; and the learning model optimization module optimizes the model and updates the dataset to ensure the model's real-time performance and accuracy. The system evaluates each indicator using preset thresholds, generates corresponding evaluation results, and proposes improvement measures, effectively reducing the risk of employee turnover and improving recruitment efficiency and employee satisfaction. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the framework structure of an enterprise data intelligent analysis system based on deep learning according to the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1

[0048] Please see Figure 1A deep learning-based enterprise data intelligence analysis system includes a data acquisition and processing module, a model building and deployment module, a performance data analysis module, a churn prediction and analysis module, a recruitment demand analysis module, and a learning model optimization module.

[0049] The data acquisition and processing module is used to collect internal human resources-related data; it integrates employee performance system, training system and attendance system through API data interface, and collects and transmits data in real time using stream processing tools.

[0050] The model building and deployment module is used to build and deploy deep learning models; it selects convolutional neural networks, trains and evaluates them, and finally deploys them.

[0051] The performance data analysis module is used to collect and analyze performance data; to analyze the performance data using a deep learning model, obtain the performance index Jxzs and evaluate it; and to generate performance evaluation results.

[0052] The turnover prediction and analysis module is used to predict employee turnover based on performance evaluation results. By collecting historical turnover data and combining it with performance evaluation results, and then using a deep learning model for analysis, high-risk employees are identified, turnover prediction results are generated, and employee retention strategies are formulated.

[0053] The recruitment demand analysis module is used to predict recruitment demand in advance based on the turnover prediction report. It extracts historical recruitment data and historical employee turnover data, and performs cleaning and feature extraction to construct a recruitment demand data set. At the same time, it calculates and evaluates the recruitment demand index Zpzs, and conducts recruitment based on the evaluation results.

[0054] The learning model optimization module is used to iterate the learning model on a fixed-period dataset, including iterating on the dataset with fixed-period updates.

[0055] In this embodiment, the data acquisition and processing module is responsible for collecting relevant human resources data within the enterprise; the model building and deployment module is used to build and deploy deep learning models for practical applications; the performance data analysis module collects and analyzes employee performance data, uses deep learning models to calculate the performance index Jxzs, and helps the enterprise identify and improve employee performance; the turnover prediction and analysis module predicts employee turnover risk based on performance evaluation results, generates turnover prediction results, and formulates corresponding employee retention strategies; the recruitment demand analysis module predicts recruitment demand in advance based on the turnover prediction report and conducts recruitment planning based on the evaluation results; and the learning model optimization module is used to optimize the learning model to ensure its predictive accuracy and adaptability to meet the ever-changing needs of the enterprise.

[0056] Example 2

[0057] Please see Figure 1 The data acquisition and processing module divides the enterprise into regions based on the number of departments and labels them accordingly, specifically QY1, QY2, QY3, ..., QY n n represents the number of segmented regions; and performance data for each department's region is collected separately.

[0058] Simultaneously, multiple API interfaces are used to connect to the employee performance system, training system, and attendance system of each department / region. Statistical analysis is performed on the information collected from these systems to obtain sets of employee performance-related statistical information, training-related statistical information, and attendance record statistical analysis information. Secondly, retry and data persistence mechanisms are established during data transmission to prevent data loss. Finally, a parallel processing strategy is adopted to process multiple data streams in parallel. The data is then cleaned, preprocessed, and stored.

[0059] In this embodiment, accurate data collection from various departments is achieved through regional division and multiple API interfaces. Data loss is prevented through retry mechanisms and data persistence, and parallel processing is used to improve data processing efficiency. Ultimately, the high quality and consistency of the data are ensured, providing reliable support for subsequent analysis.

[0060] Example 3

[0061] Please see Figure 1 The model building and deployment module is used to select convolutional neural networks as the core algorithm based on enterprise human resources data, and divides the data collected by the data acquisition and processing module into training sets, validation sets and test sets according to proportions, which are used for model training, model tuning and final model evaluation respectively; the learning model is deployed and put into use through containerization technology to monitor and evaluate the enterprise's performance data, turnover data and recruitment demand data in real time.

[0062] In this embodiment, by selecting a convolutional neural network as the core algorithm and appropriately dividing the dataset for training, validation, and evaluation, the accuracy and reliability of the model are ensured. Containerization technology is used for model deployment, providing flexible update and expansion capabilities, enabling real-time monitoring and evaluation of enterprise performance, churn, and recruitment needs data, thereby improving the intelligence level of human resource management.

[0063] Example 4

[0064] Please see Figure 1 The performance data analysis module includes a performance analysis unit and a performance evaluation unit. The performance analysis unit performs dimensionless processing on all performance data from each department, constructs a performance data set, and calculates the performance index Jxzs using the following formula:

[0065]

[0066] In the formula, Gxl represents work efficiency, Rwl represents task completion rate, Pjt represents average delivery time difference, Mcl represents monthly output; Gzl represents work quality, Tsl represents complaint rate, Jsl represents promotion rate, Lrl represents profit margin; and Cql represents attendance rate.

[0067] Among them, the task completion rate Rwl, average delivery time difference Pjt, monthly output Mcl, complaint rate Tsl, promotion rate Jsl, and profit margin Lrl were obtained through statistical analysis of employee performance-related information.

[0068] The performance evaluation unit uses a preset performance threshold Q. 1 and performance threshold Q 2 The performance index Jxzs is evaluated; and the performance threshold Q is... 1 >Performance threshold Q 2 The specific assessment content is as follows:

[0069] If the performance index Jxzs > the performance threshold Q 1 The first performance result is generated, indicating that the department's performance is satisfactory, including satisfactory work efficiency Gxl, work quality Gzl, and attendance rate Cql.

[0070] If the performance threshold Q 1 ≥Performance Index Jxzs≥Performance Threshold Q 2 The second performance result is generated. If there are non-compliance items in the department, including the department's work efficiency Gxl, work quality Gzl and attendance rate Cql, the non-compliance items are marked and sent to the relevant departments for reminders. The item with the lowest value is analyzed and improvement measures are implemented for the department.

[0071] If the performance threshold Q 2 > The performance index Jxzs generates a third performance result, departmental early warning items, including departmental work efficiency Gxl, work quality Gzl, and attendance rate Cql, and simultaneously sends notifications to relevant departments and management departments.

[0072] In this embodiment, work efficiency Gxl represents the amount of work completed by an employee per unit of time, reflecting the employee's work speed and efficiency; task completion rate Rwl represents the proportion of assigned tasks completed by an employee, reflecting the employee's task execution ability and reliability; average delivery time difference Pjt represents the difference between the actual completion time and the scheduled completion time of a task, reflecting the employee's performance in completing tasks on time; monthly output Mcl represents the amount of work completed by an employee per month, reflecting the employee's output capacity and continuous work efficiency; work quality Gzl represents the quality level of tasks completed by an employee, reflecting the accuracy and precision of the employee's work; complaint rate Tsl represents the number of complaints related to the employee's work, reflecting the employee's job satisfaction and service level; promotion rate Jsl represents the frequency of employee promotions, reflecting the employee's career development and growth within the company; profit margin Lrl represents the profit margin generated by the employee's work, reflecting the employee's contribution to the company's profitability; and attendance rate Cql represents the employee's attendance, reflecting the employee's work attendance and discipline.

[0073] Example 5

[0074] Please see Figure 1 The churn prediction and analysis module includes a churn analysis unit and a churn assessment and prediction unit. The churn analysis unit makes predictions based on the performance evaluation results. When the performance evaluation results include the third performance result, it performs the prediction of personnel churn in the department. It constructs a churn prediction data set by analyzing the historical data of departing personnel in the department and the performance index Jxzs of the third performance result, and after processing all the data in a dimensionless manner.

[0075] In this process, natural language processing technology was used to extract keywords from the historical data of departing employees of the department, and the frequency of keyword occurrence was obtained to form employee satisfaction Ymyd. The keywords included the number of times work environment appeared Znl, the number of times career development appeared Jxp, the number of times compensation and benefits appeared Jxf, and the number of times workload appeared Jst.

[0076] By extracting the churn prediction data set, the pre-churn index Ylzs is calculated. The specific calculation formula is as follows:

[0077]

[0078]

[0079] In the formula, Mlcz represents the monthly difference in the number of employees, and Dsl represents the number of employees in the current month.

[0080] The churn assessment and prediction unit uses a preset churn threshold W. 1 and churn threshold W 2 And the churn threshold W 1 > Churn threshold W 2The results were compared and evaluated with the pre-churn index Ylzs. The specific evaluation content is as follows:

[0081] If the pre-churn index Ylzs > churn threshold W 1 At this point, a third turnover result is generated, indicating that the department has an abnormal turnover risk. In addition to investigating the specific reasons for the current employees' dissatisfaction, recruitment demand forecasting should also be carried out.

[0082] If the churn threshold W 1 ≥Pre-churn index Ylzs≥Churning threshold W 2 At this point, a second churn result is generated, indicating that the department's churn risk is abnormal, but has not exceeded the minimum churn threshold W. 2 ; Take staff retention measures, including conducting in-depth investigations to understand the specific reasons for employee dissatisfaction and making adjustments based on the reasons to retain staff;

[0083] If the churn threshold W 2 >Pre-churn index Ylzs: At this point, the first churn result is generated, indicating that the churn risk of this department is normal; maintain the existing management strategy, and continue to pay attention to employee satisfaction and performance, and maintain the current work environment and corporate culture.

[0084] In this embodiment, employee satisfaction Ymyd is constructed by collecting data on the frequency of occurrences of work environment (Znl), career development (Jxp), compensation and benefits (Jxf), and workload (Jst) to understand employees' satisfaction with various aspects of the company. The frequency of occurrences of work environment (Znl) reflects employees' attention to the work environment, including aspects such as office environment, facilities, and safety. High frequency of occurrence indicates that the work environment may be one of the main factors affecting employee satisfaction or dissatisfaction. The frequency of occurrences of career development (Jxp) reflects employees' attention to career development opportunities, including promotion opportunities, training, and development. High frequency of occurrence indicates that career development opportunities may be an important consideration for employees' decision to leave or stay. The frequency of occurrences of compensation and benefits (Jxf) reflects employees' attention to compensation and benefits, including wages, bonuses, and insurance. High frequency of occurrence indicates that compensation and benefits may be an important influencing factor on employee satisfaction or a major cause of dissatisfaction. The frequency of occurrences of workload (Jst) reflects employees' attention to workload, including workload, working hours, and overtime. High frequency of occurrence indicates that workload may be a major cause of employee stress and turnover.

[0085] Employee satisfaction Ymyd represents employees' satisfaction with their work environment, salary and benefits, career development, etc., reflecting their overall satisfaction with the department; the monthly employee number difference Mlcz represents the difference between the number of employees this month and last month, reflecting the changing trend of employee turnover. A larger difference indicates a higher turnover or recruitment situation; the current month's employee number Dsl represents the total number of employees in the current month, reflecting the size and status of the department's human resources.

[0086] Example 6

[0087] Please see Figure 1 The recruitment demand analysis module includes a recruitment analysis unit and a demand assessment unit;

[0088] The recruitment analysis unit collects historical recruitment data, performs preprocessing and dimensionless processing, and finally constructs a recruitment data set. When the turnover prediction result includes a third turnover result, the pre-turnover index Ylzs in the third turnover result is simultaneously fitted with the relevant recruitment data in the recruitment data set, and the recruitment demand index Zpzs is calculated using the following formula:

[0089]

[0090] In the formula, Njz represents the average number of recruitments per year, Njb represents the average recruitment cost per year, Nls represents the average number of employees leaving per year, Nzp represents the average number of employees recruited per year, and Plsn represents the average number of years it takes for employees to leave.

[0091] Among them, the average number of recruitments per year (Njz), average recruitment cost per year (Njb), average number of employees leaving per year (Nls), average number of employees recruited per year (Nzp), and average number of years of employee turnover (Plsn) were obtained through statistical records of recruitment data sets.

[0092] The demand assessment unit uses a preset recruitment threshold E. 1 and recruitment threshold E 2 And the recruitment threshold E 1 > Recruitment Threshold E 2 The recruitment demand index Zpzs was compared and evaluated, and the following evaluation results were generated:

[0093] If the recruitment demand index Zpzs > recruitment threshold E 1 At this point, the first recruitment result is generated, indicating that the current recruitment demand is abnormal and exceeds the preset recruitment threshold E. 1 Increase recruitment budget, and increase recruitment channels and frequency;

[0094] If the recruitment threshold E 1 ≥ Recruitment Demand Index Zpzs ≥ Recruitment Threshold E 2 At this point, a second recruitment result is generated, indicating that the current recruitment demand is abnormal, but not below the recruitment threshold E. 2 At this point, pay attention to recruitment effectiveness and staff turnover, and adjust recruitment strategies accordingly.

[0095] If the recruitment threshold E 2 > Recruitment demand index Zpzs, at which point a third recruitment result is generated, indicating whether the current recruitment demand is normal or there is no recruitment demand, and the existing recruitment strategy and pace are maintained, with a focus on employee satisfaction and internal mobility;

[0096] When a warning item appears, it indicates that there are two or more non-compliant items, and a detailed analysis of the reasons is conducted. At the same time, a prediction of staff turnover for that department is made.

[0097] In this embodiment, the average annual number of recruitments (Njz) represents the average number of recruitments conducted by the company each year over the past year, reflecting the company's recruitment frequency and needs; the average annual recruitment cost (Njb) represents the average annual cost spent by the company on recruitment, reflecting the company's recruitment budget and expenditure; the average annual turnover (Nls) represents the average number of employees lost by the company each year over the past year, reflecting the company's employee turnover and stability; the average annual recruitment (Nzp) represents the average number of employees recruited by the company each year over the past year, reflecting the company's efforts in filling vacancies and meeting growth needs; and the average turnover period (Plsn) represents the average time employees stay with the company. These metrics reflect the length of time employees remain with the company and indirectly reflect the company's employee retention capacity.

[0098] Example 7

[0099] Please see Figure 1 The learning model optimization module optimizes the existing learning model through preset optimization algorithms and strategies to improve the model's prediction accuracy and adaptability. At the same time, it updates the dataset regularly and finally combines historical data and real-time data to retrain and evaluate the model through deep learning algorithms.

[0100] Based on the evaluation results, the model is further optimized, including parameter fine-tuning and architecture optimization. Finally, the model's running status and prediction results are monitored in real time.

[0101] In this embodiment, the existing learning model is continuously optimized through preset optimization algorithms and strategies, which not only improves the model's prediction accuracy and adaptability but also ensures real-time data updates and processing. The model is retrained and evaluated by combining historical and real-time data, ensuring that it always reflects the latest enterprise situation. Simultaneously, model performance is further improved through parameter fine-tuning and architecture optimization, and real-time monitoring ensures the accuracy and reliability of the model's operating status and prediction results.

[0102] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An enterprise data intelligent analysis system based on deep learning, Features: It includes data collection and processing module, model building and deployment module, performance data analysis module, churn prediction analysis module, recruitment demand analysis module and learning model optimization module; The data collection and processing module is used to collect human resources related data within the enterprise; Integrate employee performance systems, training systems, and attendance systems through API data interfaces, and use stream processing tools to collect and transmit data in real time; The model building and deployment module is used to build a deep learning model and deploy it; by selecting a convolutional neural network and performing training and evaluation, and finally deploying it; The performance data analysis module is used to collect performance data and analyze it; analyze the performance data through a deep learning model, obtain a performance index Jxzs and evaluate it; generate a performance evaluation result; The turnover prediction analysis module is used to predict employee turnover based on performance evaluation results; By collecting historical resignation data and combining it with performance evaluation results, and then using deep learning models for analysis, we can identify high-risk employees, generate resignation prediction results, and develop employee retention strategies; The recruitment demand analysis module is used to predict recruitment demand in advance according to the resignation prediction report, extract historical recruitment data and historical staff loss data, and perform cleaning and feature extraction to construct a recruitment demand data set; at the same time, calculate and evaluate the recruitment demand index Zpzs, and recruit according to the evaluation results; The learning model optimization module is used to iterate the fixed-period data set of the learning model, including iterating the fixed-period update data set.

2. According to claim 1, a deep learning-based enterprise data intelligent analysis system, Features: The data collection and processing module divides the area according to the number of departments of the enterprise and marks them accordingly, specifically QY 1 , QY 2 , QY 3 、...、QY n , n represents the number of divided areas; and the performance data of each department area is collected separately; At the same time, through multiple API interfaces, the employee performance system, training system and attendance system of each department area are connected respectively; the information collected by the employee performance system, training system and attendance system are statistically analyzed simultaneously to obtain the employee performance related statistical information set, training related statistical information set and attendance record statistical analysis information set; Secondly, during the data transmission process, a retry mechanism and a data persistence mechanism are established to prevent data loss. Finally, a parallel processing strategy is adopted to process multiple data streams in parallel. Finally, the data is cleaned, preprocessed, and stored.

3. According to claim 2, a deep learning-based enterprise data intelligent analysis system, Features: The model building and deployment module is used to select a convolutional neural network as the core algorithm based on the enterprise human resources data, and divide the data collected by the data collection and processing module into a training set, a validation set, and a test set in proportion, which are used for model training, model tuning, and final model evaluation respectively; The learning model is deployed through containerization technology and put into use to monitor and evaluate the company's performance data, churn data, and recruitment demand data in real time.

4. According to claim 3, a deep learning-based enterprise data intelligent analysis system, Features: The performance data analysis module includes a performance analysis unit and a performance evaluation unit; the performance analysis unit constructs a performance data set after dimensionless processing of all performance data of each department, and calculates the performance index Jxzs through the following formula: In the formula, Gxl represents work efficiency, Rwl represents task completion rate, Pjt represents average delivery time difference, Mcl represents monthly output; Gzl represents work quality, Tsl represents complaint rate, Jsl represents promotion rate, Lrl represents profit margin; Cql represents attendance rate.

5. According to claim 4, a deep learning-based enterprise data intelligent analysis system, Features: The performance evaluation unit presets a performance threshold Q 1 and performance threshold Q 2 , evaluate the performance index Jxzs; and the performance threshold Q 1 >Performance threshold Q 2 The specific evaluation contents are as follows: If the performance index Jxzs> performance threshold Q 1 , generating the first performance result, indicating that the department's performance is qualified, including the department's work efficiency Gxl, work quality Gzl and attendance rate Cql are qualified; If the performance threshold Q 1 ≥Performance index Jxzs≥Performance threshold Q 2 , generate the second performance result, the department has unqualified items, including the department's work efficiency Gxl, work quality Gzl and attendance rate Cql, and mark the unqualified items, and send the unqualified items to the relevant departments for prompts, analyze the item with the lowest value and implement improvement measures for the department; If the performance threshold Q 2 >The performance index Jxzs generates the third performance result, the department warning item, including the department's work efficiency Gxl, work quality Gzl and attendance rate Cql, and simultaneously sends notifications to relevant departments and management departments.

6. According to claim 5, a deep learning-based enterprise data intelligent analysis system, Features: The churn prediction analysis module includes a churn analysis unit and a churn evaluation prediction unit; the churn analysis unit makes predictions based on the performance evaluation results, and when the performance evaluation results include the third performance result, performs predictions on the staff churn of the department, and constructs a churn prediction data set by analyzing the historical data of staff leaving the department and the performance index Jxzs of the third performance result, and performing dimensionless processing on all data; Among them, the natural language processing technology is used to extract keywords from the historical data of employees who left the department, obtain the frequency of keyword occurrence, and form employee satisfaction Ymyd, where the keywords include the number of occurrences of work environment Znl, the number of occurrences of career development Jxp, the number of occurrences of salary and benefits Jxf, and the number of occurrences of workload Jst; By extracting the churn prediction data set, the pre-churn index Ylzs is calculated. The specific calculation formula is as follows: In the formula, Mlcz represents the difference in the number of employees per month, and Dsl represents the number of employees in the current month.

7. The enterprise data intelligent analysis system based on deep learning according to claim 6, Features: The churn assessment prediction unit uses a preset churn threshold W 1 and the churn threshold W 2 , and the loss threshold W 1 > Loss threshold W 2 , and compared with the pre-churn index Ylzs, the specific evaluation content is as follows: If the pre-churn index Ylzs>churn threshold W 1 ,At this time, the third loss result is generated, indicating that the loss risk of the department is abnormal. In addition to investigating the specific reasons for the current employees’ dissatisfaction, recruitment demand forecasting is also required; If the loss threshold W 1 ≥Pre-churn index Ylzs≥churn threshold W 2 , the second churn result is generated, indicating that the churn risk of the department is abnormal, but does not exceed the minimum churn threshold W 2 ; Take measures to retain employees, including conducting in-depth investigations to understand the specific reasons for employees' dissatisfaction and making adjustments based on the reasons to retain employees; If the loss threshold W 2 > Pre-churn index Ylzs, at this time the first churn result is generated, indicating that the churn risk of the department is normal; Maintain existing management strategies and continue to focus on employee satisfaction and performance, and maintain the current work environment and corporate culture.

8. According to claim 7, a deep learning-based enterprise data intelligent analysis system, Features: The recruitment demand analysis module includes a recruitment analysis unit and a demand assessment unit; The recruitment analysis unit collects historical recruitment data, performs preprocessing and dimensionless processing, and finally constructs a recruitment data set; when the resignation prediction result includes the third loss result, the pre-loss index Ylzs in the third loss result is synchronously fitted with the relevant recruitment data in the recruitment data set, and the recruitment demand index Zpzs is calculated by the following formula: In the formula, Njz represents the average annual number of recruitments, Njb represents the average annual recruitment cost, Nls represents the average annual number of attrition, Nzp represents the average annual number of recruitments, and Plsn represents the average number of years of attrition.

9. The enterprise data intelligent analysis system based on deep learning according to claim 8, Features: The demand assessment unit presets the recruitment threshold E 1 and the recruitment threshold E 2 , and the recruitment threshold E 1 >Recruitment threshold E 2 , a comparative evaluation of the recruitment demand index Zpzs was conducted to generate the following evaluation results: If the recruitment demand index Zpzs> recruitment threshold E 1 , the first recruitment result is generated at this time, indicating that the current recruitment demand is abnormal and higher than the preset recruitment threshold E 1 ; Increase recruitment budget, while increasing recruitment channels and frequency; If the recruitment threshold E 1 ≥Recruitment demand index Zpzs≥Recruitment threshold E 2 , at this time, the second recruitment result is generated, indicating that the current recruitment demand is abnormal, but not lower than the recruitment threshold E 2 ; At this time, pay attention to recruitment results and staff turnover, and adjust recruitment strategies; If the recruitment threshold E 2 >Recruitment demand index Zpzs. At this time, the third recruitment result is generated, indicating that the current recruitment demand is normal or there is no recruitment demand. The existing recruitment strategy and rhythm are maintained, focusing on employee satisfaction and internal mobility.

10. The enterprise data intelligent analysis system based on deep learning according to claim 9, Features: The learning model optimization module optimizes the existing learning model through preset optimization algorithms and strategies to improve the prediction accuracy and adaptability of the model, and regularly updates the data set. Finally, it combines historical data and real-time data to retrain and evaluate the model through deep learning algorithms; Based on the evaluation results, the model is further tuned, including parameter fine-tuning and architecture optimization. Finally, the model's operating status and prediction results are monitored in real time.

Images