An analysis method and system for employee attendance data based on artificial intelligence

Abstract

This invention discloses an artificial intelligence-based method and system for analyzing employee attendance data. It includes a data processing module that preprocesses collected attendance data and then integrates it to construct an employee attendance database; a model building module that constructs a two-layer generative adversarial network (GAN) model based on the employee attendance database; an attendance detection module that performs pattern recognition on the employee attendance database using the GAN model and performs anomaly detection on the attendance data in the database using the GAN model; and a trend prediction module that predicts attendance trends based on simulated attendance data and real-time attendance data in the employee attendance database. This invention helps enhance enterprises' ability to monitor employee attendance behavior and provides forward-looking decision support for enterprise management, ensuring the smooth operation of enterprise management.

Description

Technical Field

[0001] This invention relates to the fields of mathematical modeling and data analysis technology, specifically to a method and system for analyzing employee attendance data based on artificial intelligence. Background Technology

[0002] As businesses become increasingly digitalized and automated, employee attendance management is becoming more complex and important. Traditional attendance management methods mainly rely on manual recording and simple electronic clocking-in systems. These methods are not only inefficient, but also fail to provide in-depth data analysis, resulting in a lack of data support for human resource decisions. Furthermore, due to the lack of effective data analysis tools, companies find it difficult to extract valuable information from large amounts of attendance data, such as employees' work patterns, efficiency issues, and the underlying causes of these issues.

[0003] In existing technologies, the analysis of attendance data is mostly limited to simple statistical summarization, which cannot conduct in-depth analysis of employee behavior patterns, let alone predict future attendance trends. Therefore, developing a system that can intelligently process and analyze abnormal attendance data to deal with complex employee behavior patterns, thereby improving the depth of attendance data analysis and predictive capabilities, and providing more scientific decision support and more efficient management strategies for enterprise management has become an urgent technical problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based method and system for analyzing employee attendance data. This invention can intelligently process and analyze abnormal attendance data, thereby improving the depth of attendance data analysis and predictive ability, and providing more scientific decision support and more efficient management strategies for enterprise management.

[0005] To achieve this objective, the present invention provides an artificial intelligence-based employee attendance data analysis system, comprising:

[0006] The data processing module preprocesses the collected attendance data and then uses big data processing technology to integrate the preprocessed attendance data to build an employee attendance database.

[0007] The model building module constructs a two-layer generative adversarial network model based on the employee attendance database. The first-layer generator in the two-layer generative adversarial network model is used to generate simulated attendance data based on historical attendance data in the employee attendance database, and to generate real attendance behavior patterns based on simulated attendance data; the second-layer generator is used to generate complex attendance behavior patterns.

[0008] The attendance detection module performs pattern recognition on the employee attendance database based on a two-layer generative adversarial network model, identifying whether the employee's attendance behavior pattern is a true attendance behavior pattern or a complex attendance behavior pattern; and performs anomaly detection on the attendance data corresponding to the attendance behavior patterns in the employee attendance database based on the two-layer generative adversarial network model.

[0009] The trend prediction module predicts attendance trends based on simulated attendance data and real-time attendance data in the employee attendance database.

[0010] The beneficial effects of this invention are:

[0011] This invention utilizes an innovative abnormal behavior detection and classification algorithm to automatically identify, extract features from, and classify potential abnormal behaviors in attendance data. By comparing attendance data generated using a two-layer generative adversarial network with actual data, abnormal behaviors are detected. The system can accurately analyze the frequency, duration, and environmental characteristics of abnormal behaviors and generate detailed abnormal behavior reports. Furthermore, it predicts future attendance trends based on attendance data, identifying potential attendance fluctuations and abnormal behaviors. This not only improves the depth and breadth of attendance data analysis and enhances enterprises' monitoring capabilities of employee attendance behavior, but also provides forward-looking decision support for enterprise management, helping to promptly identify and correct abnormal behavior patterns and ensure the management order of the enterprise. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the structure of the present invention;

[0013] Figure 2 This is the overall flowchart of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 a part of the embodiments of the present invention, not all of them. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0016] Example 1

[0017] An AI-based employee attendance data analysis system, such as Figure 1 As shown, it includes:

[0018] The data processing module preprocesses the collected attendance data and then uses big data processing technology to integrate the preprocessed attendance data to build an employee attendance database.

[0019] The model building module constructs a two-layer generative adversarial network model based on the employee attendance database. The first-layer generator in the two-layer generative adversarial network model is used to generate simulated attendance data based on historical attendance data in the employee attendance database, and to generate real attendance behavior patterns based on simulated attendance data; the second-layer generator is used to generate complex attendance behavior patterns.

[0020] The attendance detection module performs pattern recognition on the employee attendance database based on a two-layer generative adversarial network model, identifying whether the employee's attendance behavior pattern is a true attendance behavior pattern or a complex attendance behavior pattern; and performs anomaly detection on the attendance data corresponding to the attendance behavior patterns in the employee attendance database based on the two-layer generative adversarial network model.

[0021] The trend prediction module predicts attendance trends based on simulated attendance data and real-time attendance data in the employee attendance database.

[0022] In the above technical solution, the preprocessing includes standardization of the timestamps of the collected attendance data, conversion of the data format, and filling in missing data.

[0023] In the above technical solution, the collected attendance data are existing attendance records in the enterprise management software time collection system used by employees, including historical attendance data and real-time attendance data.

[0024] In the above technical solution, the specific methods for obtaining various types of data in the collected attendance data are as follows:

[0025] Monitor each employee's login time through the user interface of enterprise management software. login Logout time T logout And calculate the usage time T for each login. usage :

[0026] T usage =T logout -T login ;

[0027] Among them, T login T is the time of an employee's first login to the enterprise management software. logout T represents the employee's last logout time in the enterprise management software. usage This refers to the duration of the employee's usage during a single login session;

[0028] Record and store each employee's multiple login and logout records each day, and calculate and accumulate the total usage time T for the day. total :

[0029]

[0030] Where n1 is the number of times an employee logs in that day, and T usage(i) The usage duration for the employee's i-th login;

[0031] By monitoring employees' activity levels using the enterprise management software, including mouse movements and keyboard input records, login sessions with no activity exceeding a predetermined duration are marked as "idle," and this time is deducted from the total usage time. Based on the employees' active usage time of the enterprise management software, the actual effective working time T is further calculated. active :

[0032] T active =T total -T idle ;

[0033] Among them, T idle T represents the total idle time of employees when they are not performing any operations. active This refers to the actual effective working hours of employees.

[0034] In the above technical solution, the specific processing steps for integrating the preprocessed attendance data using big data processing technology to construct an employee attendance database are as follows:

[0035] The preprocessed attendance data undergoes a standardized data format process, converting data from different sources into the standard format required by the system, including timestamp format, geographic location information format, and employee identity information format.

[0036] The preprocessed attendance data is deduplicated using a hash algorithm to avoid storing duplicate data. The hash value H(d) is used to identify a unique attendance record, where d is the original attendance record and H(d) is the unique hash value generated by it.

[0037] Based on data cleaning and filtering technology, attendance data that does not conform to the standard format or has obvious errors is removed and corrected;

[0038] Store the cleaned attendance data and build an employee attendance database (DB). attendance ;

[0039] The employee attendance database is optimized using indexing techniques. Multi-dimensional indexes based on employee identity information, timestamps, and geographic location are created to improve database query efficiency. These indexes are represented as IDX. emp,time,locWhere emp is the employee's identity information, time is the timestamp, and loc is the geographical location.

[0040] In the above technical solution, the data processing module ensures that data from different devices and systems can be processed consistently by adopting unified timestamp standardization processing, data format conversion and missing data filling technology. It uses big data processing technology to build an employee attendance database, which solves the problem of data integration difficulties caused by inconsistent data sources and formats in traditional attendance systems. The technical improvement not only improves the integrity and consistency of data, but also lays a solid foundation for subsequent intelligent analysis and prediction.

[0041] In the above technical solution, the specific steps for building the model building module are as follows:

[0042] Two-layer Generative Adversarial Network (GAN) model dual The two-layer generative adversarial network (GAN) model consists of two parts: a generator (G1) and a discriminator (D1). The first-layer generator generates simulated attendance data based on historical attendance data from the employee attendance database. The second-layer discriminator (D2) determines the authenticity of the generated simulated attendance data. The second-layer generator generates complex attendance behavior patterns, including abnormal and normal behaviors. The third-layer discriminator (D2) determines the authenticity of the generated complex attendance behavior patterns. This two-layer GAN model utilizes historical attendance data to generate simulated data and combines it with real-time collected attendance data for analysis. The overall objective function of the two-layer GAN model is L. GAN Represented as:

[0043]

[0044] Where, p data (D real This represents the distribution of actual attendance data. and These represent the noise distributions used to generate simulated attendance data and complex attendance behavior patterns, respectively. Let L represent the mean, G1 and G2 be the generators of the first and second layers respectively, and D1 and D2 be the discriminators of the first and second layers respectively. The objective function is L. GAN A two-layer generative adversarial network model is trained by minimizing the generator loss and maximizing the losses of the first and second layer discriminators.

[0045] Based on employee attendance database DB attendance Using historical attendance data, train the first-layer generator G1 to generate simulated attendance data D. sim The process of generating simulated attendance data is represented as follows:

[0046] D sim =G1(z1)=σ(W)G1 z1+b G1 );

[0047] Where z1 is the input random noise vector, W G1 and b G1 These are the weight matrix and bias vector of the first-layer generator, respectively, where σ is the activation function and D is the bias vector. sim For the generated simulated attendance data, G1 is the first-level generator, and the generator adjusts the weights W. G1 and bias b G1 Study the distribution of historical attendance data;

[0048] The first-layer discriminator D1 is trained to discriminate the generated simulated attendance data D. sim Compared with actual attendance data D real The process for determining the authenticity of something is as follows:

[0049]

[0050] Among them, D input The input attendance data is the generated simulated attendance data D. sim Or actual attendance data D real W D1 and b D1 These are the weight matrix and bias vector of the first-layer discriminator, respectively. The first-layer discriminator optimizes the accuracy of discrimination by adjusting the parameters.

[0051] Based on the training results of the first-layer generator and the first-layer discriminator, a real attendance behavior pattern is generated, and a second-layer generator G2 is constructed to integrate all real attendance behavior patterns to generate a more complex attendance behavior pattern. The process of generating a complex attendance behavior pattern is represented as follows:

[0052] B complex =G2(z2)=tanh(W G2 z2+b G2 );

[0053] Where z2 is the input random noise vector, W G2 and b G2 Here, G1 represents the weight matrix and bias vector of the second-layer generator, tanh is the activation function, and G2 is the second-layer generator. The generator generates complex attendance behavior patterns B. complex To mimic abnormal patterns in real attendance behavior;

[0054] The second-layer discriminator D2 is trained to discriminate the generated complex attendance behavior pattern B. complex The process of determining the authenticity of actual attendance behavior patterns is expressed as follows:

[0055]

[0056] Among them, B input The input behavioral pattern data is used to generate complex attendance behavior patterns B. complex Or actual attendance behavior pattern B real The data is used by the second-layer discriminator to adjust the weight matrix W. D2 and bias vector b D2 Optimize discrimination capabilities;

[0057] By repeatedly training the first and second layer generative adversarial network models, optimizing the generation capabilities of the first and second layer generators G1 and G2, as well as the discrimination accuracy of the first and second layer discriminators D1 and D2, a two-layer generative adversarial network model (GAN) capable of simulating and recognizing attendance behavior is finally obtained. dual During training, the loss functions for the generator and discriminator are as follows:

[0058]

[0059] Among them, L G1,G2 Let G1 and G2 be the loss functions for the first and second layer generators, respectively. G1 represents the probability that the generated simulated attendance data and complex attendance behavior patterns are classified as real data. Minimizing this loss function improves the realism of the generated data. D1,D2 The loss function for the first and second layers, D1 and D2, represents the discriminator's ability to correctly identify real and generated data. By repeatedly optimizing the loss function, a two-layer generative adversarial network model that simulates and identifies complex attendance behavior patterns is finally constructed.

[0060] In the above technical solution, the model building module can be implemented based on deep learning frameworks (TensorFlow, PyTorch) to support the model training and inference process of generative adversarial networks.

[0061] In the above technical solution, the model building module constructs a two-layer generative adversarial network (GAN) model to achieve complex pattern recognition and anomaly detection of employee attendance behavior. Unlike traditional rule-based simple algorithms, the two-layer GAN model uses a first-layer generator to simulate real attendance data and a first-layer discriminator to judge its authenticity. The second-layer generator further generates complex attendance behavior patterns, including potential abnormal behaviors such as falsified attendance records and abnormally frequent login and logout behaviors. The second-layer discriminator is used to verify the authenticity of these complex behavior patterns. This model significantly improves the ability to recognize complex attendance behaviors, accurately detects and classifies normal and abnormal employee behaviors, and solves the shortcomings of traditional attendance systems in deep behavior recognition.

[0062] In the above technical solution, the method for recognizing specific employee attendance behavior patterns in the attendance detection module is as follows:

[0063] The integrated employee attendance database DB attendance The data is input into a two-layer generative adversarial network (GAN) model. dual In the process, attendance behavior pattern recognition is performed:

[0064]

[0065] Among them, M recog To identify complex attendance behavior patterns, G1 and G2 are the first and second layer generators of a two-layer generative adversarial network model, respectively, D2 is the second layer discriminator, z2 is the input random noise vector, and W... G1 and W G2 These are the weight matrices for the first and second layer generators, respectively. G1 and b G2 Let f be the bias vector, h be the activation function, and f be the nonlinear mapping function. This represents the selection of a complex behavioral pattern that maximizes the probability of the discriminator's output.

[0066] Based on the identified complex attendance behavior patterns, a classification function is used to categorize them into normal working hours M. normal Overtime work M overtime Using M without logging in absent and irregular usage behavior M irregular :

[0067]

[0068] Among them, M class For the set of complex attendance behavior patterns after classification, α i g represents the weight parameters in the classification model. i (M) is the classification feature function for complex attendance behavior patterns, and k is the number of features used for classification. This indicates the selection of a complex attendance behavior pattern with the largest weighted sum of features;

[0069] Feature extraction was performed on each complex attendance behavior pattern. The extracted features included login duration, operation frequency, and login interval.

[0070]

[0071] Among them, F extract T is the extracted feature set. usage For usage duration, F freq T is the operating frequency. interval β is the login interval. jh represents the weight parameters in the feature extraction model. j (F) is the feature function for feature extraction, and n2 is the number of features used in the feature extraction model;

[0072] The extracted complex attendance behavior features are input into a pattern classifier to refine the classification of complex attendance behavior patterns. The output of the classifier is represented as follows:

[0073]

[0074] Among them, M final For the final classification result, C is the classifier function, and γ is the classifier function. l q is the weight parameter in the pattern classifier. l (F extract ) is the feature mapping function in the classifier, m1 is the number of features used in the classifier. Based on the final classification result, a report on complex employee attendance behavior patterns is automatically generated. The report includes the employee's normal working hours, overtime work, non-login usage, and irregular usage behavior.

[0075] In the above technical solution, the specific method for detecting abnormal behavior in the attendance detection module is as follows:

[0076] Anomaly detection is performed on simulated attendance data generated by a two-layer generative adversarial network model and attendance data in the employee attendance database. The detection process identifies abnormal behaviors.

[0077] Based on the detected abnormal behavior, it is classified into forged usage records, abnormally frequent login and logout behavior, and abnormal behavior of not using the software for a long time.

[0078] For each abnormal behavior pattern, feature extraction is performed to extract the frequency of occurrence of the abnormal behavior, the duration of the abnormal behavior, and the environmental features when the abnormal behavior occurs;

[0079] Based on the extracted abnormal behavior features, they are input into the abnormal behavior discriminator to further judge and classify the abnormal behavior, and finally determine the type and characteristics of the abnormal behavior.

[0080] Based on the final determination of the abnormal behavior, an abnormal behavior report is generated. The report includes the identified abnormal behavior type, frequency, duration, and environmental information of the occurrence.

[0081] In the above technical solution, the attendance detection module, through a two-layer generative adversarial network model combined with a dynamic adjustment algorithm, can quickly generate and update employee attendance behavior patterns when employee attendance behavior data changes in real time, and promptly detect and report potential abnormal behaviors, ensuring that the attendance management system is always in an efficient and accurate operating state, and greatly improving the timeliness and effectiveness of management decisions.

[0082] In the above technical solution, the specific attendance trend prediction steps of the trend prediction module are as follows: based on historical attendance data D in the employee attendance database... hist and two-layer generative adversarial network model GAN dual The generated simulation data D sim Feature analysis was performed on employees' working hours and usage frequency to extract key features F. trend :

[0083]

[0084] Where, α i The weight φ for each feature i It is a nonlinear feature mapping function. and G1 and G2 represent the working duration and usage frequency of the i-th historical period, respectively, and the first and second layer generators of the two-layer generative adversarial network model, respectively. and Let n represent the historical data and simulated data for the i-th time period, respectively, and n3 be the total number of historical time periods.

[0085] The Time Series Analysis (TSM) model is used to perform trend analysis on the extracted feature data to predict the working hours (T) of employees over a future period. future and frequency of use F future The prediction process is represented as follows:

[0086]

[0087] Where, β j The weight ψ for each time series feature j This is a time series feature transformation function. For the j-th time series feature, Δt j Let θ be the time interval. j Here, TSM is the parameter set for the time series model, and TSM is the time series analysis model function that outputs the predicted future working hours T. future and frequency of use F future ;

[0088] Based on the prediction results of the time series analysis model, an attendance trend prediction model (PTM) is generated to predict employees' future attendance patterns, including possible attendance fluctuations and potential abnormal attendance behaviors.

[0089]

[0090] Where, γ k ξ represents the weighting coefficients of the attendance trend model. k This is a mapping function for attendance trend features. and ζ represents the predicted working duration and usage frequency for the k-th future time period, respectively. k The parameter set is used to generate the attendance trend model, and PTM is the final attendance trend prediction model.

[0091] In the above technical solution, the trend prediction module effectively predicts the future attendance trend of employees by reconstructing attendance data and applying time series analysis algorithms, and generates targeted management decisions based on this, such as arranging personnel scheduling or adjusting attendance policies in advance, which significantly improves the foresight and flexibility of enterprise management and reduces the operational risks caused by improper personnel arrangements.

[0092] In the above technical solution, the results of the attendance trend prediction model are visualized to generate a predictive analysis report. The report includes future working hours trends, changes in usage frequency, and possible attendance anomalies. An attendance data analysis report is generated through visualization tools. The attendance data analysis report displays the analysis results of the attendance data in chart form and provides real-time monitoring functions. The predictive analysis results are integrated with the enterprise management system and applied to employee performance appraisal, salary calculation, and work scheduling management functions.

[0093] In the above technical solution, the prediction not only relies on attendance data but also on the complex attendance behavior patterns generated by the system and the actual attendance behavior patterns. Combining attendance behavior patterns can enable the data to make more comprehensive trend predictions.

[0094] Example 2

[0095] An AI-based method for analyzing employee attendance data, such as Figure 2 As shown, attendance data is collected by detecting the duration of employees' use of enterprise management software; the collected attendance data is preprocessed; big data processing technology is used to integrate the preprocessed attendance data to construct an employee attendance database; a two-layer generative adversarial network (GAN) model is constructed; the GAN model is used to perform pattern recognition on the employee attendance database to identify employee attendance behavior patterns; anomaly detection is performed on the attendance data in the employee attendance database based on the GAN model to identify potential abnormal behaviors; attendance trend prediction is performed based on real-time attendance data and generated simulated data to predict employees' working hours and usage frequency in the future, and predictive analysis results are provided; the predictive analysis results are integrated with the enterprise management system, and the attendance data analysis results are applied to employee performance appraisal, salary calculation, and work scheduling management functions.

[0096] The specific methods for analyzing employee attendance data include the following steps:

[0097] The collected attendance data is preprocessed, and then big data processing technology is used to integrate the preprocessed attendance data to build an employee attendance database.

[0098] A two-layer generative adversarial network model is constructed based on the employee attendance database. The first-layer generator in the two-layer generative adversarial network model is used to generate simulated attendance data based on historical attendance data in the employee attendance database, and to generate real attendance behavior patterns based on simulated attendance data; the second-layer generator is used to generate complex attendance behavior patterns.

[0099] The two-layer generative adversarial network model is used to perform pattern recognition on the employee attendance database to identify whether the employee's attendance behavior pattern is a real attendance behavior pattern or a complex attendance behavior pattern; anomaly detection is then performed on the attendance data corresponding to the attendance behavior patterns in the employee attendance database based on the two-layer generative adversarial network model.

[0100] Attendance trends are predicted based on simulated attendance data and real-time attendance data from the employee attendance database.

[0101] Example 3

[0102] A computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 2.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

[0104] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0105] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0106] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0108] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. An artificial intelligence-based employee attendance data analysis system, characterized in that, include: The data processing module preprocesses the collected attendance data and then uses big data processing technology to integrate the preprocessed attendance data to build an employee attendance database. The model building module constructs a two-layer generative adversarial network model based on the employee attendance database. The first-layer generator in the two-layer generative adversarial network model is used to generate simulated attendance data based on historical attendance data in the employee attendance database, and to generate real attendance behavior patterns based on simulated attendance data; the second-layer generator is used to generate complex attendance behavior patterns. The attendance detection module performs pattern recognition on the employee attendance database based on a two-layer generative adversarial network model, identifying whether the employee's attendance behavior pattern is a true attendance behavior pattern or a complex attendance behavior pattern; and performs anomaly detection on the attendance data corresponding to the attendance behavior patterns in the employee attendance database based on the two-layer generative adversarial network model. The trend prediction module predicts attendance trends based on simulated attendance data and real-time attendance data in the employee attendance database.

2. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The preprocessing includes standardizing the timestamps of the collected attendance data, converting the data format, and filling in missing data.

3. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The collected attendance data consists of existing attendance records in the enterprise management software time collection system used by employees, including historical attendance data and real-time attendance data.

4. The employee attendance data analysis system based on artificial intelligence according to claim 3, characterized in that: The specific methods for obtaining various types of data in the collected attendance data are as follows: Monitor the login time (T) of each employee through the user interface of the enterprise management software. login Logout time T logout And calculate the usage time T for each login. usage : T usage =T logout -T login ; Among them, T login For each employee's login time, T logout T represents the logout time for each employee. usage This refers to the duration of the employee's usage during a single login session; Record and store each employee's multiple login and logout records each day, and calculate and accumulate the total usage time T for the day. total : Where n1 is the number of times an employee logs in that day, and T usage(i) The usage duration for the employee's i-th login; Based on the active time employees spend using enterprise management software, the actual effective working time T is further calculated. active : T active =T total -T idle ; Among them, T idle T represents the total idle time of employees when they are not performing any operations. active This refers to the actual effective working hours of employees.

5. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The specific steps for integrating the preprocessed attendance data using big data processing technology to construct an employee attendance database are as follows: Standardize the data format of the pre-processed attendance data; The preprocessed attendance data is deduplicated using a hash algorithm. Based on data cleaning and filtering technology, attendance data that does not conform to the standard format or has obvious errors is removed and corrected; Store the cleaned attendance data to build an employee attendance database.

6. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The specific steps for building the model building module are as follows: Two-layer Generative Adversarial Network (GAN) model dual The two-layer generative adversarial network (GAN) model consists of two parts: a generator (G1) and a discriminator (D1). The first-layer generator generates simulated attendance data based on historical attendance data from the employee attendance database. The second-layer discriminator (D2) determines the authenticity of the generated simulated attendance data. The third-layer generator generates complex attendance behavior patterns, including abnormal and normal behaviors. The fourth-layer discriminator (D2) determines the authenticity of the generated complex attendance behavior patterns. The overall objective function of the two-layer GAN model is L. GAN Represented as: Where, p data (D real This represents the distribution of actual attendance data. and These represent the noise distributions used to generate simulated attendance data and complex attendance behavior patterns, respectively. Let L represent the mean, G1 and G2 be the generators of the first and second layers respectively, and D1 and D2 be the discriminators of the first and second layers respectively. The objective function is L. GAN A two-layer generative adversarial network model is trained by minimizing the generator loss and maximizing the losses of the first and second layer discriminators. Based on employee attendance database DB attendance Using historical attendance data, train the first-layer generator G1 to generate simulated attendance data D. sim The process of generating simulated attendance data is represented as follows: D sim =G1(z1)=σ(W G1 z1+b G1 ); Where z1 is the input random noise vector, W G1 and b G1 These are the weight matrix and bias vector of the first-layer generator, respectively, where σ is the activation function and D is the bias vector. sim For the generated simulated attendance data, G1 is the first-level generator, and the generator adjusts the weights W. G1 and bias b G1 Study the distribution of historical attendance data; The first-layer discriminator D1 is trained to discriminate the generated simulated attendance data D. sim Compared with actual attendance data D real The process for determining the authenticity of something is as follows: Among them, D input The input attendance data is the generated simulated attendance data D. sim Or actual attendance data D real , W D1 and b D1 These are the weight matrix and bias vector of the first-layer discriminator, respectively. The first-layer discriminator optimizes the accuracy of discrimination by adjusting the parameters. Based on the training results of the first-layer generator and the first-layer discriminator, a real attendance behavior pattern is generated, and a second-layer generator G2 is constructed to integrate all real attendance behavior patterns to generate a more complex attendance behavior pattern. The process of generating a complex attendance behavior pattern is represented as follows: B complex =G2(z2)=tanh(W G2 z2+b G2 ); Where z2 is the input random noise vector, W G2 and b G2 Here, G1 represents the weight matrix and bias vector of the second-layer generator, tanh is the activation function, and G2 is the second-layer generator. The generator generates complex attendance behavior patterns B. complex To mimic abnormal patterns in real attendance behavior; The second-layer discriminator D2 is trained to discriminate the generated complex attendance behavior pattern B. complex The process of determining the authenticity of actual attendance behavior patterns is expressed as follows: Among them, B input The input behavioral pattern data is used to generate complex attendance behavior patterns B. complex Or actual attendance behavior pattern B real The data is used by the second-layer discriminator to adjust the weight matrix W. D2 and bias vector b D2 Optimize discrimination capabilities; By repeatedly training the first and second layer generative adversarial network models, optimizing the generation capabilities of the first and second layer generators G1 and G2, as well as the discrimination accuracy of the first and second layer discriminators D1 and D2, a two-layer generative adversarial network model (GAN) capable of simulating and recognizing attendance behavior is finally obtained. dual During training, the loss functions for the generator and discriminator are as follows: Among them, L G1,G2 Let G1 and G2 be the loss functions of the first and second layer generators, respectively, and let L represent the probability that the generated simulated attendance data and complex attendance behavior patterns are identified as real data. D1,D2 The loss function for the first and second layers, D1 and D2, is used to distinguish between them. By repeatedly optimizing the loss function, a two-layer generative adversarial network model that simulates and identifies complex attendance behavior patterns is finally constructed.

7. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The specific method for identifying employee attendance behavior patterns in the attendance detection module is as follows: The integrated employee attendance database DB attendance The data is input into a two-layer generative adversarial network (GAN) model. dual In the process, attendance behavior pattern recognition is performed: Among them, M recog To identify complex attendance behavior patterns, G1 and G2 are the first and second layer generators of a two-layer generative adversarial network model, respectively, D2 is the second layer discriminator, z2 is the input random noise vector, and W... G1 and W G2 These are the weight matrices for the first and second layer generators, respectively. G1 and b G2 Let f be the bias vector, h be the activation function, and f be the nonlinear mapping function. This represents the selection of a complex behavioral pattern that maximizes the probability of the discriminator's output. Based on the identified complex attendance behavior patterns, a classification function is used to categorize them into normal working hours M. normal Overtime work M overtime Using M without logging in absent and irregular usage behavior M irregular : Among them, M class For the set of complex attendance behavior patterns after classification, α i g represents the weight parameters in the classification model. i (M) is the classification feature function for complex attendance behavior patterns, and k is the number of features used for classification. This indicates the selection of a complex attendance behavior pattern with the largest weighted sum of features; Feature extraction was performed on each complex attendance behavior pattern. The extracted features included login duration, operation frequency, and login interval. Among them, F extract T is the extracted feature set. usage For usage duration, F freq T is the operating frequency. interval β is the login interval. j h represents the weight parameters in the feature extraction model. j (F) is the feature function for feature extraction, and n2 is the number of features used in the feature extraction model; The extracted complex attendance behavior features are input into a pattern classifier to refine the classification of complex attendance behavior patterns. The output of the classifier is represented as follows: Among them, M final For the final classification result, C is the classifier function, and γ is the classifier function. l q is the weight parameter in the pattern classifier. l (F extract ) is the feature mapping function in the classifier, and m1 is the number of features used in the classifier.

8. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The specific method for detecting abnormal behavior in the attendance detection module is as follows: Anomaly detection is performed on simulated attendance data generated by a two-layer generative adversarial network model and attendance data in the employee attendance database. The detection process identifies abnormal behaviors. Based on the detected abnormal behavior, it is classified into forged usage records, abnormally frequent login and logout behavior, and abnormal behavior of not using the software for a long time. For each abnormal behavior pattern, feature extraction is performed to extract the frequency of occurrence of the abnormal behavior, the duration of the abnormal behavior, and the environmental features when the abnormal behavior occurs; Based on the extracted abnormal behavior features, they are input into the abnormal behavior discriminator to further judge and classify the abnormal behavior, and finally determine the type and characteristics of the abnormal behavior. Based on the final determination of the abnormal behavior, an abnormal behavior report is generated. The report includes the identified abnormal behavior type, frequency, duration, and environmental information of the occurrence.

9. The employee attendance data analysis system based on artificial intelligence according to claim 1, characterized in that: The specific attendance trend prediction steps of the trend prediction module are as follows: Based on historical attendance data D in the employee attendance database... hist and two-layer generative adversarial network model GAN dual The generated simulation data D sim Feature analysis was performed on employees' working hours and usage frequency to extract key features F. trend : Where, α i The weight φ for each feature i It is a nonlinear feature mapping function. and G1 and G2 represent the working duration and usage frequency of the i-th historical period, respectively, and the first and second layer generators of the two-layer generative adversarial network model, respectively. and Let n represent the historical data and simulated data for the i-th time period, respectively, and n3 be the total number of historical time periods. The Time Series Analysis (TSM) model is used to perform trend analysis on the extracted feature data to predict the working hours (T) of employees over a future period. future and frequency of use F future The prediction process is represented as follows: Where, β j The weight ψ for each time series feature j This is a time series feature transformation function. For the j-th time series feature, Δt j Let θ be the time interval. j Here, TSM is the parameter set for the time series model, and TSM is the time series analysis model function that outputs the predicted future working hours T. future and frequency of use F future ; Based on the prediction results of the time series analysis model, an attendance trend prediction model (PTM) is generated to predict employees' future attendance patterns, including possible attendance fluctuations and potential abnormal attendance behaviors. Where, γ k ξ represents the weighting coefficients of the attendance trend model. k This is a mapping function for attendance trend features. and ζ represents the predicted working duration and usage frequency for the k-th future time period, respectively. k The parameter set is used to generate the attendance trend model, and PTM is the final attendance trend prediction model.

10. A method for analyzing employee attendance data based on artificial intelligence, characterized in that, It involves the following steps: The collected attendance data is preprocessed, and then big data processing technology is used to integrate the preprocessed attendance data to build an employee attendance database. A two-layer generative adversarial network model is constructed based on the employee attendance database. The first-layer generator in the two-layer generative adversarial network model is used to generate simulated attendance data based on historical attendance data in the employee attendance database, and to generate real attendance behavior patterns based on simulated attendance data; the second-layer generator is used to generate complex attendance behavior patterns. The two-layer generative adversarial network model is used to perform pattern recognition on the employee attendance database to identify whether the employee's attendance behavior pattern is a real attendance behavior pattern or a complex attendance behavior pattern; anomaly detection is then performed on the attendance data corresponding to the attendance behavior patterns in the employee attendance database based on the two-layer generative adversarial network model. Attendance trends are predicted based on simulated attendance data and real-time attendance data from the employee attendance database.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method of claim 10.

Images