An intelligent attendance management system

Abstract

This invention discloses an intelligent attendance management system, characterized by comprising a data interface layer, a data processing layer, and an interaction layer. The data interface layer flexibly defines and generates structured attendance cycle rules through a dynamic cycle adaptation module, and synchronizes and standardizes raw attendance data from multiple external data sources through a raw data acquisition module. The data processing layer uses a multi-source data fusion module to resolve data conflicts according to hierarchical adjudication rules, and uses a cross-cycle calculation module to segment cross-cycle events according to cycle rules, and an anomaly judgment module to perform compliance checks and generate result records. The interaction layer provides a visual operation interface and implements role-based fine-grained access control through an identity recognition login unit. The advantages are that it solves the problems of rigid cycles, multi-source data conflicts, cross-cycle event calculation errors, and numerous invalid reminders in traditional attendance systems, thereby improving the accuracy, authenticity, and efficiency of attendance management.

Description

Technical Field

[0001] This invention relates to the field of enterprise information management technology, and in particular to an intelligent attendance management system. Background Technology

[0002] Existing attendance systems often use fixed natural months or weeks as data processing cycles, which cannot adapt to custom cycles set by companies based on their own payroll calculation or management habits (such as from the 25th of last month to the 24th of this month). This rigid cycle setting can easily lead to calculation errors when processing cross-cycle holidays (such as sick leave spanning months), affecting the accuracy of attendance statistics.

[0003] Furthermore, modern employee attendance data comes from diverse sources, including clock-in systems (such as WeChat Work), leave application systems (OA), and business trip expense reimbursement systems (such as Huilianyi). Existing systems lack the ability to integrate data from various sources, which can easily lead to double counting or omissions. For example, if an employee has both a business trip application and a leave record on the same day, the system may deduct attendance days twice, or deduct only one while ignoring the other, resulting in inaccurate final attendance results.

[0004] In terms of system management, existing attendance systems typically have rather rudimentary access control, making it difficult to achieve fine-grained permission isolation between attendance operators, department administrators, and ordinary employees. Furthermore, abnormal attendance reminder mechanisms are mostly fixed-time instant pushes, failing to adequately consider practical situations such as data synchronization delays, easily generating a large number of invalid reminders and degrading user experience. In scenarios with complex organizational structures, existing attendance systems also suffer from data matching errors caused by duplicate department names. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an intelligent attendance management system that effectively integrates multi-source attendance data, realizes refined permission management and intelligent anomaly alerts, thereby improving the accuracy of attendance statistics and management efficiency.

[0006] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: an intelligent attendance management system, including a data interface layer, a data processing layer, and an interaction layer. The data interface layer includes a dynamic cycle adaptation module and a raw data acquisition module. The dynamic cycle adaptation module performs rule-based processing on the start and end dates of the attendance cycle set by the user, generates and stores structured cycle rules. The raw data acquisition module synchronously acquires attendance-related raw data from multiple application programming interfaces connected to external data sources, and processes the time information in the raw data to obtain a standardized raw dataset with a unified time format. The data processing layer includes a multi-source data fusion module, a cross-period calculation module, and an anomaly detection module. The multi-source data fusion module performs conflict detection and adjudication on the standardized original dataset according to preset hierarchical adjudication rules to obtain a fused dataset after conflict resolution. The cross-period calculation module performs time-slice processing on events spanning attendance periods in the fused dataset according to structured periodic rules to obtain periodic event sets belonging to each independent attendance period. The anomaly detection module performs compliance detection processing on each event in the periodic event set according to preset attendance compliance rules to generate a result record containing normal records and abnormal records. The interaction layer is provided with a user interface, which includes: The attendance exception viewing and processing unit provides an interface for centralized viewing of exception records, submission of replacement cards or appeals for exception records, approval operations, and viewing of virtual attendance records. The attendance data visualization and query unit is used to provide a graphical display of periodic event sets and result records, as well as a conditional query interface; The identity recognition login unit is used to provide login functionality for users and provide corresponding operation functions based on the login information entered by the user.

[0007] Compared with existing technologies, the advantages of this invention are as follows: By generating structured cycle rules through a dynamic cycle adaptation module, the system can flexibly adapt to enterprise-defined attendance cycles, fundamentally avoiding errors in cross-cycle event attribution caused by rigid cycles and improving the accuracy of attendance statistics; by synchronously acquiring and standardizing data from multiple source APIs through a raw data acquisition module, and resolving conflicts through the hierarchical adjudication rules of a multi-source data fusion module, the system effectively solves the problem of duplicate or missed calculations in multi-source data integration, ensuring the integrity and consistency of attendance data; by performing time-slice processing on cross-cycle events in the fused data through a cross-cycle calculation module, the system achieves accurate calculation of event attribution within each attendance cycle, further improving the reliability of cross-cycle attendance event processing; in addition, the anomaly judgment module works collaboratively with the attendance anomaly viewing and processing unit in the interaction layer, enabling timely identification of attendance anomalies and providing convenient processing procedures such as card replacement and appeals, thereby enhancing the efficiency of attendance management and user experience.

[0008] Furthermore, the identity recognition and login unit includes a login module, a role definition module, and a permission allocation module. Users create and store role metadata records containing role names and their corresponding sets of operation permissions corresponding to login information through the role definition module. The permission allocation module matches the login information entered by the user in the login unit with the role metadata records stored in the role definition module, and assigns the user operation permissions corresponding to the matched role metadata for the attendance exception viewing and processing unit and the attendance data visualization and query unit. By creating and storing role metadata records containing different role names and their sets of operation permissions through the role definition module, the system can define granular operation permissions based on the actual management structure of the enterprise (such as attendance operators, department administrators, and ordinary employees). The permission allocation module dynamically assigns operation permissions corresponding only to their roles to different users by matching login information with role metadata records in real time. This achieves precise control over access to functions such as attendance exception handling, data viewing and querying, etc., effectively solving the problems of coarse and difficult-to-isolate permission management in the prior art, and improving the security and flexibility of system management.

[0009] Furthermore, the preset hierarchical adjudication rules are applied in the following order of priority: first, mandatory rules are applied; second, event type priority rules are applied; and finally, time series rules are applied. The mandatory rules are specific rules with the highest validity pre-configured by the administrator. The event type priority rules stipulate that events with an attendance status of "business trip" have higher priority than events with an attendance status of "leave," and events with an attendance status of "leave" have higher priority than events with an attendance status of "overtime." The time series rules stipulate that when event types are the same, the event with the latest timestamp is adopted. By establishing a clear and progressive logical hierarchy, a precise and unambiguous conflict resolution path is provided for the multi-source data fusion module. This ensures that when integrating multi-source data such as WeChat attendance, OA leave requests, and Huilianyi business trip data, the system can intelligently resolve data conflicts based on preset priorities (e.g., business trips take precedence over leave requests), effectively avoiding the common problems of duplicate calculations or omissions in traditional methods. Practical applications have shown that this mechanism can significantly reduce the misjudgment rate of abnormal attendance caused by data conflicts (estimated by more than 60% based on customer data), and significantly improve the accuracy and automation level of attendance statistics.

[0010] Furthermore, the anomaly determination module is also used to: temporarily store anomaly records in a buffer list; Within the preset delay processing period, continuously monitor the updates of the periodic event set; The abnormal records in the buffer list are compared with the updated periodic event set. If an abnormal record can be covered by an event in the updated periodic event set, the abnormal record is removed from the buffer list. The attendance anomaly viewing and processing unit is also used to generate reminder messages for the remaining anomaly records in the buffer list and send them to the anomaly reminder module when the preset delay processing period ends. By setting a delay processing period (e.g., one week), the system can proactively adapt to the actual situation of delayed uploads of external data (such as business trip expense reports and leave approvals), continuously compare the temporarily stored anomaly records with the latest compliance event data, and automatically filter out "pseudo-anomalies" that can be covered by subsequent events (such as supplementary business trip applications). This fundamentally changes the traditional fixed instant reminder mode, ensuring that the final push reminders are all real and valid pending items. Practice has proven that this optimization can effectively reduce invalid reminders caused by data synchronization problems by more than 50%, greatly improving management efficiency and user experience.

[0011] Furthermore, the anomaly detection module also includes a weekday overtime module and a holiday overtime module; The aforementioned workday overtime module is used to obtain the workday overtime hours based on the employee's valid clock-out time, preset benchmark clock-out time, meal deduction time, minimum starting unit, and starting duration. The aforementioned holiday overtime module is used to obtain overtime hours for holidays based on employees' clock-in records on non-working days, preset lunch break times, and the smallest calculation unit. By setting different calculation parameters and rules for working days and holidays, the system can achieve fully automated and accurate calculation of overtime hours, completely replacing inefficient and error-prone methods such as manually checking Excel spreadsheets. This not only ensures that the calculation results comply with company policies and regulations but also frees attendance specialists from tedious manual calculations. Customer feedback indicates that this improvement is expected to increase overall attendance data processing efficiency by approximately 50% while ensuring calculation accuracy.

[0012] Furthermore, the multi-source data fusion module works in conjunction with the anomaly detection module to automatically complete missing card records. The specific processing procedure is as follows: When it is determined that an employee's clock-in record is missing for a certain work period, the system will first query the fused dataset to see if there is a high-priority attendance event that covers that period and is already in effect. If it exists, a virtual attendance record consistent with the status of the high-priority event will be automatically generated to cover the missing attendance records for that work period and suppress the generation of abnormal records that require additional attendance.

[0013] Furthermore, it also includes an AI analysis layer; the AI ​​analysis layer is communicatively connected to the data processing layer and is used to perform mining and correlation analysis on the periodic dataset to generate an attendance analysis report containing attendance trends and pattern recognition results; the AI ​​analysis layer integrates a locally deployed large language model for performing the mining and correlation analysis; the user interface also includes an AI analysis report and a resilience index unit to provide an interface for triggering and viewing the attendance analysis report; the periodic dataset includes the employee's actual attendance days, required attendance days, overtime hours, reasonable personal leave, sick leave, hospitalization duration, and specific lateness thresholds.

[0014] Furthermore, the AI ​​analysis layer incorporates a sustainable attendance resilience index calculation engine. Based on a periodic dataset, it calculates the sustainable attendance resilience index using a pre-defined algorithm model to assess employee attendance health status. The AI ​​analysis report and resilience index unit also display the sustainable attendance resilience index. The sustainable attendance resilience index calculation engine, through a pre-defined algorithm model, comprehensively calculates and quantifies multi-dimensional historical attendance behavior data of employees, generating a comprehensive index characterizing their attendance stability, fatigue risk, and health resilience. The sustainable attendance resilience index integrates scattered attendance indicators into an intuitive and comparable assessment value, providing human resources departments with a new, data-driven decision support tool for employee care, risk warning, and job placement, realizing a value extension from basic attendance management to employee health and sustainable development management.

[0015] Furthermore, the preset algorithm model includes at least one of the following: The first algorithm model is a general sustainable attendance resilience index calculation model, and its formula is as follows: ,in, REALWORK This indicates the actual number of days worked; the symbol "⋅" represents a multiplication operation. NEEDWORK Indicates the number of days required to be present. OVER60 This indicates the cumulative number of times someone is late by more than 60 minutes. α , β , c , d These represent adjustable parameters; The second algorithm model is a manufacturing-specific sustainable attendance resilience index calculation model, and its formula is as follows: ,in, OVERTIME Indicates total overtime hours. α 1 , c 1 , c 2 , k This refers to adjustable parameters set for manufacturing characteristics.

[0016] Furthermore, it also includes an early warning and decision support module, which is used to: periodically acquire the sustainable attendance resilience index and visualize the results, and trigger the early warning process according to preset early warning rules; the user interface also includes an early warning and decision support unit, which is used to provide an interface for configuring early warning rules, displaying early warning information, and viewing decision suggestions. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall framework of the present invention; Figure 2 This is a schematic diagram of business trip location provided by an embodiment of the present invention via WeChat. Figure 3 This is a schematic diagram of the missed attendance summary pop-up interface provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the main interface provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the check-in details interface provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the identity recognition and login unit in this invention.

[0018] Explanation of reference numerals in the attached figures: 101. Data Interface Layer; 102. Dynamic Period Adaptation Module; 103. Raw Data Acquisition Module; 201. Data Processing Layer; 202. Multi-Source Data Fusion Module; 203. Cross-Period Calculation Module; 204. Anomaly Detection Module; 301. Interaction Layer; 302. User Interface; 303. Attendance Anomaly Viewing and Processing Unit; 304. Attendance Data Visualization and Query Unit; 305. Identity Recognition and Login Unit; 306. Login Module; 307. Role Definition Module; 308. Permission Allocation Module. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0020] like Figure 1 As shown, an intelligent attendance management system includes a data interface layer 101, a data processing layer 201, and an interaction layer 301, characterized in that: The data interface layer 101 includes a dynamic cycle adaptation module 102 and a raw data acquisition module 103. The dynamic cycle adaptation module 102 performs rule processing on the start and end dates of the attendance cycle set by the user (e.g., set to "the 26th of last month to the 25th of this month"), generates and stores structured cycle rules, for example, represented in the form of JSON objects. These rules are stored in the system configuration library for use by various modules of the data processing layer 201. The raw data acquisition module 103 synchronously acquires attendance-related raw data from multiple application programming interfaces connected to external data sources, and performs time zone normalization and format standardization on the time information in the raw data to obtain a standardized raw dataset with a unified time format. Specifically, enterprise WeChat attendance data is passed in through API, leave requests from the OA system are passed in through middleware, and business trip applications from the Huilianyi system are passed in through SOAP. XML protocol is used for data transmission, and HR master data is also transmitted through middleware. All these attendance-related raw data are sent to a unified message gateway. The core function of the message gateway is data standardization. The message gateway first parses the various types of incoming data according to their corresponding native protocols, extracting key fields such as employee ID, event occurrence time, and event type (e.g., clock-in, leave, business trip). Then, the message gateway maps the extracted information to a unified internal event model to generate standardized events. Finally, metadata is added to each standardized event, including data source and receiving timestamp, and the standardized raw events are published to a specified topic in a message queue such as Kafka, completing the delivery of data from the interface layer to the processing layer. During this process, all time information is uniformly converted to ISO 8601 format strings with time zones, such as "2024-03-05T09:00:00+08:00". If the raw data does not carry a time zone, the default time zone is UTC+8, i.e., Beijing time.

[0021] The data processing layer 201 includes a multi-source data fusion module 202, a cross-cycle calculation module 203, and an anomaly detection module 204; The multi-source data fusion module 202, acting as a consumer of the message queue, retrieves the standardized raw dataset published by the data interface layer 101 from a designated Topic. Then, according to preset hierarchical adjudication rules, it performs conflict detection and adjudication on the standardized raw dataset to obtain a conflict-resolved fused dataset. This fused dataset is then persistently stored in the core fact table of the attendance database. The preset hierarchical adjudication rules, in order of execution priority, are as follows: first, mandatory rules are applied; second, event type priority rules are applied; and finally, time series rules are applied. Mandatory rules are specific rules pre-configured by the administrator with the highest validity, such as exemption from clocking in for senior executives and invalidation of rejected leave requests. Event type priority rules stipulate that events with an attendance status of "business trip" have higher priority than events with an attendance status of "leave," and events with an attendance status of "leave" have higher priority than events with an attendance status of "overtime." Time series rules stipulate that when event types are the same, the event with the latest timestamp is adopted.

[0022] For example, if employee A submits a full day of personal leave in the OA system and also has a business trip record in the Huilianyi system on the same day (June 10, 2024), after the system detects that the times completely overlap, according to the rule that business trip > leave, the business trip record will be adopted and the leave record will be ignored. The final attendance status will be displayed as business trip, and no personal leave days will be deducted.

[0023] The cross-period calculation module 203 reads the fused dataset from the attendance database and calls the structured periodic rules generated by the data interface layer 101 to perform time-slice processing on events spanning attendance periods in the fused dataset, obtaining periodic event sets belonging to each independent attendance period. The processed periodic event sets are written back to the periodic detail table of the attendance database and marked as ready, triggering the subsequent exception handling process.

[0024] The rules for time slice segmentation include the first period allocation rule and the day-based segmentation rule; Example 1: The attendance period is from the 1st to the 30th of each month. If an employee takes leave from June 28th to July 2nd, the first cycle allocation rule will be used. The entire leave event will be attributed to the June cycle. Although there are no clock-in or leave records on July 1st and 2nd, it will not be considered as absence because it has been covered by the June cycle.

[0025] Example 2: The attendance period is from the 1st to the 30th of each month. If an employee takes leave from June 28th to July 2nd, and the leave is divided by day, then June 28th-30th belongs to the June period, which is 3 days of leave, and July 1st-2nd belongs to the July period, which is 2 days of leave.

[0026] The periodic event set is stored in the attendance database. The storage structure of each event in the periodic event set includes the original data, the processing version number (such as v20240715.001, used to track history), and the conflict marker (such as none, no conflict, overridden, conflicting, causing others to be overwritten, manual_review requires manual review).

[0027] The anomaly detection module 204 monitors the readiness status of the periodic event set, or retrieves the latest periodic event set from the attendance database according to a fixed schedule (e.g., every morning). Based on preset attendance compliance rules, it performs compliance checks on each event in the periodic event set, generating a result record containing both normal and abnormal records. Both normal and abnormal records are written to the attendance result table. Simultaneously, the system proactively queries the approval status interfaces of external systems such as OA and Huilianyi via scheduled tasks (e.g., every 10 minutes) to obtain the latest "approved" and "rejected" statuses. Once a status change is detected, the system automatically triggers a re-evaluation of the relevant periodic events, updating the records in the attendance result table to ensure real-time synchronization with external approval results.

[0028] When the preset attendance compliance rule is a time threshold rule, the compliance detection processing performed by the anomaly judgment module 204 includes: comparing the employee's clock-in time recorded in the event with the preset working time. If the duration of lateness or early departure exceeds the corresponding time threshold, an anomaly record is generated; if the duration of lateness or early departure does not exceed the corresponding time threshold, a normal record is generated. For example, comparing the clock-in time with the set working time (e.g., 8:00), a clock-in at 8:02 is considered to be late within 10 minutes, and a clock-in at 8:30 is considered to be late within 60 minutes, both of which generate anomaly records. The logic for early departure judgment is similar. When the preset attendance compliance rule is a geofencing rule, the compliance detection processing performed by the anomaly detection module 204 includes: when an employee has no clock-in record during working hours, comparing the location information reported by the employee via mobile terminal with the preset office area and customer address geofencing; if the employee is within the preset office area and customer address geofencing, a receipt is generated; otherwise, an anomaly record is generated. Figure 2 You can see the location information reported through WeChat for Business.

[0029] In addition, when multiple high-priority events exist in the same time period and cannot be automatically resolved, an exception record marked as a data conflict is generated. For example, if an employee submits a business trip application for the same day in both Huilianyi and OA, an exception record marked as a data conflict will be generated. When no valid event (i.e., leave or business trip) covers the working period and there is a missing clock-in, an abnormal record marked as needing to be made up is generated. For example, if an employee does not request leave or go on a business trip, but has no clock-in record for the whole day, an abnormal record marked as needing to be made up is generated. When an event depends on the approval status, but is currently "pending approval" or "under approval", an exception record marked as an approval request is generated. For example, if an employee submits personal leave but it has not yet been approved, an exception record marked as an approval request is generated.

[0030] In this embodiment, the anomaly determination module 204 also includes a weekday overtime module and a holiday overtime module; The weekday overtime module is used to obtain the overtime hours for weekdays based on the employee's valid clock-out time, preset benchmark clock-out time, meal deduction time, minimum calculation unit, and calculation duration. The holiday overtime module is used to obtain the overtime hours for holidays based on employees' clock-in records on non-working days, preset lunch break times, and the minimum calculation unit.

[0031] (1) Calculation of Overtime Work on Weekdays: The system first identifies the employee's valid clock-out time, which is usually the latest clock-out within the afternoon clock-out period. The calculation rule is as follows: Daylight Saving Time rules, for example, if the base time for leaving get off work is 17:00: After deducting the base time from the valid clock-out time, first subtract the fixed meal time, such as 30 minutes, and the remaining time is rounded up in half-hour increments. Overtime is calculated from a full hour, and any fraction of an hour is not counted.

[0032] Winter time rules, for example, the base time for leaving get off work is 16:30: the calculation method is the same as for daylight saving time, only the base time is different.

[0033] Example: During winter time, employees clock out at 18:30. The calculation process is as follows: 18:30 - 16:30 = 2 hours; 2 hours - 0.5 hours (meal) = 1.5 hours. Since this meets the 1-hour overtime threshold, it is ultimately counted as 1.5 hours of overtime.

[0034] (2) Calculation of holiday overtime: The system identifies the actual work attendance intervals of employees on non-working days.

[0035] If there is only one consecutive check-in in the morning or afternoon, the duration of that check-in will be calculated directly as the end time minus the start time, rounded to the nearest half hour.

[0036] If there are clock-in records in both the morning and afternoon, then the total working hours = (afternoon closing time - morning starting time) - fixed lunch break time (e.g., 1 hour). The calculation result is also rounded to the nearest half hour.

[0037] The system will also mark overtime as "weekend overtime" or "statutory holiday overtime" based on weekends and statutory holidays in order to be compatible with different pay calculation rules.

[0038] In this embodiment, the multi-source data fusion module 202 and the anomaly detection module 204 work together to automatically complete missing card records. The specific processing procedure is as follows: When it is determined that an employee's clock-in record is missing during a certain work period, the anomaly detection module 204 will send a query to the multi-source data fusion module 202. The query will prioritize checking whether there are any high-priority attendance events that cover that period in the fused data, such as approved leave requests and valid business trip records. The criteria for determining high-priority attendance events include event type priority rules. If such a record exists, a virtual attendance record with the same status as the high-priority event will be automatically generated, overwriting any missing attendance records for that work period and suppressing the generation of abnormal records requiring make-up attendance. This virtual record will be written back to the periodic event set of the attendance database as a special event and reflected in subsequent result records.

[0039] Example: Employee A did not clock in on a certain workday afternoon, but the merged dataset contains a "personal leave" record approved by the OA system that covers that afternoon period. After detecting this situation, the system will not generate an abnormal "afternoon missed clock-in" record for the employee. Instead, it will automatically complete a virtual clock-in record with the status of "personal leave" in the employee's attendance details and mark it as "system automatically completed".

[0040] The interaction layer 301 is provided with a user interface 302, which includes: The attendance anomaly viewing and processing unit 303 provides an interface for core users by calling the dedicated data service API provided by the data processing layer 201. This interface allows users to centrally view anomaly records, initiate card replacement applications or attendance appeals for anomaly records, approve related applications, and view virtual attendance records automatically generated by the system based on high-priority attendance events. All user operations (such as submitting appeals) are transmitted back to the server via this API, driving business processes to update database records. Figure 3 The pop-up window showing the summary of missed clock-ins will display a list of abnormal records that need to be processed. For example, if a clock-in is missing, employees can submit an appeal or apply for a replacement clock-in here. The attendance data visualization and query unit 304 retrieves data from the attendance database through a high-performance query service interface (such as GraphQL or RESTful API), providing a graphical interface for core users to intuitively display the daily attendance status of individuals or departments in a calendar format. It generates statistical reports and dashboards for attendance and exception rates, and supports drill-down queries for detailed attendance data based on conditions. Figure 4 The main interface and Figure 5 The attendance details interface uses different colors (such as red to highlight abnormalities and green to mark normalities) to intuitively display attendance status and key information; The identity recognition login unit 305 is used to provide login functionality for users and provide corresponding operation functions based on the login information entered by the user. like Figure 6 As shown, the identity recognition login unit 305 includes a login module 306, a role definition module 307, and a permission allocation module 308. Users create and store role metadata records containing role names and their set of operation permissions corresponding to login information through role definition module 307. Permission allocation module 308 matches the login information entered by the user in login unit 306 with the role metadata records stored in role definition module 307, and assigns the user operation permissions corresponding to the matched role metadata to attendance exception viewing and processing unit 303 and attendance data visualization and query unit 304.

[0041] When a user initiates any data query request through the interaction layer 301, the identity recognition and login unit 305 (specifically through its permission allocation module 308) injects the current user's role and associated department tree permission information as a security context before the request reaches the backend service. When processing the query, the backend service automatically converts this permission context into SQL query conditions, such as AND department_tree = 'xxx industry / xxx department / xxx section', and dynamically appends it to the WHERE clause of the query statement. This achieves row-level data filtering at the database level, ensuring that users can only access data within their authorized scope. Furthermore, the permission allocation module 308 can also control the visibility of interface columns; for example, ordinary employees cannot view expense-related data columns.

[0042] In this embodiment, the anomaly determination module 204 is further configured to: temporarily store the anomaly record in a buffer list; Within the preset delay processing period, continuously monitor the updates of the periodic event set; The abnormal records in the buffer list are compared with the updated periodic event set. If an abnormal record can be covered by an event in the updated periodic event set, the abnormal record is removed from the buffer list. The attendance exception viewing and processing unit 303 is also used to generate a reminder message for the remaining exception records in the buffer list and send it to the exception reminder module when the preset delay processing period ends, through a message push service (such as integrating DingTalk / WeChat Work robot).

[0043] The intelligent attendance management system also includes an AI analysis layer. This layer communicates with the data processing layer 201 via an internal service bus to perform mining and correlation analysis on periodic datasets, generating attendance analysis reports containing attendance trends and pattern recognition results. The periodic dataset includes employees' actual attendance days, required attendance days, overtime hours, reasonable personal leave, sick leave, hospitalization duration, and specific lateness thresholds. The AI ​​analysis layer integrates locally deployed large language models, such as the Dify workflow engine and the Ollama DeepSeek model, for performing mining and correlation analysis. The system inputs the periodic dataset in structured JSON format, which includes predefined high-value analysis point markers (such as "relationship between attendance rate and leave type" and "attendance pattern analysis of different departments") and data quality markers (completeness, consistency, etc.). Based on these guidelines, the AI ​​analysis layer performs in-depth mining and correlation analysis, ultimately outputting an analysis report containing conclusions and recommendations. This report is stored in the system's document library and is available for front-end querying and display via the AI ​​analysis report and resilience index unit API. For example, it might indicate that a department's lateness pattern is related to the Monday morning meeting, assisting management decision-making.

[0044] The AI ​​analytics layer incorporates a sustainable attendance resilience index calculation engine. Based on a periodic dataset, it calculates the sustainable attendance resilience index according to a preset algorithm model to assess the employee's attendance health status. The sustainable attendance resilience index does not simply measure "attendance days," but rather comprehensively assesses an employee's ability to maintain stable attendance between high-intensity work and reasonable leave, paying particular attention to risk signals of non-sick leave turning into sick leave. The user interface 302 also includes AI analytics reports and resilience index units, providing an interface for management to trigger or view attendance trend and pattern analysis reports generated based on the periodic dataset, and to view the sustainable attendance resilience index used to assess the employee's attendance health status.

[0045] The preset algorithm model includes at least one of the following: The first algorithm model is a general sustainable attendance resilience index calculation model (mainly used by administrative and logistics positions and technical development positions), and its formula is: ,in, REALWORK This indicates the actual number of days worked; the symbol "⋅" represents a multiplication operation. NEEDWORK Indicates the number of days required to be present. OVER60 This indicates the cumulative number of times someone is late by more than 60 minutes. α , β , c , d These represent adjustable parameters, for example α =0.6, β =0.3, c =0.8, d =1.5 orα =0.6, β =0.3, c =0.8, d =1.5, all calculations are zero-protected. The closer the SARI value is to 1, the higher the employee attendance resilience; The second algorithm model is a manufacturing-specific sustainable attendance resilience index calculation model (mainly used by workshop employees), which addresses the high requirements of on-time arrival, fatigue prevention, and work injury early warning. Its formula is: ,in, OVERTIME Indicates total overtime hours. α 1 , c 1 , c 2 , k This refers to adjustable parameters set for manufacturing characteristics, such as... α 1 =0.5, c 1 =1.0, c 2 =1.2, k =2.0; or α 1 =0.5, c 1 =1.0, c 2 =1.2, k =2.0, and set the overtime compensation cap at 2 days to highlight the high risk of work-related injuries and hospitalization.

[0046] The intelligent attendance management system also includes an early warning and decision support module, used for: subscribing to sustainable attendance resilience index updates released by the AI ​​analysis layer; periodically obtaining the sustainable attendance resilience index and visualizing the results; triggering early warning processes according to preset early warning rules; pushing early warning information to relevant responsible persons through integrated messaging services; and using the sustainable attendance resilience index for decision support in scheduling optimization, flexible work system selection, or team health assessment. The system also supports securely exporting the final attendance results, analysis reports, and early warning information to the enterprise's existing payroll system or ERP platform via standard APIs.

[0047] The user interface 302 also includes an early warning and decision support unit, which provides an interface for managers to configure early warning rules based on the sustainable attendance resilience index, receive and process early warning information triggered by the system, and view decision suggestions generated based on attendance data and the resilience index.

[0048] The intelligent attendance management system is applied as follows: the system automatically calculates each employee's SARI or SARI-M monthly and incorporates the results into the human resources health dashboard. Administrators can set warning thresholds, such as automatically triggering a care process when SARI decreases by more than 15% for two consecutive months. SARI can also be used for management scenarios such as scheduling optimization, selection for flexible work system pilots, and team health assessment.

[0049] As shown in Table 1, through actual comparative testing, enterprises using the system of this invention have achieved significant improvements in key indicators compared to systems using traditional fixed-cycle and simple aggregation techniques. Specifically, the accuracy rate of cross-cycle holiday calculation has increased from 65% to 98%, the error rate of abnormal attendance has decreased from 28% to 11%, the overall attendance processing efficiency (i.e., the amount of work done per unit of man-hours) has increased by more than 150%, and the coverage rate of attendance health warnings has increased by 100%.

[0050] Table 1

Claims

1. An intelligent attendance management system, comprising a data interface layer, a data processing layer, and an interaction layer, characterized in that: The data interface layer includes a dynamic cycle adaptation module and a raw data acquisition module. The dynamic cycle adaptation module performs rule-based processing on the start and end dates of the attendance cycle set by the user, generates and stores structured cycle rules. The raw data acquisition module synchronously acquires attendance-related raw data from multiple application programming interfaces connected to external data sources, and processes the time information in the raw data to obtain a standardized raw dataset with a unified time format. The data processing layer includes a multi-source data fusion module, a cross-period calculation module, and an anomaly detection module. The multi-source data fusion module performs conflict detection and adjudication on the standardized original dataset according to preset hierarchical adjudication rules to obtain a fused dataset after conflict resolution. The cross-period calculation module performs time-slice segmentation on events spanning attendance periods in the fused dataset according to structured periodic rules to obtain periodic event sets belonging to each independent attendance period. The anomaly detection module performs compliance detection on each event in the periodic event dataset according to preset attendance compliance rules to generate a result record containing normal and abnormal records. The interaction layer is provided with a user interface, which includes: The attendance exception viewing and processing unit provides an interface for centralized viewing of exception records, submission of replacement cards or appeals for exception records, approval operations, and viewing of virtual attendance records. The attendance data visualization and query unit is used to provide a graphical display of periodic event sets and result records, as well as a conditional query interface; The identity recognition login unit is used to provide login functionality for users and provide corresponding operation functions based on the login information entered by the user.

2. The intelligent attendance management system according to claim 1, characterized in that, The identity recognition and login unit includes a login module, a role definition module, and a permission allocation module. Users create and store role metadata records containing role names and their set of operation permissions corresponding to login information through the role definition module. The permission allocation module matches the login information entered by the user in the login unit with the role metadata records stored in the role definition module, and assigns the user operation permissions corresponding to the matched role metadata to the attendance exception viewing and processing unit and the attendance data visualization and query unit.

3. The intelligent attendance management system according to claim 1, characterized in that, The preset hierarchical adjudication rules are applied in the following order of priority: first, mandatory rules are applied; second, event type priority rules are applied; and finally, time series rules are applied. The mandatory rules are specific rules with the highest validity pre-configured by the administrator. The event type priority rules stipulate that events with an attendance status of "business trip" have higher priority than events with an attendance status of "leave," and events with an attendance status of "leave" have higher priority than events with an attendance status of "overtime." The time series rules stipulate that when event types are the same, the event with the latest timestamp is adopted.

4. The intelligent attendance management system according to claim 1, characterized in that, The aforementioned anomaly detection module is also used to: temporarily store anomaly records in a buffer list; Within the preset delay processing period, continuously monitor the updates of the periodic event set; The abnormal records in the buffer list are compared with the updated periodic event set. If an abnormal record can be covered by an event in the updated periodic event set, the abnormal record is removed from the buffer list. The attendance exception viewing and processing unit is also used to generate a reminder message for the remaining exception records in the buffer list and send it to the exception reminder module when the preset delay processing period ends.

5. The intelligent attendance management system according to claim 1, characterized in that, The aforementioned anomaly detection module also includes a weekday overtime module and a holiday overtime module; The aforementioned workday overtime module is used to obtain the workday overtime hours based on the employee's valid clock-out time, preset benchmark clock-out time, meal deduction time, minimum starting unit, and starting duration. The holiday overtime module is used to obtain the overtime hours for holidays based on employees' clock-in records on non-working days, preset lunch break times, and the minimum calculation unit.

6. The intelligent attendance management system according to claim 1, characterized in that, The multi-source data fusion module works in conjunction with the anomaly detection module to automatically complete missing card records. The specific processing procedure is as follows: When it is determined that an employee's clock-in record is missing for a certain work period, the system will first query the fused dataset to see if there is a high-priority attendance event that covers that period and is already in effect. If it exists, a virtual attendance record consistent with the status of the high-priority event will be automatically generated to cover the missing attendance records for that work period and suppress the generation of abnormal records that require additional attendance.

7. The intelligent attendance management system according to claim 1, characterized in that, It also includes an AI analysis layer; the AI ​​analysis layer is communicatively connected to the data processing layer and is used to perform mining and correlation analysis on the periodic dataset to generate an attendance analysis report containing attendance trends and pattern recognition results; the AI ​​analysis layer integrates a locally deployed large language model for performing the mining and correlation analysis; the user interface also includes an AI analysis report and a resilience index unit to provide an interface for triggering and viewing the attendance analysis report; the periodic dataset includes the employee's actual attendance days, required attendance days, overtime hours, reasonable personal leave, sick leave, hospitalization duration, and specific lateness thresholds.

8. The intelligent attendance management system according to claim 7, characterized in that, The AI ​​analysis layer has a built-in sustainable attendance resilience index calculation engine, which calculates the sustainable attendance resilience index based on the periodic dataset and according to the preset algorithm model to evaluate the employee's attendance health status; the AI ​​analysis report and resilience index unit are also used to display the sustainable attendance resilience index.

9. The intelligent attendance management system according to claim 8, characterized in that, The preset algorithm model includes at least one of the following: The first algorithm model is a general sustainable attendance resilience index calculation model, and its formula is as follows: ,in, REALWORK This indicates the actual number of days worked; the symbol "⋅" represents multiplication. NEEDWORK Indicates the number of days required to be present. OVER60 This indicates the cumulative number of times someone is late by more than 60 minutes. α , β , γ , δ These represent adjustable parameters; The second algorithm model is a manufacturing-specific sustainable attendance resilience index calculation model, and its formula is as follows: ,in, OVERTIME Indicates total overtime hours. α 1 , γ 1 , γ 2 , κ This refers to adjustable parameters set for manufacturing characteristics.

10. The intelligent attendance management system according to claim 8, characterized in that, It also includes an early warning and decision support module, which is used to: periodically acquire the sustainable attendance resilience index and visualize the results, and trigger the early warning process according to preset early warning rules; the user interface also includes an early warning and decision support unit, which is used to provide an interface for configuring early warning rules, displaying early warning information, and viewing decision suggestions.

Images