Intelligent employment project OA management system based on cloud computing

Abstract

The invention discloses an intelligent employment project OA management system based on cloud computing. The intelligent employment project OA management system comprises the following steps: S1, data acquisition and preprocessing; s2, generating an optimal scheduling scheme based on an improved bacterial foraging optimization algorithm; s3, carrying out wavelet transformation and island model evolution learning, and generating risk early warning; s4, integrating scheduling and early warning information, and dynamically adjusting a scheduling scheme; s5, displaying a scheduling and early warning result, and receiving user feedback; and S6, storing the data, and implementing encryption and authority control. According to the invention, while intelligent scheduling and risk perception linkage management of intelligent employment tasks are realized, the stability and response speed of a scheduling scheme are significantly improved, and the accuracy of anomaly detection and the timeliness of early warning are effectively enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent labor systems, and particularly to an OA management system for intelligent labor projects based on cloud computing. Background Art

[0002] In the context of the increasing complexity of current enterprise human resource management and project execution, intelligent labor has become an important direction for achieving flexible and efficient scheduling. Most traditional labor management systems rely on static processes, rule configurations, and manual management methods, showing obvious limitations in dealing with project scenarios involving multiple tasks, multiple personnel, different regions, and multiple constraints. Especially in an environment with numerous tasks, limited resources, and high response time requirements, traditional scheduling systems often struggle to achieve precise matching and dynamic adjustment of human resources, thus affecting the overall project execution efficiency and quality.

[0003] As an important part of enterprise informatization, the office automation system plays a fundamental supporting role in daily office processes, approvals, task assignments, working hour statistics, etc. However, the existing OA systems generally suffer from the "information island" phenomenon, lacking a real-time linkage mechanism between task scheduling, business data, and the status of personnel resources. During the system operation, it is unable to perceive the dynamic changes in task execution and cannot make intelligent responses based on business risks. Especially when facing emergencies or potential business anomalies, traditional systems can only rely on manual identification and intervention, resulting in insufficient system intelligence and lagging responses, which are likely to cause waste of labor resources or task delays.

[0004] In the existing field of intelligent labor, some systems have also tried to introduce scheduling algorithms such as genetic algorithms and particle swarm optimization to achieve preliminary resource optimization scheduling. However, such methods often face problems such as slow algorithm convergence, easy entrapment in local optima, and poor robustness in dynamic change scenarios. In addition, most of these algorithms operate in an offline environment, making it difficult to adapt to the changes in real-time business flows and lacking the ability to adaptively handle system abnormal states. At the same time, for anomaly monitoring and risk identification, most existing systems rely on traditional methods such as rule matching and threshold warning, and are unable to deeply mine and distinguish the complex and variable characteristics of business data, especially weak in the identification of fine-grained abnormal behaviors during task execution.

[0005] More importantly, there is a lack of an effective linkage closed-loop mechanism between task scheduling and anomaly detection in existing systems. Even if some abnormal signals can be identified, it is difficult to transmit the abnormal feedback in real time to the task scheduling system for self-adjustment, resulting in the scheduling system still operating according to the original rules when facing risks, losing system flexibility and adaptability. Such a split design limits the application ability of intelligent labor systems in a highly uncertain and dynamically evolving environment.

[0006] In addition, during the process of users participating in system operations, there is often a lack of visual understanding of the task scheduling decision logic and anomaly recognition mechanism. The system feedback is relatively passive and single, which is not conducive to managers making comprehensive judgments and policy corrections based on system information. At the same time, a unified and reliable mechanism for data storage security and access management has not been formed, posing potential compliance risks and data abuse risks.

[0007] In summary, the existing OA management system for intelligent employment projects still has many technical bottlenecks and deficiencies in aspects such as the intelligence level of task scheduling algorithms, the accuracy and sensitivity of anomaly detection mechanisms, the closed-loop collaborative ability between scheduling and detection modules, as well as user participation in interaction and data security. Therefore, there is an urgent need for an OA management system for intelligent employment projects constructed in a cloud computing environment that can integrate dynamic task scheduling, real-time anomaly detection, intelligent linkage feedback, and security interaction mechanisms to achieve true intelligent perception, adaptive scheduling, and risk control. This system should have high scalability and real-time response capabilities, be able to effectively integrate business data streams and decision models, and significantly improve employment management efficiency, scheduling quality, and system stability.

[0008] Therefore, how to provide an OA management system for intelligent employment projects based on cloud computing is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0009] An object of the present invention is to propose an OA management system for intelligent employment projects based on cloud computing, which fully integrates an improved bacterial foraging optimization algorithm, wavelet transform, and an adaptive anomaly detection method of island model evolutionary learning, and details the overall technical solution for realizing task intelligent scheduling, risk real-time warning, and closed-loop dynamic response in employment projects. This system completes the optimal matching of tasks through intelligent modeling of task data and human resource data; at the same time, it performs multi-scale feature extraction and anomaly recognition on approval processes, working hour records, and task feedback data, and feeds back the anomaly information for scheduling adjustment to achieve self-optimization and linkage control at the system level. Compared with the existing technology, the present invention has the advantages of high intelligence level, strong scheduling optimization ability, accurate anomaly recognition, timely response, high system stability, etc., and can significantly improve the human resource allocation efficiency and management decision-making level in the intelligent employment scenario.

[0010] An OA management system for intelligent employment projects based on cloud computing according to an embodiment of the present invention includes:

[0011] A data collection and preprocessing module for collecting data and completing standardized preprocessing;

[0012] A task scheduling module for generating an optimal task scheduling plan based on an improved bacterial foraging optimization algorithm;

[0013] Anomaly detection and early warning module, which is used to detect anomalies and generate early warning information through wavelet transform and island model evolution learning algorithm;

[0014] Closed-loop linkage module, which is used to dynamically adjust the scheduling strategy according to the anomaly results and realize the linkage feedback between scheduling and detection;

[0015] User interaction module, which is used to display scheduling and early warning information and receive user feedback and instructions;

[0016] Data storage and security module, which is used to store various types of data and implement access control and encryption protection.

[0017] Optionally, the modules are implemented through the following methods:

[0018] S1. Collect labor task data, human resource data, approval processes, working hour records and task execution data through the data collection and preprocessing module to construct a data set;

[0019] S2. Process the data in the data set through the dynamic task scheduling module based on the improved bacterial foraging optimization algorithm, establish the task scheduling objective function and constraint conditions, and output the scheduling plan with the minimum scheduling objective function;

[0020] S3. Implement wavelet transform and adaptive anomaly detection algorithm based on island model evolution learning on the data set through the real-time anomaly detection and risk early warning module, extract multi-scale time-frequency features, and generate real-time risk early warning information;

[0021] S4. Integrate the optimal task scheduling plan and risk early warning information through the closed-loop linkage module, analyze the scheduling plan, and feedback it to the OA system business logic layer to achieve closed-loop management;

[0022] S5. Display the integrated information on the user interface through the user interaction and feedback module, and receive user input and feedback;

[0023] S6. Store the collected data, task scheduling plan, risk early warning information and user feedback data in the cloud database through the data storage and security management module, and implement access control and encryption measures on the data.

[0024] Optionally, the data set includes the following data subsets: The labor task data includes task ID, task type, task start time, task end time, task priority, and task completion status. The human resource data includes personnel ID, personnel capabilities, personnel working hours, current task assignment status of personnel, and personnel skill tags. The approval process data includes approval process ID, approval time, approval status, and approver ID. The working hours record data includes personnel ID, working hours record, working hours type, and working hours date. The task execution data includes task ID, task executor ID, task execution start time, task execution end time, and task completion percentage.

[0025] Optionally, S2 specifically includes:

[0026] S21. Establish a scheduling variable matrix representation based on the collected data subsets;

[0027] S22. Set the number of bacteria populations, the maximum number of iterations, the adaptive step size factor, the task constraint conditions, and the personnel ability boundaries, and initialize the current position solution of each bacteria individual. The current position solution corresponds to a feasible task scheduling plan that satisfies the constraint conditions. Among them, each bacteria individual represents a scheduling strategy, and this scheduling strategy forms the position vector of the bacteria individual through vectorizing and encoding the task scheduling variable matrix. Specifically, the scheduling variable matrix is a two-dimensional array, where the rows represent the human resource numbers and the columns represent the task numbers. The element values in the matrix are 0 or 1, indicating whether a certain task is assigned to a certain personnel. This matrix is sequentially expanded into a one-dimensional vector by columns or rows as the position representation of the bacteria individual, and participates in the chemotaxis, peristaltic mutation, and elimination and dispersion calculations in the scheduling optimization process;

[0028] S23. In the chemotaxis movement stage, calculate the bacteria population scheduling solution. The bacteria population scheduling solution is defined as the assignment strategy of bacteria individuals to each task:

[0029]

[0030] Among them, X (k) represents the assignment strategy of bacteria individuals to each task in the k-th iteration. Based on the task priority factor γ i and gradient feedback, the movement formula for updating the bacteria position is:

[0031]

[0032] Among them, X (k+1) is the assignment strategy of bacteria individuals to each task in the (k + 1)-th iteration. C is the adaptive step size factor, δ is the perturbation adjustment coefficient, rand(q) is a random number function generated in the interval [-1, 1], represents the gradient direction of the fitness function;

[0033] S24. Discretize the updated allocation strategy X (k+1) into a scheduling matrix form, denoted as X. Introduce a local mutation operation during the creeping stage to achieve local search optimization through the exclusive - or perturbation matrix M. The specific operation is as follows:

[0034]

[0035] Among them, represents the exclusive - or operator, and M is the perturbation mask matrix;

[0036] S25. In the elimination of dispersion stage, calculate the fluctuation entropy value η of the current bacterial population scheduling solution. When η < θ, perform bacterial re - initialization. θ is the set entropy threshold, and the calculation formula of η is:

[0037]

[0038] Among them, p k represents the probability distribution of the bacterial individual falling into the k - th local area, and log represents the logarithmic function;

[0039] S26. When the number of iterations reaches the maximum value or the optimal solution has not been updated for several consecutive generations, output the scheduling plan that minimizes the task - scheduling objective function, and transmit it to the business logic layer of the OA system for execution. Among them, in each round of bacterial foraging iteration process, the system calculates the corresponding task - scheduling objective function value based on the scheduling variable matrix, and uses this objective function value as the individual fitness index to evaluate the quality of the scheduling plan. The smaller its value, the better the comprehensive scheduling effect of the plan in terms of task - completion efficiency, resource - allocation balance, and personnel - load rationality.

[0040] Optionally, the specific steps of S3 are as follows:

[0041] S31. Based on the scheduling plan currently output by the task - scheduling module, extract the behavior data related to scheduling tasks, execution personnel, and approval nodes, and construct a unified multi - dimensional time - series data set. The data set includes the approval - process duration sequence, working - hour record sequence, and task - execution feedback sequence;

[0042] S32. Select orthogonal wavelet basis functions and perform J - layer discrete wavelet decomposition on the multi - dimensional time - series data set to obtain a set of scale components;

[0043] S33. Construct feature vectors based on each scale component, and splice all scale feature vectors to form the final input feature set F;

[0044] S34. Allocate the feature set F to M heterogeneous sub - populations {Ω1, Ω2, …, Ω MIn the isolated island evolution model composed of}, each sub - population adopts different individual coding strategies and local evolution mechanisms to independently learn the abnormal boundary, specifically including:

[0045] S341. Initialize the heterogeneous sub - population Ω m The set of individuals where represents the boundary model parameters of the k - th individual in the feature space;

[0046] S342. The density - driven adaptive mutation mechanism is applied to the isolated island evolution learning process of the anomaly detection module to dynamically update the individual boundary model parameters in each sub - population. The mechanism combines the behavioral feature distribution changes of the approval process duration sequence, working hour record sequence, and task execution feedback sequence data in the intelligent labor scenario, and implements targeted updates to the individual parameters. The density - driven adaptive mutation mechanism is used to update the individual set parameters, and the update formula is:

[0047]

[0048] where randn() is a standard normal distribution random perturbation, α is the perturbation amplitude coefficient, β is the density - sensitive adjustment factor, Δρ is the current sample density change, and ρ base is the initial reference density;

[0049] S343. Define the fitness function Φ m (Z) to evaluate the individual performance and calculate the fitness of the updated individual fitness;

[0050] S344. In the individual selection of each generation, retain the abnormal score individuals higher than the preset value in the previous generation according to the calculated fitness Perform mutation operations in its neighborhood and retain the historical key abnormal boundaries to achieve long - term memory retention of abnormal features;

[0051] S35. Perform cross - population migration operations every fixed number of generations τ, and inject some abnormal score individuals higher than the preset value into other populations for boundary fusion to obtain the integrated abnormal score function:

[0052]

[0053] where A m (t) is the abnormal score of the population Ω m for the input data at time t;

[0054] S36. Set the abnormal decision threshold θ A , and generate an abnormal alarm signal:

[0055]

[0056] Return R(t) to the closed-loop linkage module. After meeting the requirements, feedback it to adjust the scheduling plan and respond to system risks.

[0057] Optionally, the step S4 specifically includes:

[0058] S41. Conduct correlation analysis on the abnormal alarm signal R(t) output by the anomaly detection module and the current minimum scheduling plan, and identify the abnormal impact factors according to the abnormal type and related tasks, including:

[0059] Identification of high-risk task units: If the data time period corresponding to the abnormal signal completely coincides with the task execution time and the abnormal score A(t) ≥ θ A , then mark the task unit to which the task belongs as a high-risk task unit;

[0060] Identification of unstable resource nodes: If risks or anomalies continuously occur in the tasks assigned to a certain human resource, then mark this resource as an unstable node;

[0061] Extraction of sensitive task patterns: If multiple schedules of a certain task trigger abnormal scores in different time periods, then mark this task as a sensitive type;

[0062] S42. Adjust the scheduling constraints and scheduling objective function according to the identification results, and trigger the adaptive adjustment mechanism of task scheduling:

[0063] When a high-risk task unit appears, increase its weight coefficient in the objective function and limit the resource allocation range;

[0064] When an unstable resource node is detected, impose resource load limit constraints and adjust the upper limit of tasks that can be assigned to personnel;

[0065] When a sensitive task pattern is identified, trigger a reconstruction strategy, generate a new schedule, and update the task assignment rules;

[0066] S43. Re-execute the improved bacterial foraging optimization algorithm on the updated task scheduling objective function and constraint conditions to generate an optimized scheduling plan X ′ ;

[0067] S44. Evaluate the difference between the optimized scheduling plan and the current scheduling plan. The difference evaluation indicators include:

[0068] Task assignment difference rate

[0069] Personnel reassignment ratio

[0070] Resource stability change coefficient Δ S = σ new - σ orig ;

[0071] Among them, Task orig represents the task allocation set in the original scheduling scheme, and Task new represents the task allocation set of the optimized scheduling scheme, and Task orig △Task new The △ in it is the symmetric difference operator, representing the new set composed of the different elements in the two sets. represents the indicator function. If the task set of person h j changes, it is 1, where m is the total number of human resources. is the original task set of personnel. is the new task set of personnel, and σ orig is the standard deviation of the number of tasks per person under the original scheduling scheme, and σ new is the standard deviation under the new scheduling scheme;

[0072] S45. If the difference degree is within the acceptable range, then replace the original scheduling scheme X ′ with the optimized scheduling scheme X, write it into the scheduling result database, otherwise retain the original scheme and record the exception feedback for subsequent policy correction;

[0073] S46. Synchronously update the scheduling decision log, write the log into the task scheduling module, and feedback it to the OA system business logic layer.

[0074] The beneficial effects of the present invention are as follows:

[0075] (1) By introducing an improved bacterial foraging optimization algorithm, the present invention realizes the optimal matching of labor tasks and human resources on the basis of considering multi-dimensional constraints such as task priority, personnel load, and resource capacity. Compared with the traditional static scheduling method, the present invention can dynamically adjust the scheduling scheme in multi-task and multi-change scenarios, has stronger self-adaptability and convergence efficiency, and significantly improves the scientificity and execution efficiency of human resource allocation.

[0076] (2) By combining wavelet transform and island model evolutionary learning algorithm, the present invention extracts multi-scale time-frequency features and intelligently identifies time series data such as approval processes, working hour records, and task feedback, and can effectively capture potential abnormal behaviors and abnormal task patterns. Compared with the traditional early warning methods based on thresholds or rules, the present invention has higher accuracy and robustness in identifying weak anomalies, continuous attack behaviors, and complex behavior patterns, which helps enterprises detect risks early and formulate intervention measures.

[0077] (3) The closed-loop linkage module proposed by the present invention can feedback the abnormal detection results to the scheduling module in real time, dynamically adjust the scheduling objective function and constraint conditions according to the abnormal type, task impact range and risk level, regenerate the scheduling plan, and conduct difference evaluation and automatic replacement. This mechanism breaks the isolation between the scheduling and detection modules, enabling the system to have the capabilities of continuous learning, self-repair and optimized scheduling, thereby improving the intelligent level and operation stability of the overall labor employment system. Description of the Drawings

[0078] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0079] Figure 1 is a method flowchart of an OA management system for a cloud computing-based intelligent labor employment project proposed by the present invention;

[0080] Figure 2 is a system flowchart of an OA management system for a cloud computing-based intelligent labor employment project proposed by the present invention. Detailed Embodiments

[0081] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic way, so they only show the components related to the present invention.

[0082] Refer to Figure 1 and Figure 2 , an OA management system for a cloud computing-based intelligent labor employment project includes the following steps:

[0083] A data collection and preprocessing module, used to collect data and complete standardized preprocessing;

[0084] A task scheduling module, used to generate an optimal task scheduling plan based on an improved bacterial foraging optimization algorithm;

[0085] An abnormal detection and warning module, used to detect abnormalities and generate warning information through wavelet transform and island model evolution learning algorithm;

[0086] A closed-loop linkage module, used to dynamically adjust the scheduling strategy according to the abnormal results and achieve the linkage feedback between scheduling and detection;

[0087] A user interaction module, used to display scheduling and warning information and receive user feedback and instructions;

[0088] A data storage and security module, used to store various types of data and implement access control and encryption protection.

[0089] The present invention constructs a smart labor OA management system with six major functional modules including data collection, task scheduling, anomaly warning, closed-loop linkage, user interaction, and data security under the cloud computing architecture, realizing the full-process intelligent management of human resource scheduling and project execution. The modules are highly integrated and operate collaboratively on the cloud platform. Compared with the problems of fragmented modules and single functions in traditional OA systems, this structure significantly improves the scalability of the system and the ability to manage cross-regional tasks, supporting flexible deployment and efficient operation in scenarios with large scale, multiple projects, and frequent dynamic changes in the workforce. In this embodiment, the modules are implemented through the following methods:

[0090] S1. Collect employment task data, human resource data, approval processes, working hour records, and task execution data through the data collection and preprocessing module, and construct a data set;

[0091] S2. Process the data in the data set through the dynamic task scheduling module based on an improved bacterial foraging optimization algorithm, establish a task scheduling objective function and constraint conditions, and output a scheduling plan with the minimum scheduling objective function;

[0092] S3. Implement wavelet transform and an adaptive anomaly detection algorithm based on the evolution learning of the island model on the data set through the real-time anomaly detection and risk warning module, extract multi-scale time-frequency features, and generate real-time risk warning information;

[0093] S4. Integrate the optimal task scheduling plan and risk warning information through the closed-loop linkage module, analyze the scheduling plan, and feedback it to the business logic layer of the OA system to achieve closed-loop management;

[0094] S5. Display the integrated information on the user interface through the user interaction and feedback module, and receive user input and feedback;

[0095] S6. Store the collected data, task scheduling plan, risk warning information, and user feedback data in the cloud database through the data storage and security management module, and implement access control and encryption measures for the data.

[0096] The S1 - S6 processes of the present invention construct an efficient linkage business closed-loop between multiple cloud nodes from task data collection to result feedback and encrypted storage. Especially in smart labor projects, manpower scheduling needs to be frequently adjusted according to task progress and risks. Through this method process, a dynamic adaptive control path of real-time perception - optimized scheduling - risk warning - scheduling reconstruction - user feedback - result writing can be achieved, effectively avoiding the core drawbacks of traditional OA systems such as static scheduling being difficult to cope with sudden changes and lacking system self-optimization capabilities.

[0097] In this embodiment, the specific content of S2 includes:

[0098] S21. Establish a scheduling variable matrix representation based on the collected data subset;

[0099] S22. Set the number of bacteria populations, the maximum number of iterations, the adaptive step factor, the task constraint conditions, and the personnel ability boundaries, and initialize the current position solutions of each bacteria individual. The current position solution corresponds to a feasible task scheduling scheme that satisfies the constraint conditions. Among them, each bacteria individual represents a scheduling strategy, and this scheduling strategy forms the position vector of the bacteria individual through vectorized encoding of the task scheduling variable matrix. Specifically, the scheduling variable matrix is a two-dimensional array, where the rows represent the human resource numbers, the columns represent the task numbers, and the element values in the matrix are 0 or 1, indicating whether a certain task is assigned to a certain person. This matrix is sequentially expanded into a one-dimensional vector by column or by row as the position representation of the bacteria individual, and participates in the chemotaxis, peristaltic mutation, and elimination dispersion calculations in the scheduling optimization process;

[0100] S23. In the chemotaxis movement stage, calculate the bacteria population scheduling solution. The bacteria population scheduling solution is defined as the allocation strategy of bacteria individuals for each task:

[0101]

[0102] Among them, X (k) represents the allocation strategy of bacteria individuals for each task in the k-th iteration. Based on the task priority factor γ i and gradient feedback, the movement formula for updating the bacteria position is:

[0103]

[0104] Among them, X (k+1) is the allocation strategy of bacteria individuals for each task in the (k + 1)-th iteration, C is the adaptive step factor, δ is the perturbation adjustment coefficient, rand(q) is a random number function generated in the interval [-1, 1], represents the gradient direction of the fitness function;

[0105] S24. Discretize the updated allocation strategy X (k+1) into a scheduling matrix form, denoted as X. In the peristalsis stage, introduce a local mutation operation to achieve local search optimization through the exclusive OR perturbation matrix M. The specific operation is:

[0106]

[0107] Among them, represents the exclusive OR operator, and M is the perturbation mask matrix;

[0108] S25. In the elimination dispersion stage, calculate the fluctuation entropy value η of the current bacteria population scheduling solution. When η < θ, perform bacteria re-initialization. θ is the set entropy threshold, and the calculation formula of η is:

[0109]

[0110] where p k represents the probability distribution of a bacterial individual falling into the k-th local area, and log represents the logarithmic function;

[0111] S26. When the number of iterations reaches the maximum value or the optimal solution does not update for several consecutive generations, output the scheduling plan that minimizes the task scheduling objective function and transmit it to the business logic layer of the OA system for execution. Among them, in each round of bacterial foraging iteration process, the system calculates the corresponding task scheduling objective function value based on the scheduling variable matrix, and uses this objective function value as the individual fitness index to evaluate the quality of the scheduling plan. The smaller the value, the better the comprehensive scheduling effect of the plan in terms of task completion efficiency, resource allocation balance, and personnel load rationality.

[0112] The present invention first introduces an improved bacterial foraging optimization algorithm into the human resource task allocation in the intelligent labor scenario, and constructs a dynamic scheduling model considering multiple constraints such as task priority, personnel ability boundary, and resource load. The algorithm improves the solution space exploration ability through chemotaxis, adaptive perturbation, and local mutation strategies, adapts to the actual enterprise project management requirements of high task density, large human resource fluctuations, and frequent scheduling conflicts, effectively improves the scheduling convergence speed and allocation rationality, and reduces the task reallocation rate and the frequency of manual intervention.

[0113] In this embodiment, the specific steps of S3 are as follows:

[0114] S31. Based on the scheduling plan currently output by the task scheduling module, extract the behavior data related to the scheduling tasks, execution personnel, and approval nodes, and construct a unified multi-dimensional time series data set, which includes the approval process duration sequence, working hour record sequence, and task execution feedback sequence;

[0115] S32. Select an orthogonal wavelet basis function and perform J-layer discrete wavelet decomposition on the multi-dimensional time series data set to obtain a set of scale components;

[0116] S33. Construct feature vectors based on each scale component, and splice all scale feature vectors to form the final input feature set F;

[0117] S34. Assign the feature set F to the island evolution model composed of M heterogeneous sub-populations {Ω1, Ω2, …, Ω M}. Each sub-population independently performs abnormal boundary learning using different individual coding strategies and local evolution mechanisms, specifically including:

[0118] S341. Initialize the individual set of the heterogeneous sub-population Ω m of where Denote the boundary model parameters of the k-th individual in the feature space;

[0119] S342. The density-driven adaptive mutation mechanism is applied to the island evolution learning process of the anomaly detection module to dynamically update the individual boundary model parameters of each sub-population. The mechanism combines the behavioral feature distribution changes of the approval process duration sequence, working hour record sequence, and task execution feedback sequence data in the intelligent labor scenario, and implements targeted updates to the individual parameters. The density-driven adaptive mutation mechanism is used to update the individual set parameters, and the update formula is:

[0120]

[0121] where randn() is a standard normal distribution random perturbation, α is the perturbation amplitude coefficient, β is the density sensitivity adjustment factor, Δρ is the current sample density change amount, and ρ base is the initial reference density;

[0122] S343. Define the fitness function Φ m (Z) to evaluate the individual performance and calculate the updated individual fitness;

[0123] S344. In the selection of individuals in each generation, retain the individuals with anomaly scores higher than the preset value in the previous generation according to the calculated fitness Perform mutation operations in their neighborhoods, retain the historical key anomaly boundaries, and achieve long-term memory retention of anomaly features;

[0124] S35. Perform cross-population migration operations every fixed number of generations τ, and inject some individuals with anomaly scores higher than the preset value into other populations for boundary fusion to obtain an integrated anomaly scoring function:

[0125]

[0126] where A m (t) is the anomaly score of the population Ω m for the input data at time t;

[0127] S36. Set the anomaly determination threshold θ A , and generate an anomaly warning signal:

[0128]

[0129] Return R(t) to the closed-loop linkage module, and after meeting the requirements, feedback it to adjust the scheduling plan and respond to system risks.

[0130] The present invention combines the multi-scale feature extraction ability of wavelet transform with the heterogeneous evolution characteristics of the island model to construct a high-precision anomaly detection module for intelligent employment process behavior data. By enabling multiple sub-populations to learn the weak signal boundaries separately and fusing the anomaly scoring results, the system can accurately identify problems such as task delays, personnel overload, and approval bottlenecks in complex, non-linear, and sudden business data, significantly improving the employment risk identification rate and early warning timeliness, and meeting the urgent need for refined risk control in dynamic project management.

[0131] In this embodiment, S4 specifically includes:

[0132] S41. Conduct correlation analysis on the anomaly warning signal R(t) output by the anomaly detection module and the current minimum scheduling plan, and identify the anomaly impact factors according to the anomaly type and associated tasks, including:

[0133] High-risk task unit identification: If the data time period corresponding to the anomaly signal completely coincides with the task execution time and the anomaly score A(t)≥θ A , then mark the task unit to which the task belongs as a high-risk task unit;

[0134] Unstable resource node identification: If risks or anomalies continuously occur in the tasks assigned to a certain human resource, mark this resource as an unstable node;

[0135] Sensitive task pattern extraction: If anomalies are triggered in multiple schedules of a certain task in different time periods, mark this task as a sensitive type;

[0136] S42. Adjust the scheduling constraints and scheduling objective function according to the recognition results, and trigger the task scheduling adaptive adjustment mechanism:

[0137] When a high-risk task unit appears, increase its weight coefficient in the objective function and limit the resource allocation range;

[0138] When an unstable resource node is detected, impose resource load limit constraints and adjust the upper limit of tasks that personnel can be assigned;

[0139] When a sensitive task pattern is recognized, trigger a reconstruction strategy, generate a new schedule, and update the task assignment rules;

[0140] S43. Re-execute the improved bacterial foraging optimization algorithm on the updated task scheduling objective function and constraint conditions to generate an optimized scheduling plan X ′ ;

[0141] S44. Evaluate the difference degree between the optimized scheduling plan and the current scheduling plan. The difference evaluation indicators include:

[0142] Task assignment difference rate

[0143] Personnel reassignment ratio

[0144] Resource stability change coefficient Δ S = σ new - σ orig ;

[0145] Among them, Task orig represents the task assignment set in the original scheduling scheme, and Task new represents the task assignment set of the optimized scheduling scheme. The △ in Task orig △Task new is the symmetric difference operator, representing a new set composed of different elements in the two sets. represents the indicator function. If the task set of personnel h j changes, it is 1. m is the total number of human resources. is the original personnel task set. is the new personnel task set, and σ orig is the standard deviation of the number of tasks of each personnel under the original scheduling scheme, and σ new is the standard deviation under the new scheduling scheme;

[0146] S45. If the difference degree is within the acceptable range, then replace the original scheduling scheme X ′ with the optimized scheduling scheme X, write it into the scheduling result database, otherwise retain the original scheme and record the exception feedback, and perform policy correction later;

[0147] S46. Synchronously update the scheduling decision log, write the log into the task scheduling module, and feedback it to the OA system business logic layer.

[0148] In the present invention, by real-time feedback of the risk signal output by the anomaly detection module to the task scheduling module and adjusting the scheduling objective function and constraint conditions according to the anomaly type, the system can quickly reconstruct the scheduling scheme and evaluate its difference degree, ensuring that the updated policy is both reasonable and stable. This closed-loop linkage mechanism is particularly applicable to the environment in the intelligent labor project where tasks are closely dependent, sudden anomalies are frequent, and the scheduling real-time requirement is high, avoiding problems such as task delay and manpower waste caused by the risk signal being ignored or the response lag, and enhancing the self-adaptability and anti-interference ability of the system.

[0149] Example:

[0150] To verify the feasibility of this invention, it was applied to the project scheduling and management system at the headquarters of a large internet technology company. Located in Beijing's Zhongguancun Software Park, the company employs over 5,000 people, including approximately 1,300 project-based personnel. They must manage over 80 employment projects simultaneously on a daily basis, characterized by fragmented task granularity, frequent cross-departmental collaboration, complex approval processes, and frequent personnel changes. Previously, the company used a traditional OA system for employment project management, but issues such as task conflicts, redundant scheduling, and delayed response to exceptions were prominent during peak periods, severely impacting project efficiency and human resource utilization.

[0151] During peak project periods, such as from September to November 2024, the company launched 27 major platform redesign projects, totaling 1,450 tasks and involving 876 employees. The original system relied on rule-based templates for scheduling, with frequent manual intervention. This resulted in a high task reassignment rate of 22.6%, an average task approval cycle of 3.5 days, an average resource response delay of over 26 hours, and an 18.3% temporary employee replacement rate. More seriously, the company experienced a task execution delay risk event in mid-October 2024. The system failed to promptly identify anomalies in work time records, causing multiple tasks to fail in tandem and ultimately delaying customer delivery by over a week.

[0152] In order to solve the above problems, the company introduced the cloud computing-based smart employment project OA management system proposed in the present invention, and deployed it on its private cloud platform, opening up key data channels such as task dispatching, manpower scheduling, approval working hours, task feedback, etc., and integrating the "dynamic task scheduling module", "abnormal detection and risk warning module" and "closed-loop linkage module" in the present invention.

[0153] After system deployment, the data acquisition and preprocessing module collects project task descriptions, human resource histories, current work hours, and approval progress in real time. The system dynamically schedules tasks based on an improved bacterial foraging optimization algorithm, taking into account task priority, labor load, and capability tags, and outputs the optimal scheduling solution. During task execution, the system uses wavelet transforms and island evolution anomaly recognition models to analyze work hours records and feedback data, promptly identifying weak anomalies such as abnormal approval delays and execution deviations. This anomaly information is then transmitted to the closed-loop linkage module in real time, and the scheduling solution is automatically reconstructed through difference analysis and objective function adjustment mechanisms.

[0154] The following are key comparison data before and after system deployment, as shown in Table 1:

[0155] Table 1: Comparison of the scheduling effects between the traditional system and the system of the present invention

[0156]

[0157]

[0158] According to the statistical results of Table 1, "Comparison Data Table of Scheduling Effects between Traditional System and the System of the Present Invention", it can be clearly seen that the application effect of the present invention in the intelligent labor employment scenario is significantly better than that of the traditional OA system. Especially in key indicators such as task scheduling efficiency, human resource utilization rate, and exception handling ability, significant optimizations have been achieved.

[0159] First of all, from the perspective of approval efficiency, the average task approval cycle of the traditional system is 3.5 days, while the system of the present invention has compressed the approval cycle to 1.6 days, a reduction of 54.3%, significantly accelerating the task turnover speed and improving the overall operation efficiency. The task reallocation rate has also decreased from the original 22.6% to 5.1%, a decrease of 77.4%. This means that the scheduling plan is more stable and accurate, reducing the management costs and communication burdens brought by repeated adjustments.

[0160] In terms of human resources, the human replacement rate under the traditional system is as high as 18.3%, while that of the system of the present invention is only 6.7%, a decrease of 63.4%. This fully shows that the system of the present invention has a more refined control over the matching relationship between personnel capabilities and tasks, reducing the temporary replacement frequency during the labor employment process. And the average utilization rate of human resources has also increased from 74.2% to 92.4%, an increase of 24.5%, indicating that human resources have been more fully and reasonably scheduled and used, reducing the phenomena of idleness and repeated investment.

[0161] In terms of project execution risk control, the incidence of project delays in the traditional system is 11.5%, while that of the system of the present invention is controlled at 1.2%, a reduction of nearly 90%. This shows that the scheduling strategy of the system of the present invention is more forward-looking and stable, effectively avoiding potential execution obstacles. In addition, the accuracy rate of exception warnings has been greatly increased from 34.2% to 91.8%. The system can accurately identify potential risks and respond quickly, shortening the average time limit for exception response from 26.1 hours to 3.4 hours, a reduction of 86.9%, greatly improving the emergency handling efficiency and enhancing the enterprise's ability to handle sudden problems.

[0162] Finally, in terms of user feedback, the system satisfaction has increased from the original 62% to 91%, reflecting that the system has also been widely recognized in terms of interface friendliness, scheduling transparency, and intelligent recommendation.

[0163] The OA management system for the intelligent labor employment project proposed by the present invention not only has obvious technical advantages in scheduling algorithms and exception detection mechanisms, but also demonstrates excellent implementation effects and business adaptation capabilities in practical applications. It is a comprehensive improvement of the existing labor employment management system and has high promotion value and market prospects.

[0164] This embodiment fully demonstrates that the intelligent labor-using OA system proposed by the present invention can effectively solve key problems such as slow scheduling, untimely abnormal identification, and poor system linkage of traditional systems, improve scheduling efficiency, risk control ability, and human resource utilization rate, and has good engineering adaptability and commercial promotion value.

[0165] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered within the protection scope of the present invention.

Claims

1. An OA management system for a smart employment project based on cloud computing, characterized in that, Including: A data acquisition and preprocessing module, which is used to collect data and complete standardized preprocessing; A task scheduling module, which is used to generate an optimal task scheduling scheme based on an improved bacterial foraging optimization algorithm; An anomaly detection and early warning module, which is used to detect anomalies and generate early warning information through wavelet transform and island model evolution learning algorithm; A closed-loop linkage module, which is used to dynamically adjust the scheduling strategy according to the anomaly results and realize the linkage feedback between scheduling and detection; A user interaction module, which is used to display scheduling and early warning information and receive user feedback and instructions; A data storage and security module, which is used to store various types of data and implement access control and encryption protection.

2. An OA management system for a smart employment project based on cloud computing, characterized in that, The modules are implemented through the following methods: S1. The labor task data, human resource data, approval process, working hour records and task execution data are collected through the data acquisition and preprocessing module to construct a data set; S2. The data in the data set is processed by the dynamic task scheduling module based on the improved bacterial foraging optimization algorithm, the task scheduling objective function and constraint conditions are established, and the scheduling scheme with the minimum scheduling objective function is output; S3. The wavelet transform and the adaptive anomaly detection algorithm based on island model evolution learning are implemented on the data set through the real-time anomaly detection and risk early warning module, multi-scale time-frequency features are extracted, and real-time risk early warning information is generated; S4. The optimal task scheduling scheme and the risk early warning information are integrated through the closed-loop linkage module, the scheduling scheme is analyzed, and the result is fed back to the business logic layer of the OA system to realize closed-loop management; S5. The integrated information is displayed on the user interface through the user interaction and feedback module, and user input and feedback are received; S6. The collected data, task scheduling scheme, risk early warning information and user feedback data are stored in the cloud database through the data storage and security management module, and access control and encryption measures are implemented on the data.

3. The OA management system for the intelligent employment project based on cloud computing according to claim 2, wherein, The data set includes the following data subsets: the labor task data includes task ID, task type, task start time, task end time, task priority, and task completion status; the human resource data includes personnel ID, personnel ability, personnel working time, personnel current task assignment status, and personnel skill tags; the approval process data includes approval process ID, approval time, approval status, and approval personnel ID; the working hour record data includes personnel ID, working hour record, working hour type, and working hour date; the task execution data includes task ID, task execution personnel ID, task execution start time, task execution end time, and task completion percentage.

4. The OA management system for the intelligent employment project based on cloud computing according to claim 2, wherein, The specific content of S2 includes: S21. Establish a scheduling variable matrix representation based on the collected data subset; S22. Set the number of bacteria population, the maximum number of iterations, the adaptive step factor, the task constraint conditions and the personnel ability boundary, and initialize the current position solution of each bacterial individual. The current position solution corresponds to a feasible task scheduling scheme that meets the constraint conditions. Among them, each bacterial individual represents a scheduling strategy, and this scheduling strategy forms the position vector of the bacterial individual through vectorized encoding of the task scheduling variable matrix. Specifically, the scheduling variable matrix is a two-dimensional array, where the rows represent the human resource numbers, the columns represent the task numbers, and the element values in the matrix are 0 or 1, indicating whether a certain task is assigned to a certain person. This matrix is sequentially expanded into a one-dimensional vector by column or by row order as the position representation of the bacterial individual, and participates in the chemotaxis, peristaltic mutation, and elimination dispersion calculations in the scheduling optimization process; S23. In the chemotaxis movement stage, calculate the bacterial population scheduling solution. The bacterial population scheduling solution is defined as the assignment strategy of bacterial individuals to each task: Among them, X (k) represents the allocation strategy of each bacterium individual to each task in the k-th iteration. Based on the task priority factor γ i and gradient feedback, the motion formula for updating the bacterium position is: Among them, X (k+1) is the allocation strategy of each bacterium individual to each task in the (k + 1)-th iteration, C is the adaptive step size factor, δ is the perturbation adjustment coefficient, rand(q) is a random number function generated within the interval [-1, 1], represents the gradient direction of the fitness function; S24. Update the allocation policy X (k+1) Discretize it into the form of a scheduling matrix, denoted as X. Introduce a local mutation operation during the peristaltic stage to achieve local search optimization through the XOR perturbation matrix M. The specific operation is as follows: Among them, represents the exclusive OR operator, and M is the perturbation mask matrix; S25. In the elimination dispersion stage, calculate the fluctuation entropy value η of the current bacterial population scheduling solution. When η < θ, perform bacterial re-initialization, where θ is the set entropy threshold, and the calculation formula of η is: where p k represents the probability distribution of a bacterial individual falling into the k-th local region, and log represents the logarithmic function; S26. When the number of iterations reaches the maximum value or the optimal solution does not update for several consecutive generations, output the scheduling scheme that minimizes the task scheduling objective function and transmit it to the OA system business logic layer for execution.

5. The OA management system for the intelligent employment project based on cloud computing according to claim 2, characterized in that, The specific content of S3 includes: S31. Based on the scheduling scheme currently output by the task scheduling module, extract the behavior data related to the scheduling tasks, executing personnel, and approval nodes, and construct a unified multi-dimensional time series data set. The data set includes the approval process duration sequence, the working hour record sequence, and the task execution feedback sequence; S32. Select the orthogonal wavelet basis function and perform J-layer discrete wavelet decomposition on the multi-dimensional time series data set to obtain a set of scale components; S33. Construct feature vectors based on each scale component, and splice all scale feature vectors to form the final input feature set F; S34. Assign the feature set F to the isolated island evolution model composed of M heterogeneous sub-populations {Ω1, Ω2, …, Ω M}, and each sub-population independently performs abnormal boundary learning using different individual coding strategies and local evolution mechanisms, specifically including: S341. Initialize the heterogeneous population Ω m of the individual set where represents the boundary model parameters of the k-th individual in the feature space; S342. The density-driven adaptive mutation mechanism is applied to the island evolution learning process of the anomaly detection module to dynamically update the individual boundary model parameters in each sub-population. The mechanism combines the changes in the behavior feature distribution of the approval process duration sequence, working hour record sequence, and task execution feedback sequence data in the intelligent employment scenario, and implements targeted updates to the individual parameters. The density-driven adaptive mutation mechanism is used to update the individual set parameters, and the update formula is: where randn() is a standard normal distribution random perturbation, α is the perturbation amplitude coefficient, β is the density sensitivity adjustment factor, Δρ is the current sample density change, and ρ base is the initial reference density; S343. Define the fitness function Φ m (Z) Evaluate the individual performance and calculate the updated individual Fitness; S344. In the selection of individuals in each generation, retain the abnormal score individuals in the previous generation that are higher than the preset value according to the calculated fitness Perform a mutation operation in its neighborhood, retain the historical key abnormal boundaries, and achieve long-term memory retention of abnormal features; S35. Perform cross-group migration operations every fixed number of generations τ, and inject some individuals with abnormal scores higher than the preset value into other groups for boundary fusion to obtain an integrated abnormal score function: Among them, A m (t) is the anomaly score of the input data for the population Ω m at time t; S36. Set the abnormal determination threshold θ A , and generate an abnormal alarm signal: Return R(t) to the closed-loop linkage module, and after meeting the requirements, feedback it to adjust the scheduling scheme and respond to system risks.

6. The OA management system for the intelligent employment project based on cloud computing according to claim 2, characterized in that, The specific content of S4 includes: S41. Perform correlation analysis on the anomaly warning signal R(t) output by the anomaly detection module and the current minimum scheduling scheme, and identify the anomaly impact factors according to the anomaly type and related tasks, including: Identification of high-risk task units: If the data time period corresponding to the abnormal signal completely coincides with the task execution time and the abnormal score A(t) ≥ θ A , then mark the task unit to which the task belongs as a high-risk task unit; Unstable resource node identification: If a resource is continuously at risk or abnormal in the tasks assigned to it, mark this resource as an unstable node; Sensitive task pattern extraction: If multiple schedules of a certain task trigger anomaly scores at different time periods, mark this task as a sensitive type; S42. Adjust the scheduling constraints and the scheduling objective function according to the identification results, and trigger the task scheduling adaptive adjustment mechanism: When a high-risk task unit appears, increase its weight coefficient in the objective function and limit the resource allocation range; When an unstable resource node is detected, impose resource load limit constraints and adjust the upper limit of tasks that can be assigned to personnel; When a sensitive task pattern is identified, trigger a reconstruction strategy, generate a new schedule and update the task assignment rules; S43. Re - execute the improved bacterial foraging optimization algorithm for the updated task scheduling objective function and constraint conditions to generate an optimized scheduling plan X ′ ; S44. Evaluate the difference between the optimized scheduling plan and the current scheduling plan. The difference evaluation indicators include: Task assignment difference rate Personnel reassignment ratio Resource stability change coefficient Δ S = σ new - σ orig ; Among them, Task orig represents the task assignment set in the original scheduling plan, and Task new represents the task assignment set of the optimized scheduling plan. For Task orig △Task new the △ in it is the symmetric difference operator, representing the new set composed of the different elements in the two sets. represents the indicator function. If the task set of person h j changes, it is 1. m is the total number of human resources. is the original person-task set. is the new person-task set. σ orig is the standard deviation of the number of tasks of each person under the original scheduling plan, and σ new is the standard deviation under the new scheduling plan. S45. If the difference degree is within the acceptable range, then replace the original scheduling plan X with the optimized scheduling plan X, and write it into the scheduling result database; otherwise, retain the original plan, record the exception feedback, and perform policy correction later. ′ Replace the original plan X and write it into the scheduling result database; otherwise, retain the original plan, record the exception feedback, and perform policy correction later. S46. Synchronously update the scheduling decision log, write the log into the task scheduling module, and feedback it to the business logic layer of the OA system.

Images