Digital employee management method, system and device, storage medium and program product thereof

Abstract

The invention discloses a digital employee management method, system and device, a storage medium and a program product thereof, and relates to the technical field of data processing, and the digital employee management method comprises the steps: obtaining the real-time business data of digital employees; based on real-time business data, a performance evaluation result is obtained through a preset multi-dimensional performance evaluation model based on multi-dimensional factor evaluation influencing performance, dynamic adaptability is improved, and evaluation comprehensiveness is improved at the same time; based on the real-time business data, a causal relationship between the real-time business data and future performance is deeply mined through a performance prediction model based on causal reasoning driving, and a performance prediction result is accurately predicted and obtained; according to the real-time business data, the actual performance evaluation result and the predicted performance prediction result, generating a performance improvement suggestion; according to the performance improvement suggestion, accurate performance improvement suggestions can be provided for digital employees, and accurate optimization of the digital employees is realized, so that the working efficiency of the digital employees is improved, and the resource utilization rate is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a digital employee management method, system, device, storage medium, and program product thereof. Background Art

[0002] Digital employees refer to automated workflows achieved through technologies such as software robots (Robotic process automation, RPA) and artificial intelligence, which can replace manual labor to complete repetitive and regular work tasks.

[0003] Currently, the management of digital employees usually relies on post-event statistical data. This method is often lagging and cannot fully reflect the real-time work status and results of digital employees. It has high calculation delays and lacks dynamic adaptability. Moreover, it fails to fully consider the multi-dimensional factors affecting performance and their causal relationships, resulting in inaccurate performance evaluation of digital employees.

[0004] Based on the above, problems such as poor dynamic adaptability and inaccurate performance evaluation will affect the optimization of digital employees, reduce the work efficiency of digital employees, and ultimately reduce resource utilization. Summary of the Invention

[0005] The main purpose of this application is to provide a digital employee management method, system, device, storage medium and program product thereof, aiming to solve the technical problem of low resource utilization of digital employees.

[0006] To achieve the above objectives, this application proposes a digital employee management method, which includes:

[0007] Get real-time business data from digital employees;

[0008] Based on the real-time business data, a performance evaluation result is obtained by presetting a multi-dimensional performance evaluation model;

[0009] Based on the real-time business data, a performance prediction is performed using a performance prediction model driven by causal reasoning to obtain a performance prediction result;

[0010] Based on the real-time business data, the performance evaluation results and the performance prediction results, performance improvement suggestions are generated.

[0011] In addition, to achieve the above objectives, the present application also proposes a digital employee management system, which includes a data layer, a core processing layer, and an application layer;

[0012] The data layer is used to manage multi-source heterogeneous data, which includes real-time business data, performance evaluation results and performance improvement suggestions;

[0013] The core processing layer includes a performance management module, which is used to obtain performance evaluation results based on the real-time business data using a preset multi-dimensional performance evaluation model, perform performance prediction based on the real-time business data using a performance prediction model driven by causal reasoning to obtain performance prediction results, and generate performance improvement suggestions based on the real-time business data, the performance evaluation results, and the performance prediction results;

[0014] The application layer is used to visually display performance evaluation results, performance prediction results and performance improvement suggestions for managers to monitor.

[0015] In addition, to achieve the above-mentioned purpose, the present application also proposes a digital employee management device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the digital employee management method described above.

[0016] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the digital employee management method described above are implemented.

[0017] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the digital employee management method described above.

[0018] One or more technical solutions proposed in this application have at least the following technical effects:

[0019] This application obtains real-time business data of digital employees; based on the real-time business data, a performance evaluation result is obtained by pre-setting a multi-dimensional performance evaluation model based on the multi-dimensional factors affecting performance, thereby reducing calculation delays, improving dynamic adaptability, and improving the comprehensiveness of the evaluation; based on the real-time business data, a performance prediction model driven by causal reasoning is used to deeply explore the causal relationship between real-time business data and future performance, and predict performance prediction results, thereby improving the accuracy of performance prediction results; further, based on the real-time business data, actual performance evaluation results and predicted performance prediction results, targeted, comprehensive and accurate performance improvement suggestions are dynamically generated; the performance improvement suggestions can provide digital employees with accurate performance improvement suggestions, achieve accurate optimization of digital employees, thereby effectively improving the work efficiency of digital employees, and further improving resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] Figure 1 A first flow chart of the first embodiment of the digital employee management method of the present application is provided;

[0023] Figure 2 A schematic diagram of the first scenario provided in Example 1 of the digital employee management method of this application;

[0024] Figure 3 A schematic diagram of a second scenario provided in Example 1 of the digital employee management method of this application;

[0025] Figure 4 A second flow chart provided for Example 2 of the digital employee management method of this application;

[0026] Figure 5 A schematic diagram of the third scenario provided in Example 2 of the digital employee management method of this application;

[0027] Figure 6 A third flow chart provided for the third embodiment of the digital employee management method of this application;

[0028] Figure 7 A schematic diagram of the fourth scenario provided in Example 3 of the digital employee management method of this application;

[0029] Figure 8 A fourth flow chart provided for the fourth embodiment of the digital employee management method of this application;

[0030] Figure 9 A schematic diagram of the fifth scenario provided in Example 4 of the digital employee management method of this application;

[0031] Figure 10 This is a schematic diagram of the device structure of the hardware operating environment involved in the digital employee management method in the embodiment of the present application.

[0032] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0033] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0034] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0035] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution capabilities, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device capable of implementing the above functions, a digital employee management system, etc. This embodiment and the following embodiments will be described below using the digital employee management system as an example.

[0036] Based on this, the embodiment of the present application provides a digital employee management method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the digital employee management method of this application.

[0037] In this embodiment, the digital employee management method includes steps S10 to S40:

[0038] Step S10, obtaining real-time business data of digital employees;

[0039] It should be noted that the digital employee management method can be applied to a digital employee management system, which includes a data layer, a core processing layer, and an application layer. The core processing layer includes a performance management module. Figure 2 , through the digital employee management system, the complete process from data collection, analysis and processing to decision support can be realized.

[0040] Current digital workforce management typically relies on post-hoc statistical analysis. This approach is often delayed and fails to fully reflect the real-time status and effectiveness of digital workers. It suffers from high computational latency and lacks dynamic adaptability. Furthermore, it fails to fully consider the multi-dimensional factors influencing performance and their inter-causal relationships, resulting in inaccurate performance evaluations of digital workers. These issues, such as poor dynamic adaptability and inaccurate performance evaluations, hinder digital worker optimization, reduce their efficiency, and ultimately decrease resource utilization.

[0041] This embodiment obtains real-time business data of digital employees through the data layer, laying a data foundation for subsequent real-time dynamic management.

[0042] Among them, the real-time business data of digital employees includes task execution data, interaction records, resource usage, etc.

[0043] Step S20, obtaining a performance evaluation result based on the real-time business data by using a preset multi-dimensional performance evaluation model;

[0044] Since traditional digital employee management usually adopts static evaluation methods and is difficult to adapt to the rapidly changing business environment, the performance management module of this embodiment analyzes real-time business data through a preset multi-dimensional performance evaluation model, analyzes the relationship between business data in various dimensions and performance results, and can obtain more comprehensive performance evaluation results.

[0045] Among them, the preset multi-dimensional performance evaluation model can be a framework that comprehensively considers multiple performance evaluation dimensions, and the multiple performance evaluation dimensions can be dimensions such as efficiency, quality, reliability and cost; the preset multi-dimensional performance evaluation model can be obtained by training the PPO (Proximal Policy Optimization) algorithm or the Analytic Hierarchy Process (AHP) based on historical business data.

[0046] Step S30, based on the real-time business data, a performance prediction is performed using a performance prediction model driven by causal reasoning to obtain a performance prediction result;

[0047] Since traditional digital employee management methods cannot accurately capture complex causal relationships, they limit the accuracy of performance forecasts and the effectiveness of management decisions. The performance management module of this embodiment analyzes real-time business data through a performance forecasting model driven by causal reasoning, explores the causal relationship between business data and performance results, and obtains accurate performance forecast results.

[0048] Among them, the performance prediction model driven by causal reasoning can be obtained by training a structural equation model (SEM) or a causal forest based on historical business data.

[0049] Step S40: generating performance improvement suggestions based on the real-time business data, the performance evaluation results and the performance prediction results.

[0050] Since traditional digital employee management methods usually lack a highly targeted improvement suggestion mechanism, it is difficult to meet the personalized management needs of digital employees, which affects the effectiveness of optimization measures. The performance management module of this embodiment generates performance improvement suggestions based on the real-time business data of digital employees, the performance evaluation results obtained by real-time evaluation based on a preset multi-dimensional performance evaluation model, and the performance prediction results obtained by a performance prediction model driven by causal reasoning.

[0051] That is, by analyzing the real-time business data of digital employees in real time, and combining the analysis results to provide targeted performance improvement suggestions for the digital employees; the performance improvement suggestions may include optimization suggestions at the task execution level (for example, optimization suggestions for process processing logic, optimization suggestions for exception handling mechanisms), resource scheduling and load balancing suggestions (for example, task allocation suggestions, task scheduling suggestions, etc.), performance optimization suggestions (internal software update suggestions, script optimization suggestions), etc.; it can be understood that the performance improvement suggestions can provide digital employees with accurate performance improvement suggestions, achieve accurate optimization of digital employees, thereby effectively improving the work efficiency of digital employees, and then improving resource utilization.

[0052] Specifically, the implementation method of generating performance improvement suggestions based on the real-time business data, the performance evaluation results, and the performance prediction results may be:

[0053] Leverage data analysis tools and techniques to conduct in-depth analysis of real-time business data to identify key factors influencing performance; combine historical performance data with current performance evaluation results to uncover the causes of performance fluctuations; and develop personalized performance improvement recommendations based on the specific challenges faced by different business areas or teams.

[0054] In addition, refer to Figure 3 The implementation method of generating performance improvement suggestions based on the real-time business data, performance evaluation results and performance prediction results may also be:

[0055] Based on the real-time business data, a preset knowledge graph is updated, wherein the preset knowledge graph is constructed based on entities and their first attributes and the relationships between entities, and the entities include employees, skills and tasks; features are extracted from the updated knowledge graph through a graph attention network to obtain node embeddings; and performance improvement suggestions are generated based on the node embeddings, the performance evaluation results and the performance prediction results.

[0056] It should be noted that the preset knowledge graph is constructed based on entities, their first attributes and the relationships between entities, where the first attribute is a metadata attribute used to describe the entity, and the entity (corresponding to the first attribute) includes employees (ID, name, department, position and other attributes), skills (name, category, proficiency level and other attributes), tasks (ID, name, description, difficulty level and other attributes), tools (name, type, version and other attributes) and projects (name, start date, end date, status and other attributes); the relationships between entities include the mastery relationship between employees and skills, the responsibility, participation, completion and other relationships between employees and tasks, the familiarity, use and need to learn relationship between employees and tools, the need and help between skills and tasks, the dependence and optional use between tasks and tools, and the participation and leadership between employees and projects.

[0057] Neo4j (a graph database) can be used to store and manage pre-set knowledge graphs. During initialization, basic data is imported from HR (Human Resources System, an information platform) systems, project management systems, and other systems. The pre-set knowledge graph is updated weekly by synchronizing the latest data and capturing events in real time, ensuring that the nodes and edges in the graph reflect the latest business conditions and individual performance.

[0058] Furthermore, the updated knowledge graph is subjected to feature extraction through a Graph Attention Network (GAT), and an implementation method for obtaining node embeddings can be to perform feature initialization on the nodes in the knowledge graph to obtain node embeddings:

[0059] Employee node: department and position encoded in read-hot format, and years of experience and performance scores output as continuous values;

[0060] Skill node: embedding vector of skill name and skill category encoded in read-hot encoding;

[0061] Task node: embedding vector of task description and difficulty level encoded in read-hot encoding;

[0062] Tool node: Embedding vector of tool name and tool type encoded in read-hot encoding.

[0063] The above approach can capture the role of employees in the preset knowledge graph and their relationship with other entities (such as skills, tasks, tools, etc.), thereby obtaining a more comprehensive node embedding representation.

[0064] The graph attention network includes an input layer, a hidden layer, and an output layer. The attention mechanism can be applied in the hidden layer to calculate the attention weights between nodes, and the parameters of the graph attention network to be trained are optimized by minimizing the cross entropy loss of the node classification task during the training process to obtain a trained graph attention network.

[0065] Furthermore, the node embeddings output by the graph attention network can be combined with performance evaluation and performance prediction results as input to the sequence-to-sequence (Seq2Seq) encoder. Bahdanau (attention mechanism) is used to enhance the model's focus on key information, helping to more accurately understand an employee's current status and potential improvement points. Specific performance improvement suggestion text is generated through the sequence-to-sequence model's decoder (which uses a unidirectional LSTM (Long Short-Term Memory) structure to generate specific optimization suggestions for each employee based on attention weights. The final layer of the decoder is a vocabulary-sized softmax (activation function) layer, which is used to generate suggestion text in natural language). This process not only considers the effectiveness of historical optimization suggestions, but also incorporates current real-time business data, real-time performance evaluation results, and performance prediction results, making the proposed performance improvement suggestions more personalized and forward-looking. To ensure the effectiveness and relevance of performance improvement suggestions, a feedback collection mechanism can be established to record the adoption status and implementation effect score of each suggestion, and the sequence-to-sequence model can be updated using incremental learning and periodic full retraining. At the same time, set performance monitoring indicators (such as recommendation adoption rate, average implementation effect score, etc.), and compare the performance of different models through A / B (comparative testing) to ensure that the optimal model can be used in real time. Figure 3 .

[0066] Combining knowledge graphs and deep learning technologies, a knowledge graph of digital employee capabilities and task characteristics is constructed; a graph attention network (GAT) is used for feature extraction, and a sequence-to-sequence (Seq2Seq) model is adopted to generate personalized optimization suggestions; through continuous learning and updating, the quality and relevance of suggestions are continuously improved; this method can accurately capture the unique characteristics, skill level and work performance of each digital employee, and combine current tasks and project requirements to generate highly personalized optimization suggestions, achieving precise optimization of digital employees, thereby effectively improving the work efficiency of digital employees and significantly improving resource utilization.

[0067] This embodiment obtains real-time business data of digital employees; based on the real-time business data, a performance evaluation result is obtained by evaluating the multi-dimensional factors affecting performance through a preset multi-dimensional performance evaluation model, thereby reducing calculation delay, improving dynamic adaptability, and improving the comprehensiveness of the evaluation; based on the real-time business data, a performance prediction model driven by causal reasoning is used to deeply explore the causal relationship between real-time business data and future performance, and predict performance prediction results, thereby improving the accuracy of performance prediction results; further, based on the real-time business data, the actual performance evaluation results and the predicted performance prediction results, targeted, comprehensive and accurate performance improvement suggestions are dynamically generated; the performance improvement suggestions can provide accurate performance improvement suggestions for digital employees, achieve accurate optimization of digital employees, thereby effectively improving the work efficiency of digital employees, and further improving resource utilization.

[0068] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 4 , step S20 includes steps S01 to S03:

[0069] Step S01, defining a state space containing the real-time business data, wherein the real-time business data includes performance evaluation indicators, historical performance data, and external environmental factors;

[0070] It should be noted that the preset multi-dimensional performance evaluation model includes a strategy network.

[0071] In order to accurately analyze the performance evaluation results, this embodiment integrates multi-dimensional real-time business data (such as current performance evaluation indicators, historical performance data and external environmental factors) by constructing a dynamic Bayesian network (DBN), which can more comprehensively reflect the working status of digital employees. Figure 10 .

[0072] Real-time business data includes performance evaluation indicators, historical performance data, and external environmental factors:

[0073] Performance evaluation indicators include the number of task executions, customer satisfaction, execution time, execution success rate, implementation coverage, user activity, user retention rate, update frequency, value creation, productivity index, etc.; historical performance data can be historical performance data, which includes daily performance values ​​of the past 30 days, forming a 30x10 matrix (30 days, 10 indicators); external environmental factors include working day / holiday marks (0 or 1), seasonal indicators (spring, summer, autumn and winter, represented by 0-3), market volatility index (floating point number in the range of 0-1), team size changes (percentage relative to the benchmark), degree of technology update (floating point number in the range of 0-1), etc.

[0074] Total state space dimensions: 10 (current performance evaluation indicators) + 300 (historical performance data) + 5 (external environmental factors) = 315 dimensions.

[0075] Step S02: Based on the state space, a strategy network is used to calculate and obtain weight adjustment suggestions for multiple evaluation dimensions;

[0076] Furthermore, based on the state space, the strategy network is used to calculate and obtain weight adjustment suggestions for multiple evaluation dimensions, where the multiple evaluation dimensions include efficiency, quality, innovation, collaboration, growth, etc. That is, the action space is defined as the weight adjustment of these multiple evaluation dimensions, each weight range is between 0 and 1, and the softmax (activation function) function is used to ensure that the sum of the weights of the multiple evaluation dimensions is 1. Figure 10 .

[0077] The structure of the policy network may include:

[0078] Input layer: 315 neurons (corresponding to the dimension of the state space);

[0079] Hidden layer 1: 256 neurons, ReLU (activation function) activation;

[0080] Hidden layer 2: 128 neurons, ReLU (activation function) activation;

[0081] Output layer: 5 neurons (corresponding to 5 evaluation dimensions), softmax (activation function) activation.

[0082] It is understandable that the weights of multiple evaluation dimensions can be flexibly adjusted according to real-time business data and external influencing factors to make the evaluation more in line with the actual situation.

[0083] Step S03: determining a performance evaluation result based on the weight adjustment suggestion and the real-time business data.

[0084] Since static performance evaluation methods are currently widely used, it is difficult to adjust evaluation standards in real time according to the rapidly changing business environment, resulting in the performance evaluation results being unable to accurately reflect the actual performance of digital employees; this embodiment combines real-time business data and external environmental factors, and dynamically adjusts the weight of performance evaluation through the proximal policy optimization (PPO) algorithm. It can quickly respond to changes in the business environment, adjust the evaluation strategy in a timely manner, ensure the effectiveness of the evaluation results, and thus provide more accurate and meaningful evaluation results.

[0085] Further, refer to Figure 5The preset multi-dimensional performance evaluation model further includes a value network. After determining the performance evaluation result based on the weight adjustment suggestion and the real-time business data, the following steps may be performed:

[0086] Based on the state space and the weight adjustment suggestion, the value network is used to predict the total amount of future cumulative rewards that can be obtained by applying the adjusted weights for performance evaluation; manual feedback results are obtained for the performance evaluation results; based on the performance evaluation results and the manual feedback results, the reward value is calculated through a preset reward function; based on the total amount of future cumulative rewards and the reward value, the preset multi-dimensional performance evaluation model is updated.

[0087] The structure of the value network can include:

[0088] Input layer: 315 neurons;

[0089] Hidden layer 1: 256 neurons, ReLU (activation function) activation;

[0090] Hidden layer 2: 128 neurons, ReLU (activation function) activation;

[0091] Output layer: 1 neuron, linear activation.

[0092] The manual feedback results include manual evaluation scores, and the preset reward function R can be:

[0093] R = 0.6*business goal achievement + 0.3*manual evaluation score + 0.1*(1-weight change range).

[0094] Business goal achievement = Σ(wi*(1-|KPIi-Targeti| / Targeti)) / Σwi;

[0095] Among them, wi is the importance weight of each performance appraisal indicator, KPIi is the current performance appraisal indicator, and Targeti is the target performance appraisal indicator; the manual evaluation score is the management's satisfaction with the performance evaluation results (1-5 points, normalized to 0-1); the weight change range is the Euclidean distance between the weights before and after adjustment, normalized to 0-1.

[0096] It should be noted that the value network, based on the state space and the weight adjustment suggestions, predicts the total future cumulative rewards that can be obtained by applying the adjusted weights for performance evaluation. This not only focuses on short-term results but also predicts long-term effects. The estimation of future rewards provides important reference information for guiding the learning direction of the policy network and continuously optimizing model performance.

[0097] Furthermore, by obtaining manual feedback results for the performance evaluation results, it can be used as an additional source of information to supplement key details or subjective evaluations that may be missed by automatic evaluation, thereby increasing the transparency of the evaluation process.

[0098] Based on the performance evaluation results and manual feedback results, the reward value is calculated through a preset reward function, taking into account multiple aspects such as the achievement of business goals, manual evaluation and weight stability, and calculating the reward value of the preset reward function. Based on the sum of the future cumulative rewards and the reward value, the preset multi-dimensional performance evaluation model is updated. That is, by continuously learning and adjusting the model parameters, a model that can accurately evaluate performance is obtained.

[0099] Specifically, based on the sum of the future accumulated rewards and the reward value, an implementation method of updating the preset multi-dimensional performance evaluation model may be: updating the preset multi-dimensional performance evaluation model using stored experience data every week:

[0100] For each mini-batch (divide the training dataset into multiple mini-batches of size 64):

[0101] Calculate the advantage estimate: A(s,a)=R+γV(s')-V(s);

[0102] Calculate the policy loss: L = E[min(ratio*A,clip(ratio,1-ε,1+ε)*A)];

[0103] where ratio = πnew(a|s) / πold(a|s), and ε is the clipping parameter (set to 0.2);

[0104] Calculate the value loss: L_v = MSE(V(s), R + γV(s'));

[0105] Calculate the entropy reward: S = -Σπ(a|s)logπ(a|s);

[0106] Total loss: L_total = -L + 0.5*L_v - 0.01*S;

[0107] The preset multi-dimensional performance evaluation model is updated based on the total loss; offline large-scale training is performed once a month using data from the past 3 months; the exploration rate ε is set to 0.1, and the weights are randomly adjusted with a probability of 10% to maintain the model's exploration ability.

[0108] This embodiment uses a multi-dimensional performance evaluation model that can dynamically adjust the weights of different evaluation dimensions in real time based on real-time business data. It can respond to changes in the business environment in real time and dynamically adjust the evaluation weights, so that the evaluation results are more in line with actual business needs, thereby significantly improving the accuracy and adaptability of digital employee performance evaluation. At the same time, the model's adaptive learning ability enables it to continuously optimize evaluation strategies, reducing the need for human intervention.

[0109] Based on the first and second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction and will not be described in detail later. Figure 6 Before step S30, the digital employee management method further includes steps A01 to A05:

[0110] Step A01: Based on the performance evaluation indicators of digital employees, external environmental factors, and the causal relationship between the performance evaluation indicators, define nodes and their conditional probability distributions;

[0111] It should be noted that real-time business data specifically also includes multi-dimensional performance indicators such as work efficiency, work quality, innovation ability, team collaboration, learning and growth, as well as factors that affect these performance indicators such as workload, skill level, tool efficiency, team atmosphere, training investment, and market environment; the preset knowledge graph is constructed based on entities and their first attributes and the relationships between entities, and the entities include employees, skills, and tasks.

[0112] Specifically, refer to Figure 7 , the implementation method of constructing a preset knowledge graph can be:

[0113] Define the nodes of the preset knowledge graph: determine the performance indicator nodes (such as work efficiency, work quality, etc.) and influencing factor nodes (such as workload, skill level, etc.), and set the current value and conditional probability distribution for each node.

[0114] Establish edges of the preset knowledge graph: Initialize and update the causal relationship between nodes through domain expert knowledge and PC (Peter-Clar, constraint-based causal discovery method) algorithm, reflecting the identification and construction of causal relationships.

[0115] Step A02: constructing a dynamic Bayesian network based on the nodes and their conditional probability distributions, and the time dependencies between the nodes;

[0116] Furthermore, the time dependency between nodes can be determined: two time slices, t and t+1, are created for each node, and the time dependency between nodes is defined (for example, how the workload at time t affects the work efficiency at time t+1); this further strengthens the expression of causal relationships in the time dimension.

[0117] Step A03, generating an initial particle set for each time step, wherein each particle in the initial particle set is used to represent a state of each variable in the dynamic Bayesian network;

[0118] Reference Figure 7 ,Particle initialization: Generate an initial set of particles (e.g., 1000 particles) for each time step, where each particle represents a possible state of all variables in the dynamic Bayesian network; predict the state at time t+1 based on the conditional probability distribution of the dynamic Bayesian network, and use the forward sampling method to generate the prediction.

[0119] Step A04, predicting the state of each variable at the next time point using a particle filter algorithm based on the conditional probability distribution;

[0120] For each particle (particle), the conditional probability distribution of the dynamic Bayesian network is used, and the forward sampling method is adopted to generate the predicted state (particle_state) at time t+1.

[0121] Step A05: Based on the states of the variables at the next time point, the performance prediction model to be trained is trained to obtain a performance prediction model driven by causal reasoning.

[0122] Through the above method, time series analysis and dynamic Bayesian networks are combined for deep causal analysis, and a particle filter algorithm is used for online reasoning and prediction. Based on the state of each of the variables at the next time point, the performance prediction model to be trained is trained. The resulting performance prediction model driven by causal reasoning not only provides accurate performance predictions, but also explains the reasons behind the predictions.

[0123] Specifically, refer to Figure 7 The implementation method of training the performance prediction model to be trained based on the state of each of the variables at the next time point to obtain a performance prediction model driven by causal reasoning may be:

[0124] Intervene in the target variables in the dynamic Bayesian network to modify the dynamic Bayesian network; perform reasoning based on the modified dynamic Bayesian network to obtain the intervention effect produced by the modification; perform sensitivity analysis based on the intervention effect to obtain the causal influence relationship of each external environmental factor on the performance; and train the performance prediction model to be trained based on the state of each variable at the next time point and the causal influence relationship to obtain a performance prediction model driven by causal reasoning.

[0125] Through do-calculus, the target variable X is intervened, the dynamic Bayesian network is modified to reflect the intervention, and the intervention effect is recalculated by reasoning; based on the intervention effect, a sensitivity analysis is performed to evaluate the causal impact of various external environmental factors on performance; the importance is ranked based on the size of the impact, and the reference Figure 7 .

[0126] Data collection and storage: Actual performance data and environmental factor data are collected daily and stored in InfluxDB (a time series database). New data is used weekly to update the conditional probability distribution of the dynamic Bayesian network and optimize the model parameters of the performance prediction model driven by causal reasoning.

[0127] During the model application process, the performance prediction is performed based on the real-time business data using a performance prediction model driven by causal reasoning. The implementation method for obtaining the performance prediction result may be:

[0128] The particle weights are updated based on the real-time business data and the state of each variable at the next time point, and effective particles are determined based on the updated particle weights; the current state of the dynamic Bayesian network is estimated based on the weighted average of the effective particles; and with the current state as the starting point, performance prediction is performed using the causal reasoning-driven performance prediction model based on the real-time business data and the dynamic Bayesian network to obtain a performance prediction result.

[0129] It should be noted that when real-time business data is received, the likelihood of each particle is calculated based on the state of each variable at the next time point, and the particle weight is updated based on the real-time business data and the state of each variable at the next time point: L(particle)=ΠP(observation|

[0130] particle_state) - Update particle weight (w_t = w{t-1}*L(particle)); Calculate the effective number of particles (N_eff): N_eff = 1 / Σ(w_i^2); If N_eff < 500, perform resampling: use the system resampling method to select particles according to weight probability, copy high-weight particles, discard low-weight particles, and reset all particle weights to 1 / 1000.

[0131] Furthermore, the weighted average of the effective particles is used to estimate the current state of the dynamic Bayesian network: Σ(w_i*particle_state_i) / Σw_i; taking the current state as the starting point, a forward simulation is performed based on the real-time business data and the dynamic Bayesian network, and performance prediction is performed through the performance prediction model driven by causal reasoning to obtain a performance prediction result.

[0132] Among them, the performance forecast results output by the performance forecast model driven by causal reasoning include the performance forecast for the next 7 days and its 95% confidence interval, as well as a causal analysis report (used to provide a ranking of the importance of different factors on performance indicators to support decision-making).

[0133] This embodiment adopts the above-mentioned method to significantly improve the dynamic adaptability of performance evaluation, the depth of data analysis, and the flexibility of anomaly detection. Combined with the dynamic Bayesian network and particle filtering algorithm, it provides accurate performance prediction and in-depth causal analysis, and provides reliable decision support for personalized recommendations for digital employees.

[0134] Based on the first, second and third embodiments of the present application, in the fourth embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction and will not be described in detail later. Figure 8 After step S10, the digital employee management method further includes steps B01 to B06:

[0135] Step B01: performing anomaly detection on the real-time business data using a graph neural network to obtain anomaly detection results, wherein the graph neural network is obtained by iterative training based on a graph structure, and the graph structure is constructed based on preset anomaly detection rules and their second attributes and logical associations between the rules, wherein the second attributes include conditions, weights, and activation states;

[0136] It should be noted that the core processing layer also includes an exception management module, refer to Figure 2 , by adding an exception management module, it aims to identify, evaluate and respond to abnormal situations when digital employees perform a series of operations or processes, thereby reducing the negative impact of exceptions on business processes.

[0137] Specifically, the real-time business data also includes response time, CPU usage, memory usage, etc., and the anomaly detection results include long response time, high CPU usage, high memory usage, long problem duration, high degree of business interruption, etc.

[0138] In order to more clearly analyze the relationship between detection rules, each anomaly detection rule can be encoded as a node in the graph structure. The node attributes (second attributes) include: conditions ("response time > 5s"), weights, activation status, etc. The initial rule set can be generated based on domain expert knowledge and historical anomaly data.

[0139] The logical associations between rules are used as edges between nodes. Based on the constructed graph structure and historical business data, the graph neural network can be iteratively trained to learn the complex interactive relationships between rules. Figure 9 .

[0140] Step B02, obtaining the rule representation of each anomaly detection rule in the graph structure;

[0141] In order to improve the update adaptability of the graph neural network to abnormal events, it can be retrained with newly collected data every week to keep the graph neural network dynamically updated.

[0142] Specifically, the rule performance of each anomaly detection rule in the graph structure can be monitored in real time: that is, the triggering frequency, accuracy, and impact of each anomaly detection rule can be monitored.

[0143] Step B03: determining the importance score of each anomaly detection rule in the graph structure based on the rule performance;

[0144] Based on the rule performance, the importance score of each anomaly detection rule in the graph structure is calculated using a sliding time window (the last 30 days);

[0145] Step B04: adjusting the weight of each of the anomaly detection rules based on the importance score;

[0146] Dynamically adjust rule weights based on importance scores.

[0147] Step B05: updating each of the anomaly detection rules based on the adjusted weights to obtain an updated graph structure;

[0148] Based on the adjusted weights, each of the anomaly detection rules is updated to obtain an updated graph structure. Specifically, the weights of less important rules can be automatically reduced, while the weights of more important rules can be increased. Alternatively, the graph structure can be optimized monthly to remove rules that have not been triggered for a long time (90 days) or whose accuracy is below a threshold, merge highly relevant rules to reduce redundancy, or generate new candidate rules based on historical data and add them to the graph structure to obtain an updated graph structure.

[0149] In addition, before performing anomaly detection on the real-time business data using a graph neural network to obtain anomaly detection results, the following steps may be performed:

[0150] Acquire historical abnormal events; extract abnormal features based on the historical abnormal events through a preset feature extraction model; generate new abnormality detection rules based on the abnormal features through a preset meta-learning model; and add the new abnormality detection rules to the graph structure.

[0151] The preset meta-learning model can be obtained by iterative training using the MAML (Model-Agnostic Meta-Learning) algorithm. The model input of the preset meta-learning model is the abnormal event feature vector (CPU usage, memory usage, response time, etc.), and the model output of the preset meta-learning model is the parameters of the new rule (condition threshold, weight, etc.). Figure 9 .

[0152] The training process of the preset meta-learning model can be: building a task library, each task corresponds to a specific type of anomaly detection scenario, using the data in the task library to train the meta-model, learning the ability to quickly adapt to new tasks, and updating the meta-model every quarter using newly collected anomaly data.

[0153] The anomaly management module obtains historical anomaly events; based on the historical anomaly events, it extracts anomaly features through a preset feature extraction model; based on the anomaly features, it generates new anomaly detection rules through a preset meta-learning model, and verifies the effectiveness of the new anomaly detection rules in a simulation environment. If effective, the new anomaly detection rules are added to the graph structure.

[0154] Furthermore, we can track the actual performance of newly generated rules, use the reinforcement learning method policy gradient to optimize the rule generation strategy, regularly evaluate and eliminate poorly performing rules, achieve continuous optimization of the graph structure, and adapt to changes in abnormal events.

[0155] Step B06: Iteratively optimize the graph neural network based on the updated graph structure to obtain an optimized graph neural network, and perform anomaly detection based on the optimized graph neural network.

[0156] In order to dynamically detect and evaluate operational anomalies, this embodiment iteratively optimizes the graph neural network based on the updated graph structure in real time to obtain an optimized graph neural network. By performing anomaly detection based on the optimized graph neural network, it is possible to quickly respond to newly emerging abnormal patterns and improve real-time response capabilities.

[0157] Specifically, refer to Figure 9 Before obtaining real-time business data of digital employees, you can also:

[0158] Obtaining work orders to be assigned; calculating the priority scores of the work orders to be assigned based on the work orders to be assigned by using a preset multi-dimensional priority scoring model; and determining the optimal scheduling solution based on the priority scores and the objective function by using a multi-objective genetic algorithm.

[0159] It should be noted that the digital employee management system of this embodiment also includes a work order distribution system. Since the work order priority is fixed and difficult to adapt to real-time business demand changes during the execution of tasks by digital employees, operational efficiency and resource utilization are reduced.

[0160] This embodiment first defines multiple scoring dimensions: urgency, scope of impact, difficulty of resolution, customer importance, etc., and designs a scoring function for each dimension. For example, urgency = f (problem duration, degree of business interruption), scope of impact = g (number of affected users, system importance involved).

[0161] Based on the work orders to be assigned, a weighted summation method is used to calculate the priority scores of the work orders to be assigned by presetting a multi-dimensional priority scoring model (including a scoring function corresponding to each dimension). The initial weights can be set based on historical data and expert experience, and key business indicators (customer satisfaction, system availability) can be monitored in real time. When the indicators deviate from the target values ​​by more than a threshold, the weight adjustment is triggered, and the gradient descent method is used to optimize the weights to minimize the difference between the indicators and the target values.

[0162] Furthermore, the work order queue is encoded as a chromosome of a multi-objective genetic algorithm, and the work order priority is used as the gene of the multi-objective genetic algorithm; the objective function is defined: for example, minimizing the average processing time, maximizing resource utilization, maximizing customer satisfaction, etc.; based on the priority score and the objective function, the optimal order scheduling plan is determined through the multi-objective genetic algorithm.

[0163] Among them, the multi-objective genetic algorithm can be NSGA-II (Non-dominated Sorting Genetic Algorithm II, fast non-dominated sorting genetic algorithm II), NSOOA (Non-dominated Sorting Optimizer Algorithm, non-dominated sorting optimization algorithm), etc. The NSGA-II algorithm can be used to generate a series of non-dominated solutions (Pareto front), and the algorithm can be rerun every 15 minutes to adapt to the latest status.

[0164] Based on the priority score and the objective function, the implementation method of determining the optimal scheduling plan through a multi-objective genetic algorithm can be to generate a series of non-dominated solutions through a multi-objective genetic algorithm based on the priority score and the objective function, and determine the optimal scheduling plan from the non-dominated solutions based on a preset decision model.

[0165] Among them, deep reinforcement learning (Deep Q-Network, DQN) can be used to train the decision model:

[0166] Define the state space as the current work order queue status, resource status, and business KPIs (Key Performance Indicators).

[0167] The action space is defined as selecting the optimal solution from the solution set generated by NSGA-II;

[0168] The reward function is defined as a comprehensive consideration of the achievement of multiple business goals.

[0169] In addition, it should be noted that the core processing layer of this embodiment also includes an integrated decision-making module, which is used to integrate cross-module data, such as obtaining performance improvement suggestions from the performance management module, obtaining anomaly detection results from the anomaly management module, etc., and building a global view (for example, "performance decline is strongly correlated with a certain type of anomaly"); the integrated decision-making module is also used to generate global optimization strategies, for example, formulating cross-departmental collaboration strategies based on cross-module data (such as "adjusting the scheduling plan to balance the load"); the integrated decision-making module is also used for decision execution and feedback, for example, automatically issuing strategies (such as resource allocation instructions) and tracking the effects of closed-loop optimization.

[0170] The application layer of this embodiment can also provide a visualization interface, through which the application layer can display performance evaluation results, performance prediction results and performance improvement suggestions for managers to monitor; it can also display the expected effects of different plans in the performance improvement suggestions, thereby facilitating manual intervention and allowing administrators to manually adjust priorities; it can also record the actual processing time, resource consumption and processing results of each work order, compare the actual results with the expected effects, calculate the prediction error, and use the prediction error to update the DQN model to improve the accuracy of future decisions, analyze system performance on a monthly basis, and identify improvement points.

[0171] By real-time monitoring of work order processing results and changes in business indicators, this embodiment enables the system to dynamically adjust work order priorities, ensuring that resources are reasonably allocated to the most critical issues. This optimizes the work order processing process and effectively responds to complex and changing operating environments, thereby improving the work efficiency of digital employees. At the same time, by optimizing resource allocation, operating costs can be significantly reduced.

[0172] It should be noted that the above examples are only used to understand this application and do not constitute a limitation on the digital employee management method of this application. More simple transformations based on this technical concept are all within the scope of protection of this application.

[0173] This application also provides a digital employee management system, please refer to Figure 2 , the system includes a data layer, a core processing layer and an application layer;

[0174] The data layer is used to manage multi-source heterogeneous data, which includes real-time business data, performance evaluation results and performance improvement suggestions;

[0175] The core processing layer includes a performance management module, which is used to obtain performance evaluation results based on the real-time business data using a preset multi-dimensional performance evaluation model, perform performance prediction based on the real-time business data using a performance prediction model driven by causal reasoning to obtain performance prediction results, and generate performance improvement suggestions based on the real-time business data, the performance evaluation results, and the performance prediction results;

[0176] The application layer is used to visually display performance evaluation results, performance prediction results and performance improvement suggestions for managers to monitor.

[0177] In one embodiment, the preset multi-dimensional performance evaluation model includes a strategy network, and the performance management module includes:

[0178] A definition submodule, configured to define a state space containing the real-time business data, wherein the real-time business data includes performance evaluation indicators, historical performance data, and external environmental factors;

[0179] A suggestion submodule, configured to calculate, based on the state space, through a policy network, and obtain weight adjustment suggestions for multiple evaluation dimensions;

[0180] The performance evaluation submodule is used to determine a performance evaluation result based on the weight adjustment suggestion and the real-time business data.

[0181] In one embodiment, the preset multi-dimensional performance evaluation model further includes a value network. After the step of determining the performance evaluation result based on the weight adjustment suggestion and the real-time business data, the performance management module further includes:

[0182] A prediction submodule, configured to predict, through a value network, a total future cumulative reward that can be obtained by applying the adjusted weights for performance evaluation based on the state space and the weight adjustment suggestion;

[0183] A result acquisition submodule, used to obtain manual feedback results for the performance evaluation results;

[0184] A reward value calculation submodule, configured to calculate a reward value based on the performance evaluation result and the manual feedback result using a preset reward function;

[0185] The model updating submodule is used to update the preset multi-dimensional performance evaluation model based on the sum of the future accumulated rewards and the reward value.

[0186] In one embodiment, before the step of performing performance forecasting based on the real-time business data using a causal reasoning-driven performance forecasting model to obtain a performance forecast result, the performance management module further includes:

[0187] Define submodules for defining nodes and their conditional probability distributions based on the performance evaluation indicators of digital employees, external environmental factors, and the causal relationship between various performance evaluation indicators;

[0188] A network construction submodule, configured to construct a dynamic Bayesian network based on the nodes and their conditional probability distributions, and the temporal dependencies between the nodes;

[0189] A particle generation submodule, configured to generate an initial particle set for each time step, wherein each particle in the initial particle set is used to represent a state of each variable in the dynamic Bayesian network;

[0190] A state prediction submodule, configured to predict the state of each of the variables at the next time point using a particle filter algorithm based on the conditional probability distribution;

[0191] The model training submodule is used to train the performance prediction model to be trained based on the state of each of the variables at the next time point to obtain a performance prediction model driven by causal reasoning.

[0192] In one embodiment, the model training submodule includes:

[0193] An intervention unit, configured to intervene in a target variable in the dynamic Bayesian network to modify the dynamic Bayesian network;

[0194] An inference unit, used for performing inference based on the modified dynamic Bayesian network to obtain the intervention effect produced by the modification;

[0195] A sensitivity analysis unit, used to perform sensitivity analysis based on the intervention effect to obtain the causal impact of various external environmental factors on performance;

[0196] The model training unit is used to train the performance prediction model to be trained based on the state of each variable at the next time point and the causal influence relationship, so as to obtain a performance prediction model driven by causal reasoning.

[0197] In one embodiment, the performance management module further includes:

[0198] An effective particle determination submodule, configured to update particle weights based on the real-time service data and the state of each variable at a next time point, and determine effective particles based on the updated particle weights;

[0199] a state estimation submodule, configured to estimate the current state of the dynamic Bayesian network based on a weighted average of the effective particles;

[0200] The performance prediction submodule is used to perform performance prediction based on the current state, the real-time business data and the dynamic Bayesian network, and the causal reasoning-driven performance prediction model to obtain a performance prediction result.

[0201] In one embodiment, the performance management module further includes:

[0202] A graph updating submodule, configured to update a preset knowledge graph based on the real-time business data, wherein the preset knowledge graph is constructed based on entities, their first attributes, and relationships between entities, wherein the entities include employees, skills, and tasks;

[0203] The feature extraction submodule is used to extract features from the updated knowledge graph through the graph attention network to obtain node embeddings;

[0204] The suggestion generation submodule is used to generate performance improvement suggestions based on the node embedding, the performance evaluation results and the performance prediction results.

[0205] In one embodiment, the system further includes an exception management module, wherein the exception management module includes:

[0206] an anomaly management submodule, configured to perform anomaly detection on the real-time business data using a graph neural network to obtain anomaly detection results, wherein the graph structure is constructed based on preset anomaly detection rules and their second attributes and logical associations between the rules, wherein the second attributes include conditions, weights, and activation states;

[0207] A performance acquisition submodule, configured to acquire the rule performance of each anomaly detection rule in the graph structure;

[0208] an importance scoring submodule, configured to determine an importance score of each anomaly detection rule in the graph structure based on the rule performance;

[0209] A weight adjustment submodule, configured to adjust the weight of each of the anomaly detection rules based on the importance score;

[0210] A rule updating submodule, configured to update each of the anomaly detection rules based on the adjusted weights to obtain an updated graph structure;

[0211] The anomaly detection submodule is used to iteratively optimize the graph neural network based on the updated graph structure to obtain an optimized graph neural network, so as to perform anomaly detection based on the optimized graph neural network.

[0212] In one embodiment, before the step of performing anomaly detection on the real-time business data using a graph neural network to obtain anomaly detection results, the anomaly management module further includes:

[0213] The event acquisition submodule is used to obtain historical abnormal events;

[0214] A feature extraction submodule is used to extract abnormal features based on the historical abnormal events through a preset feature extraction model;

[0215] A rule generation submodule is used to generate new anomaly detection rules based on the anomaly features through a preset meta-learning model;

[0216] The rule adding submodule is used to add the new anomaly detection rule to the graph structure.

[0217] In one embodiment, the system further includes an exception management module, and the exception management module further includes:

[0218] The work order acquisition submodule is used to obtain the work orders to be assigned;

[0219] A priority scoring submodule, configured to calculate the priority score of the work order to be assigned based on the work order to be assigned by using a preset multi-dimensional priority scoring model;

[0220] The order scheduling scheme determination submodule is used to determine the optimal order scheduling scheme based on the priority score and the objective function through a multi-objective genetic algorithm.

[0221] The digital employee management system provided in this application, employing the digital employee management method described in the aforementioned embodiments, can address the technical issue of low resource utilization for digital employees. Compared to the prior art, the digital employee management system provided in this application offers the same beneficial effects as the digital employee management method described in the aforementioned embodiments. Other technical features of the digital employee management system are the same as those disclosed in the aforementioned embodiments and are not further elaborated upon here.

[0222] The present application provides a digital employee management device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the digital employee management method in the above-mentioned embodiment 1.

[0223] Reference below Figure 10, which shows a schematic diagram of the structure of a digital employee management device suitable for implementing an embodiment of the present application. The digital employee management device in the embodiment of the present application may include, but is not limited to, mobile terminals such as mobile phones, tablet computers, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PMPs (Portable Media Players), in-vehicle terminals (such as in-vehicle navigation terminals), and fixed terminals such as digital televisions and desktop computers. Figure 10 The digital employee management device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0224] like Figure 10 As shown, the digital employee management device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the digital employee management device. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and communication device 1009. The communication device 1009 can allow the digital employee management device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows a digital employee management device with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems can be implemented or have alternatively.

[0225] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0226] The digital employee management device provided in this application, utilizing the digital employee management method described in the aforementioned embodiments, can address the technical issue of low resource utilization for digital employees. Compared to the prior art, the beneficial effects of the digital employee management device provided in this application are the same as those of the digital employee management method described in the aforementioned embodiments. Other technical features of the digital employee management device are the same as those disclosed in the aforementioned embodiments and are not further elaborated upon here.

[0227] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0228] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0229] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer program) stored thereon, wherein the computer-readable program instructions are used to execute the digital employee management method in the above-mentioned embodiment.

[0230] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0231] The computer-readable storage medium may be included in the digital employee management device, or may exist independently without being incorporated into the digital employee management device.

[0232] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the digital employee management device, the digital employee management device is enabled to execute the digital employee management method.

[0233] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0234] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to the various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the prescribed logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the prescribed function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0235] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0236] The computer-readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned digital employee management method, thereby resolving the technical issue of low resource utilization in digital employees. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the digital employee management method provided in the aforementioned embodiments, and are not further elaborated here.

[0237] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned digital employee management method when executed by a processor.

[0238] The computer program product provided in this application can solve the technical problem of low resource utilization of digital employees. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the digital employee management method provided in the above embodiment, and will not be repeated here.

[0239] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A digital employee management method, characterized in that: The method includes: Get real-time business data from digital employees; Based on the real-time business data, a performance evaluation result is obtained by presetting a multi-dimensional performance evaluation model; Based on the real-time business data, a performance prediction is performed using a performance prediction model driven by causal reasoning to obtain a performance prediction result; Based on the real-time business data, the performance evaluation results and the performance prediction results, performance improvement suggestions are generated.

2. The method according to claim 1, wherein The preset multi-dimensional performance evaluation model includes a policy network. The step of obtaining a performance evaluation result based on the real-time business data through the preset multi-dimensional performance evaluation model includes: Defining a state space containing the real-time business data, wherein the real-time business data includes performance evaluation indicators, historical performance data, and external environmental factors; Based on the state space, a policy network is used to calculate and obtain weight adjustment suggestions for multiple evaluation dimensions; A performance evaluation result is determined based on the weight adjustment suggestion and the real-time business data.

3. The method according to claim 2, wherein The preset multi-dimensional performance evaluation model further includes a value network. After the step of determining the performance evaluation result based on the weight adjustment suggestion and the real-time business data, the method further includes: Predicting, using the value network, the sum of future cumulative rewards that can be obtained by applying the adjusted weights for performance evaluation based on the state space and the weight adjustment suggestion; Obtaining manual feedback results for the performance evaluation results; Calculating a reward value based on the performance evaluation result and the manual feedback result using a preset reward function; The preset multi-dimensional performance evaluation model is updated based on the sum of the future accumulated rewards and the reward value.

4. The method according to claim 1, wherein Before the step of performing performance prediction based on the real-time business data using a performance prediction model driven by causal reasoning to obtain a performance prediction result, the method further includes: Based on the performance evaluation indicators of digital employees, external environmental factors, and the causal relationship between each performance evaluation indicator, define nodes and their conditional probability distribution; Constructing a dynamic Bayesian network based on the nodes and their conditional probability distributions, and the time dependencies between the nodes; generating an initial particle set for each time step, wherein each particle in the initial particle set is used to represent a state of each variable in the dynamic Bayesian network; Based on the conditional probability distribution, predict the state of each variable at the next time point through a particle filtering algorithm; Based on the states of the variables at the next time point, the performance prediction model to be trained is trained to obtain a performance prediction model driven by causal reasoning.

5. The method according to claim 4, wherein The step of training the performance prediction model to be trained based on the state of each of the variables at the next time point to obtain a performance prediction model driven by causal reasoning includes: Intervening in a target variable in the dynamic Bayesian network to modify the dynamic Bayesian network; Inference is performed based on the modified dynamic Bayesian network to obtain the intervention effect produced by the modification; Conduct sensitivity analysis based on the intervention effect to obtain the causal relationship between each external environmental factor and performance; Based on the states of the variables at the next time point and the causal influence relationship, the performance prediction model to be trained is trained to obtain a performance prediction model driven by causal reasoning.

6. The method according to any one of claims 4 to 5, characterized in that The step of performing performance prediction based on the real-time business data by using a performance prediction model driven by causal reasoning to obtain a performance prediction result includes: updating particle weights based on the real-time business data and the state of each variable at a next time point, and determining valid particles based on the updated particle weights; estimating a current state of the dynamic Bayesian network based on a weighted average of the effective particles; Taking the current state as a starting point, based on the real-time business data and the dynamic Bayesian network, performance prediction is performed through the performance prediction model driven by causal reasoning to obtain a performance prediction result.

7. The method according to claim 1, wherein The step of generating performance improvement suggestions based on the real-time business data, the performance evaluation results, and the performance prediction results includes: Based on the real-time business data, updating a preset knowledge graph, wherein the preset knowledge graph is constructed based on entities and their first attributes and relationships between entities, the entities including employees, skills, and tasks; The updated knowledge graph is subjected to feature extraction through the graph attention network to obtain node embedding; A performance improvement suggestion is generated based on the node embedding, the performance evaluation result, and the performance prediction result.

8. The method according to claim 1, wherein After the step of obtaining real-time business data of digital employees, the following steps are also included: performing anomaly detection on the real-time business data using a graph neural network to obtain an anomaly detection result, wherein the graph neural network is obtained by iterative training based on a graph structure, and the graph structure is constructed based on preset anomaly detection rules and their second attributes and logical associations between the rules, wherein the second attributes include conditions, weights, and activation states; Obtaining rule representations of each anomaly detection rule in the graph structure; Determining an importance score for each anomaly detection rule in the graph structure based on the rule performance; Adjusting the weight of each of the anomaly detection rules based on the importance score; Based on the adjusted weights, each of the anomaly detection rules is updated to obtain an updated graph structure; The graph neural network is iteratively optimized based on the updated graph structure to obtain an optimized graph neural network, so as to perform anomaly detection based on the optimized graph neural network.

9. The method according to claim 8, wherein Before the step of performing anomaly detection on the real-time business data through a graph neural network to obtain an anomaly detection result, the method further includes: Get historical abnormal events; Based on the historical abnormal events, abnormal features are extracted using a preset feature extraction model; Based on the abnormal features, a new anomaly detection rule is generated through a preset meta-learning model; The new anomaly detection rule is added to the graph structure.

10. The method according to claim 1, wherein Before the step of obtaining the real-time business data of the digital employee, the method further includes: Get the work orders to be assigned; Based on the work order to be assigned, calculating the priority score of the work order to be assigned by using a preset multi-dimensional priority scoring model; Based on the priority score and the objective function, the optimal order scheduling scheme is determined through a multi-objective genetic algorithm.

11. A digital employee management system, characterized in that: The system includes a data layer, a core processing layer and an application layer; The data layer is used to manage multi-source heterogeneous data, including real-time business data, performance evaluation results and performance improvement suggestions; The core processing layer includes a performance management module, which is used to obtain performance evaluation results based on the real-time business data using a preset multi-dimensional performance evaluation model, perform performance prediction based on the real-time business data using a performance prediction model driven by causal reasoning to obtain performance prediction results, and generate performance improvement suggestions based on the real-time business data, the performance evaluation results, and the performance prediction results; The application layer is used to visually display performance evaluation results, performance prediction results and performance improvement suggestions for managers to monitor.

12. A digital employee management device, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the digital employee management method according to any one of claims 1 to 10.

13. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the digital employee management method according to any one of claims 1 to 10 are implemented.

14. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the digital employee management method according to any one of claims 1 to 10 are implemented.

Images