Human resource information processing method based on RPA robot

Abstract

The invention relates to the field of data processing, and discloses a human resource information processing method based on an RPA robot, and the method comprises the following steps: collecting graph structure data at the beginning of each stage of a project, the graph structure data comprises nodes and edges connected with the nodes, one node represents one employee or one project, and the other node represents one employee or one project; one edge represents the relationship between the employee and the project; encoding the collected graph structure data into sequence data, wherein the sequence data comprises n sequence units; the process model comprises a first module, a second module and an output module, one sequence unit is input into the first module, the first module outputs a first hidden state of the sequence unit, and the second module inputs first hidden states of all the sequence units and outputs a second hidden state; and the output layer inputs the second hidden state and outputs the employees allocated to the project. According to the invention, a plurality of appropriate employees can be drawn for a project to form a project group for completing project work.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a human resources information processing method based on an RPA robot. Background Art

[0002] In modern enterprises, more and more work projects require the combination and collaboration of employees across departments and fields. For example, in the process of developing new products, a large manufacturing company needs to form a virtual work team with employees from multiple departments such as R&D, marketing, production, and supply chain to work closely together. However, due to the obvious differences in knowledge, skills, work habits, and communication methods among employees in different departments and fields, problems such as poor communication, inconsistent work rhythms, and conflicts in concepts and cognition often occur, which seriously affects the efficiency of cross-departmental collaboration, slows down the progress of R&D, and reduces product quality.

[0003] In order to solve the above-mentioned problem of inefficient cross-departmental and cross-field collaboration, enterprises urgently need an intelligent analysis and decision-making system based on employee information and project information, which can select qualified employees as project team members for each project team when forming a work team to ensure that the project can be completed within the expected time. Summary of the invention

[0004] The present invention provides a human resource information processing method based on RPA robots. By collecting employee and project information at different project stages and constructing graph structure data, a neural network model is used to match suitable employees into project teams according to the needs of different project stages, and the employee composition in the project team is further optimized according to the overall progress of the project and the degree of harmony among employees, thereby solving the problems of project delays and suspensions caused by non-technical factors for employees drawn from general projects.

[0005] The present invention provides a human resource information processing method based on an RPA robot, comprising the following steps:

[0006] Step 100, collecting graph structure data at the beginning of each stage of the project, the graph structure data including nodes and edges connecting the nodes, a node represents an employee or a project, an edge represents the relationship between the employee and the project;

[0007] There are edges between nodes corresponding to employees working on the same project;

[0008] Two nodes, one node represents an employee, and the other node represents a project. If the employee belongs to the project, there is an edge between the two nodes, otherwise there is no edge between the two nodes;

[0009] The node characteristics of the node representing an employee are equal to the employee characteristics, and the node characteristics of the node representing a project are equal to the project characteristics; the employee characteristics are obtained by encoding the collected background information of the employee, and the project characteristics are obtained by encoding the collected project information of the project;

[0010] Step 200, encoding the collected graph structure data into sequence data, the sequence data includes n sequence units, and the t-th sequence unit represents the graph structure data collected in the t-th stage;

[0011] Step 300, input the sequence data into the process model, the process model includes a first module, a second module and an output module, each sequence unit is input into the first module in turn, the first module outputs the first hidden state of the sequence unit; the second module inputs the first hidden state of all sequence units and outputs the second hidden state; the output module inputs the second hidden state and outputs the employee assigned to the project.

[0012] Furthermore, the background information of the employees and the project information of the projects are both text information;

[0013] The encoding method of text information is to convert the text information into a word sequence, then one-hot encode the word sequence and input it into the word embedding model to obtain the embedding representation of the words in the word sequence, and concatenate the embedding representations of all words to obtain the encoding result of the text information.

[0014] Furthermore, the calculation formula of the first module is as follows:

[0015] ;

[0016] in represents the hidden state of the vth node, represents the weight matrix, represents the bias term, is the set of nodes that have edges with node v, represents the node feature of the u-th node, is a normalizing constant, It is a S-shaped function;

[0017] A node feature represents an employee feature corresponding to an employee node or a project feature corresponding to a project node;

[0018] ;

[0019] in represents the first hidden state of the output, represents the hidden state of the vth node, It means concatenating the hidden states of all nodes in the graph structure data, N represents the set of all nodes in the graph structure data, is the weight parameter, is the bias parameter, Represents a sigmoid function.

[0020] Furthermore, the calculation formula of the second module is as follows:

[0021] ;

[0022] in represents the t-th second hidden state, represents the t-1th second hidden state, represents the first hidden state output when the t-th sequence unit enters the first module, and is the weight parameter, is the bias parameter and tanh is the hyperbolic tangent function.

[0023] Furthermore, the calculation formula of the output module is as follows:

[0024] ;

[0025] in The output is the employee vector assigned to the project. The i-th component of the employee vector assigned to the project represents the probability value of the i-th action. The set of all actions is called the action space, which contains all possible actions that may be selected. represents the nth second hidden state, is the weight parameter, is the bias parameter, Represents a sigmoid function.

[0026] Furthermore, an action is represented as a vector, and the j-th component of the vector represents the j-th employee among the employees assigned to the project.

[0027] Furthermore, the steps of training the process model include:

[0028] Step 101, building a target network, and replicating a process model of the target network;

[0029] Step 102: randomly obtain an experience from the experience pool, the experience includes the graph structure data of the z stage , Actions performed in the z phase , graph structure data at stage z+1 , Execute actions Rewards ;

[0030] Graph structure data at stages z and z+1 and They represent the sequence data of the zth and z+1th stages of the process model input respectively;

[0031] Actions performed in z phase express The output actions when inputting the process model;

[0032] Then calculate the graph structure data The output of the target network corresponds to The probability value ;

[0033] ;

[0034] Representing graph structured data The maximum probability value of the target network output;

[0035] Step 103, according to and Calculate the loss value and update the process model;

[0036] ;

[0037] Representing graph structured data The output of the following process model corresponds to The probability value of

[0038] is the discount factor, a value between 0 and 1;

[0039] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current process model;

[0040] Step 105, until the number of collected experiences reaches a set threshold or the process model converges, the step is terminated.

[0041] Furthermore, rewards The calculation formula is as follows:

[0042] ;

[0043] in Represents the actual workload of the project from stage z to z+1, Represents the expected workload of the project from phases z to z+1.

[0044] Furthermore, the human resources information processing system based on RPA robots includes:

[0045] Data collection module, used to collect graph structure data at the beginning of each stage of the project;

[0046] A data encoding module, used to encode the collected graph structure data into sequence data;

[0047] The data processing module inputs the graph structure data corresponding to each sequence unit in the sequence data into the process model to obtain the value of the selected action;

[0048] The action decoding module obtains the employee assigned to the project based on the selected action decoding.

[0049] The beneficial effects of the present invention are:

[0050] 1. Combining employees and projects into graph structure data can better express the complex relationship between employees and projects, and more richly depict the different ways of association between employees and projects in the graph structure, which not only enriches the semantics but also has good scalability;

[0051] 2. Encoding graph structure data at different stages into sequence data can better use the network model to capture local and global contextual information, which is crucial for accurately understanding and representing complex graph structures. When outputting, all information in the input sequence will be comprehensively considered to predict an output sequence that is more in line with the contextual semantics, making the output content more accurate and meaningful. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of a human resources information processing method based on an RPA robot of the present invention. DETAILED DESCRIPTION

[0053] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0054] In one embodiment of the present invention, a human resources information processing method based on an RPA robot is provided. Figure 1 As shown, including:

[0055] Step 100, collecting graph structure data at the beginning of each stage of the project, the graph structure data including nodes and edges connecting the nodes, a node represents an employee or a project, an edge represents the relationship between the employee and the project;

[0056] There are edges between nodes corresponding to employees working on the same project;

[0057] Two nodes, one node represents an employee, and the other node represents a project. If the employee belongs to the project, there is an edge between the two nodes, otherwise there is no edge between the two nodes;

[0058] The node characteristics of the node representing an employee are equal to the employee characteristics, and the node characteristics of the node representing a project are equal to the project characteristics; the employee characteristics are obtained by encoding the collected background information of the employee, and the project characteristics are obtained by encoding the collected project information of the project;

[0059] In one embodiment of the present invention, an employee feature includes:

[0060] Employee ID, skills, education, work experience, communication skills, teamwork skills, innovative thinking skills, professional certificates, training experience, current salary, and job level;

[0061] In one embodiment of the present invention, a project feature includes:

[0062] Project ID, project type, project field, project scale, project complexity, skill requirements, project budget, and project timeline;

[0063] In one embodiment of the present invention, the background information of the employee and the project information of the project are both text information;

[0064] The encoding method of text information is to convert the text information into a word sequence, then one-hot encode the word sequence and input it into the word embedding model to obtain the embedding representation of the words in the word sequence, and concatenate the embedding representations of all words to obtain the encoding result of the text information.

[0065] Of course, other text encoding methods can also be used.

[0066] Step 200, encoding the collected graph structure data into sequence data, the sequence data includes n sequence units, and the t-th sequence unit represents the graph structure data collected in the t-th stage;

[0067] Step 300, inputting sequence data into a process model, the process model includes a first module, a second module and an output module, inputting each sequence unit into the first module in sequence, the first module outputting the first hidden state of the sequence unit; the second module inputting the first hidden state of all sequence units and outputting the second hidden state; the output module inputting the second hidden state and outputting the employee assigned to the project;

[0068] In one embodiment of the present invention, the calculation formula of the first module is as follows:

[0069] ;

[0070] in represents the hidden state of the vth node, represents the weight matrix, represents the bias term, is the set of nodes that have edges with node v, represents the node feature of the u-th node, is a normalizing constant, It is a S-shaped function;

[0071] A node feature represents an employee feature corresponding to an employee node or a project feature corresponding to a project node;

[0072] ;

[0073] in represents the first hidden state of the output, represents the hidden state of the vth node, It means concatenating the hidden states of all nodes in the graph structure data, N represents the set of all nodes in the graph structure data, is the weight parameter, is the bias parameter, Represents a sigmoid function.

[0074] In one embodiment of the present invention, the calculation formula of the second module is as follows:

[0075] ;

[0076] in represents the t-th second hidden state, represents the t-1th second hidden state, represents the first hidden state output when the t-th sequence unit enters the first module, and is the weight parameter, is the bias parameter and tanh is the hyperbolic tangent function.

[0077] In one embodiment of the present invention, the calculation formula of the output module is as follows:

[0078] ;

[0079] in The output is the employee vector assigned to the project. The i-th component of the employee vector assigned to the project represents the probability value of the i-th action. The set of all actions is called the action space, which contains all possible actions that may be selected. represents the nth second hidden state, is the weight parameter, is the bias parameter, Represents a sigmoid function.

[0080] In one embodiment of the present invention, an action is represented by a vector, and the j-th component of the vector represents the j-th employee among the employees assigned to the project.

[0081] In one embodiment of the present invention, the step of training the process model includes:

[0082] Step 101, building a target network, and replicating a process model of the target network;

[0083] Step 102: randomly obtain an experience from the experience pool, the experience includes the graph structure data of the z stage , Actions performed in the z phase , graph structure data at stage z+1 , Execute actions Rewards ;

[0084] Graph structure data at stages z and z+1 and They represent the sequence data of the zth and z+1th stages of the process model input respectively;

[0085] Actions performed in z phase express The output actions when inputting the process model;

[0086] Then calculate the graph structure data The target network output corresponds to The probability value ;

[0087] ;

[0088] Representing graph structured data The maximum probability value of the target network output;

[0089] Step 103, according to and Calculate the loss value and update the process model;

[0090] ;

[0091] Representing graph structured data The output of the following process model corresponds to The probability value of

[0092] is the discount factor, a value between 0 and 1;

[0093] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current process model;

[0094] Step 105, until the number of collected experiences reaches a set threshold or the process model converges, the step is terminated.

[0095] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0096] In one embodiment of the present invention, the reward The calculation formula is as follows:

[0097] ;

[0098] in Represents the actual workload of the project from stage z to z+1, Represents the expected workload of the project from phases z to z+1.

[0099] In one embodiment of the present invention, the first module performs pre-training, and during the pre-training, the pre-training output layer is connected, the pre-training output layer inputs the node features of two employees in the project, and outputs a result indicating whether the two employees will have a work conflict.

[0100] Work-related conflicts refer to the number of times two employees have disputes during their work that require a third party to mediate more than three times.

[0101] The calculation formula of the pre-trained output layer is as follows:

[0102] ;

[0103] in, It represents the probability that there is a conflict between employee i and employee j. If it is greater than the set threshold, it means there is a conflict, otherwise there is no conflict. D represents the set of employee nodes in the graph structure data. represents the concatenation of the hidden state of the i-th employee node and the hidden state of the j-th employee node, is the weight parameter, is the bias parameter, represents the S-type function, Represents the transpose of a matrix.

[0104] By pre-training the first module in this way, the first module can identify the degree of harmony of work cooperation between employees, reduce the probability of work conflicts between employees assigned to the same project in the process model, and ensure project progress.

[0105] In one embodiment of the present invention, a human resources information processing system based on an RPA robot includes:

[0106] Data collection module, used to collect graph structure data at the beginning of each stage of the project;

[0107] A data encoding module, used to encode the collected graph structure data into sequence data;

[0108] The data processing module inputs the graph structure data corresponding to each sequence unit in the sequence data into the process model to obtain the value of the selected action;

[0109] The action decoding module obtains the employee assigned to the project based on the selected action decoding.

[0110] In one embodiment of the present invention, a storage medium is provided, which stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps in the aforementioned human resources information processing method based on an RPA robot can be executed.

[0111] After the first module is pre-trained, the process model is added and the parameters are updated during the process model training.

[0112] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A human resources information processing method based on RPA robots, It is characterized in that The following steps are involved: Step 100, collecting graph structure data at the beginning of each stage of the project, the graph structure data including nodes and edges connecting the nodes, a node represents an employee or a project, an edge represents the relationship between the employee and the project; There are edges between nodes corresponding to employees working on the same project; Two nodes, one node represents an employee, and the other node represents a project. If the employee belongs to the project, there is an edge between the two nodes, otherwise there is no edge between the two nodes; The node characteristics of the node representing an employee are equal to the employee characteristics, and the node characteristics of the node representing a project are equal to the project characteristics; Employee characteristics are obtained by coding the collected background information of employees, and project characteristics are obtained by coding the collected project information of projects; Step 200, encoding the collected graph structure data into sequence data, the sequence data includes n sequence units, and the t-th sequence unit represents the graph structure data collected in the t-th stage; Step 300, input the sequence data into the process model, the process model includes a first module, a second module and an output module, each sequence unit is input into the first module in turn, the first module outputs the first hidden state of the sequence unit; the second module inputs the first hidden state of all sequence units and outputs the second hidden state; the output module inputs the second hidden state and outputs the employee assigned to the project.

2. According to claim 1, a method for processing human resources information based on an RPA robot, It is characterized in that The background information of employees and the project information of projects are all text information; The encoding method of text information is to convert the text information into a word sequence, then one-hot encode the word sequence and input it into the word embedding model to obtain the embedding representation of the words in the word sequence, and concatenate the embedding representations of all words to obtain the encoding result of the text information.

3. According to claim 1, a method for processing human resources information based on an RPA robot, It is characterized in that The calculation formula for the first module is as follows: ； in represents the hidden state of the vth node, represents the weight matrix, represents the bias term, is the set of nodes that have edges with node v, represents the node feature of the u-th node, is a normalizing constant, It is a S-shaped function; A node feature represents an employee feature corresponding to an employee node or a project feature corresponding to a project node; ； in represents the first hidden state of the output, represents the hidden state of the vth node, It means concatenating the hidden states of all nodes in the graph structure data, N represents the set of all nodes in the graph structure data, is the weight parameter, is the bias parameter, Represents a sigmoid function.

4. According to claim 3, a method for processing human resources information based on an RPA robot, It is characterized in that The calculation formula of the second module is as follows: ； in represents the t-th second hidden state, represents the t-1th second hidden state, represents the first hidden state output when the t-th sequence unit enters the first module, and is the weight parameter, is the bias parameter and tanh is the hyperbolic tangent function.

5. According to claim 4, a method for processing human resources information based on an RPA robot, It is characterized in that The calculation formula for the output module is as follows: ； in The output is the employee vector assigned to the project. The i-th component of the employee vector assigned to the project represents the probability value of the i-th action. The set of all actions is called the action space, which contains all possible actions that may be selected. represents the nth second hidden state, is the weight parameter, is the bias parameter, Represents a sigmoid function.

6. According to claim 5, a method for processing human resources information based on an RPA robot, It is characterized in that An action is represented as a vector, and the j-th component of the vector represents the j-th employee among the employees assigned to the project.

7. According to claim 1, a method for processing human resources information based on an RPA robot, It is characterized in that The steps of process model training include: Step 101, building a target network, and replicating a process model of the target network; Step 102: randomly obtain an experience from the experience pool, the experience includes the graph structure data of the z stage , Actions performed in the z phase , graph structure data at stage z+1 , Execute actions Rewards ; Graph structure data at stages z and z+1 and They represent the sequence data of the zth and z+1th stages of the process model input respectively; Actions performed in z phase express The output actions when inputting the process model; Then calculate the graph structure data The target network output corresponds to The probability value ; ； Representing graph structured data The maximum probability value of the target network output; Step 103, according to and Calculate the loss value and update the process model; ； Representing graph structured data The output of the following process model corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current process model; Step 105, until the number of collected experiences reaches a set threshold or the process model converges, the step is terminated.

8. According to claim 7, a method for processing human resources information based on an RPA robot, It is characterized in that award The calculation formula is as follows: ； in Represents the actual workload of the project from stage z to z+1, Represents the expected workload of the project from phases z to z+1.

9. A human resources information processing system based on RPA robots, It is characterized in that It is used to execute a human resource information processing method based on an RPA robot as described in any one of claims 1 to 8. The human resource information processing system based on the RPA robot includes: Data collection module, used to collect graph structure data at the beginning of each stage of the project; A data encoding module, used to encode the collected graph structure data into sequence data; The data processing module inputs the graph structure data corresponding to each sequence unit in the sequence data into the process model to obtain the value of the selected action; The action decoding module obtains the employee assigned to the project based on the selected action decoding.

10. A storage medium storing non-transitory computer-readable instructions, It is characterized in that When the non-transitory computer-readable instructions are executed by a computer, the steps in the human resources information processing method based on an RPA robot as described in any one of claims 1 to 8 can be executed.

Images