Intelligent process task assignment method and system based on combination of activiti and machine learning

Abstract

The application provides a kind of intelligent process task allocation method and system based on Activiti and machine learning combination, it is related to machine learning technical field, the method steps include collecting multi-dimensional roadshow meeting process data and carrying out data cleaning;The multi-dimensional roadshow meeting process data is processed and calculated, and the roadshow meeting type feature, task urgency feature and compliance approval personnel task load feature are extracted;Task allocation model is constructed, and the task allocation model is trained based on the extracted features, the model is used to predict the task approval time of different approval personnel;The trained task allocation model is integrated into the Activiti process engine, and when the process task reaches the approval node, the model is called to calculate the predicted approval time of the candidate approval personnel;According to the predicted approval time, dynamically allocate tasks, select the compliance approval personnel with the shortest predicted approval time to perform the current approval task;Periodically collect new approval data, retrain the task allocation model to iteratively optimize the prediction effect.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to an intelligent process task allocation method and system based on the combination of Activiti and machine learning. Background Technology

[0002] With increasing informatization, more and more enterprises need to digitize various business approval processes online. Activiti, an open-source workflow and business process management engine based on Java, follows the Apache 2.0 open-source license and conforms to the BPMN 2.0 standard. It is favored for its lightweight nature and ease of integration with mainstream frameworks such as Spring, facilitating the visual design, automated execution, and monitoring of business processes.

[0003] However, the native Activiti engine primarily relies on fixed allocation when distributing approval tasks. This means that administrators assign tasks to specific approvers based on experience and judgment, or according to simple rules. This allocation method is inefficient, making it difficult to quickly respond to a large number of complex approval tasks. Secondly, it results in uneven distribution, failing to effectively assess the actual workload and processing capacity of approvers, leading to underutilization of resources. Finally, it lacks foresight, making it difficult to analyze and predict based on historical data, thus hindering the optimization of task allocation in the approval process and ultimately resulting in low approval efficiency and quality. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent process task allocation method and system based on the combination of Activiti and machine learning, in order to solve the problems mentioned in the background art, that the native Activiti engine mainly relies on fixed allocation or allocation according to simple rules when allocating approval tasks, which is not only inefficient, but also has uneven allocation and lack of predictability.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent process task allocation method based on Activiti and machine learning, comprising the following steps: collecting multi-dimensional roadshow meeting process data and performing data cleaning; processing and calculating the multi-dimensional roadshow meeting process data, and extracting roadshow meeting type features, task urgency features, and compliance approval personnel task load features; constructing a task allocation model and training the task allocation model based on the extracted features, wherein the model is used to predict the task approval time of different approvers; integrating the trained task allocation model into the Activiti process engine, and when a process task reaches an approval node, calling the model to calculate the predicted approval time of candidate approvers; dynamically allocating tasks according to the predicted approval time, selecting the compliance approver with the shortest predicted approval time to execute the current approval task; periodically collecting new approval data and retraining the task allocation model to iteratively optimize the prediction effect.

[0006] Optionally, the preprocessing steps specifically include: during the integration of multi-dimensional roadshow conference process data, recording the data source and processing logic, desensitizing sensitive information, retaining business-related features, and removing missing data values; using one-hot encoding to convert categorical data into unique binary vectors, and normalizing numerical data to make it fall within the [0,1] interval.

[0007] Optionally, the multi-dimensional roadshow meeting process data includes: roadshow meeting type, task submission time, meeting start time, approver information, professional field matching degree, online status, approval duration, and the number of tasks pending approval by the approver; the roadshow meeting type is used to determine the task workload based on different types; the task submission time and the meeting start time are used to calculate the urgency of the task; the professional field matching degree is used to assess the suitability of the approver and the task type; the online status is used to determine the real-time availability of the approver, and the approval duration and the number of tasks pending approval by the approver are used to calculate the approver's workload.

[0008] Optionally, the step of determining whether the approver is compliant by combining the professional field matching degree and the online status specifically includes: the professional field type corresponds one-to-one with the meeting type; determining whether the professional field type matches the meeting type, if yes, it is recorded as 1, otherwise as 0; determining whether the approver is online, if online, it is recorded as 1, otherwise as 0; if the professional field type matches the meeting type and the approver is online, then the approver is determined to be compliant.

[0009] Optionally, the calculation steps for the task urgency feature specifically include: obtaining the time difference between the task submission time and the meeting start time; using a time decay function to convert the time difference into an urgency value, where the smaller the time difference, the higher the urgency value; and introducing random noise conforming to a normal distribution during the calculation process to simulate the uncertainty in actual approval scenarios.

[0010] Optionally, the calculation method for the task load characteristics of compliance approvers includes: counting the number of pending tasks for the current approver; calculating the urgency of each pending task and sorting them from high to low urgency; and calculating a normalized task load value based on the current task's position in the sort and the historical maximum number of tasks.

[0011] Optionally, the steps of constructing the task allocation model specifically include: constructing a linear regression model based on task type characteristics, task urgency characteristics, and task load characteristics of compliance approvers; performing a logarithmic transformation on the approval time to ensure that the predicted time is positive, and introducing cross-term features to optimize the prediction; and solving the model parameters using the least squares method to obtain the optimal parameters.

[0012] Optionally, the step of integrating the trained task allocation model into the Activiti process engine specifically includes: configuring a custom listener in the BPMN file, and triggering the listener when the process task reaches the approval node.

[0013] Optionally, the steps of the custom listener specifically include: inheriting the TaskListener interface of the Activiti engine and overriding the notify method; obtaining the context information of the current task in the notify method; calling the API of the task allocation model, inputting task type characteristics, task urgency characteristics, and compliance approver task load characteristics, obtaining the predicted approval time of candidate approvers, and determining the optimal approver; dynamically setting the approver variable through the runtime service interface of the Activiti engine, and updating the task allocation through the RuntimeService interface.

[0014] On the other hand, the present invention also provides an intelligent process task allocation system based on Activiti and machine learning, comprising: a data collection module for collecting multi-dimensional roadshow meeting process data and performing data cleaning; a feature extraction module for processing and calculating the multi-dimensional roadshow meeting process data and extracting roadshow meeting type features, task urgency features, and compliance approval personnel task load features; a model building module for building a task allocation model and training the task allocation model based on the extracted features, wherein the model is used to predict the task approval time of different approvers; an integration module for integrating the trained task allocation model into the Activiti process engine, and calling the model to calculate the predicted approval time of candidate approvers when a process task reaches an approval node; an allocation module for dynamically allocating tasks according to the predicted approval time and selecting the compliance approval personnel with the shortest predicted approval time to execute the current approval task; and an optimization module for periodically collecting new approval data and retraining the task allocation model to iteratively optimize the prediction effect.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] This application combines the Activiti workflow engine and machine learning models to dynamically allocate tasks, reducing approval waiting time and approval time, and significantly improving the efficiency of the approval process. Based on the approver's historical performance and current work status, tasks are intelligently allocated to achieve a balanced distribution of workload and avoid resource waste. Utilizing machine learning technology, the system can continuously learn and optimize allocation strategies to adapt to changes in approval needs, achieving intelligent process management. This intelligent task allocation method, combining the Activiti workflow engine and machine learning, not only improves the efficiency and quality of the approval process but also enhances the system's intelligence level through dynamic optimization of resource allocation, providing enterprises with a more efficient and accurate workflow task allocation solution. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0018] Figure 2 This is a schematic diagram of the system structure of the present invention.

[0019] In the diagram: 10 - Data collection module, 20 - Feature extraction module, 30 - Model building module, 40 - Integration module, 50 - Allocation module, 60 - Optimization module. Detailed Implementation

[0020] The present invention will now be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0023] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0024] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Please refer to Figure 1This invention discloses an intelligent process task allocation method based on Activiti and machine learning, comprising the following steps: collecting multi-dimensional roadshow meeting process data and performing data cleaning; processing and calculating the multi-dimensional roadshow meeting process data, and extracting roadshow meeting type features, task urgency features, and compliance approval personnel task load features; constructing a task allocation model and training the task allocation model based on the extracted features, wherein the model is used to predict the task approval time of different approvers; integrating the trained task allocation model into the Activiti process engine, and calling the model to calculate the predicted approval time of candidate approvers when a process task reaches an approval node; dynamically allocating tasks according to the predicted approval time, and selecting the compliance approver with the shortest predicted approval time to execute the current approval task; periodically collecting new approval data and retraining the task allocation model to iteratively optimize the prediction effect.

[0027] Specifically, by collecting multi-dimensional roadshow meeting process data, key features such as roadshow meeting type characteristics, task urgency characteristics, and compliance approver task load characteristics are extracted to construct a task allocation model for task allocation. This model is integrated into the Activiti process engine. When a process task reaches an approval node, the model is invoked to calculate the predicted approval time for candidate approvers. Tasks are dynamically allocated based on the predicted approval time, and the compliance approver with the shortest predicted approval time is selected to execute the current approval task. This mechanism of automatically selecting the optimal approver significantly shortens the approval process time, achieving dynamic and intelligent allocation of approval tasks and overcoming the shortcomings of traditional fixed allocation methods, such as low efficiency and uneven resource distribution. New approval data is collected periodically to retrain the task allocation model for iterative optimization of prediction results. This application combines the Activiti workflow engine and machine learning models to dynamically allocate tasks, reducing approval waiting time and approval time, and significantly improving the efficiency of the approval process. Based on the approver's historical performance and current work status, tasks are intelligently allocated to achieve a balanced distribution of workload and avoid resource waste. Utilizing machine learning technology, the system can continuously learn and optimize allocation strategies to adapt to changes in approval needs, achieving intelligent process management. By combining the Activiti process engine with machine learning for intelligent task allocation, not only is the efficiency and quality of the approval process improved, but the system's intelligence level is also enhanced through dynamic optimization of resource allocation, providing enterprises with a more efficient and accurate process task allocation solution.

[0028] In some embodiments, the preprocessing steps specifically include: during the integration of multi-dimensional roadshow conference process data, recording the data source and processing logic, desensitizing sensitive information, retaining business-related features, and removing missing data values; using one-hot encoding to convert categorical data into unique binary vectors, and normalizing numerical data to make it fall within the [0,1] interval.

[0029] Specifically, the collected multi-dimensional roadshow meeting process data undergoes standardized data cleaning. Data quality is improved through anonymization and missing value removal, providing reliable input for model training. A data traceability mechanism enhances the traceability of feature engineering and reduces the impact of data anomalies on allocation results. The collected multi-dimensional roadshow meeting process data is processed as follows: matching the professional field is recorded as 1, otherwise as 0; online approval is recorded as 1, otherwise as 0; approval by the approver is recorded as 1, otherwise as 0; roadshow meeting types include three categories: industry viewpoint presentation, company viewpoint presentation, and company performance exchange, denoted as: In the formula, Different types of roadshow meetings require different information for compliance approval personnel to confirm and check, resulting in varying time consumption. One-hot encoding is used to convert categorical data into unique binary vectors, and numerical data is normalized to fall within the [0,1] interval.

[0030] In some embodiments, the multi-dimensional roadshow meeting process data includes: roadshow meeting type, task submission time, meeting start time, approver information, professional field matching degree, online status, approval duration, and the number of tasks pending approval by the approver; the roadshow meeting type is used to determine the workload of tasks based on different types; the task submission time and the meeting start time are used to calculate the urgency of tasks; the professional field matching degree is used to assess the suitability of approvers with task types; the online status is used to determine the real-time availability of approvers; and the combination of the professional field matching degree and the online status is used to determine whether approvers are compliant; the approval duration and the number of tasks pending processing by the approver are used to calculate the approver's workload.

[0031] Specifically, a comprehensive task allocation indicator system is constructed by evaluating multiple dimensions of data, including the type of roadshow meeting, task submission time, meeting start time, approver information, professional field matching degree, online status, approval time, and the number of tasks pending approval by approvers. Through a joint judgment mechanism of professional field matching degree, online status, task urgency, and approver workload, the efficiency and quality of the approval process are improved while ensuring compliance.

[0032] In some embodiments, the step of determining whether the approver is compliant by combining the professional field matching degree and the online status specifically includes: the professional field type corresponds one-to-one with the meeting type; determining whether the professional field type matches the meeting type, if yes, it is recorded as 1, otherwise as 0; determining whether the approver is online, if online, it is recorded as 1, otherwise as 0; if the professional field type matches the meeting type and the approver is online, then the approver is determined to be compliant.

[0033] Specifically, a binary quantification standard is adopted to achieve efficient compliance screening. A candidate list is accurately generated through professional matching and online status verification, which effectively reduces the allocation of invalid tasks.

[0034] In some embodiments, the calculation steps for the task urgency feature specifically include: obtaining the time difference between the task submission time and the meeting start time; converting the time difference into an urgency value using a time decay function, wherein the smaller the time difference, the higher the urgency value; and introducing random noise conforming to a normal distribution into the calculation process to simulate the uncertainty in actual approval scenarios.

[0035] Specifically, a custom time decay function is used to convert the time interval into task urgency, and the calculation formula is as follows: In the formula, Depending on the urgency of the task, The start time of the roadshow meeting. For the submission time of the process task, For the standard overall distribution and This application uses an urgency quantification model based on a time decay function, combined with random noise to simulate real-world scenarios, to more accurately identify differences in task priorities and ensure that urgent tasks are handled in a timely manner.

[0036] In some embodiments, the calculation method for the task load characteristics of compliance approvers includes: counting the number of pending tasks for the current approver; calculating the urgency of each pending task and sorting them from high to low urgency; and calculating a normalized task load value based on the current task's position in the sort and the historical maximum number of tasks.

[0037] Specifically, after an approval task is submitted, the compliance approver's pending approval tasks are: Given n tasks, calculate the urgency of each task and sort them from highest to lowest urgency. The sorted results are: The formula for calculating the workload of compliance approval personnel is: In the formula, As a workload for compliance approval personnel This represents the position number after sorting tasks by urgency from highest to lowest. This represents the maximum historical number of pending tasks (n) for all compliance approvers. This application employs a load assessment method involving dynamic sorting and normalization to reflect the workload of approvers in real time, achieving intelligent and balanced allocation of tasks.

[0038] In some embodiments, the steps of constructing the task allocation model specifically include: constructing a linear regression model based on task type characteristics, task urgency characteristics, and task load characteristics of compliance approvers; performing a logarithmic transformation on the approval time to ensure that the predicted time is positive, and introducing cross-term features to optimize the prediction; and solving the model parameters using the least squares method to obtain the optimal parameters.

[0039] Specifically, the intelligent task allocation model is constructed, and its calculation formula is as follows: In the formula, y represents the approval time, and logarithmic variation is used to ensure that the predicted time is a positive value. For constant terms, For roadshow conference type The corresponding matrix coefficients are in the form of , Based on the urgency of the mission The corresponding coefficients, Task load for compliance approval personnel Corresponding coefficients. Because the urgency of the task affects the workload of compliance approval personnel, cross-terms are introduced. Its corresponding coefficient is , It is a standard normal distribution and .

[0040] First, construct the coefficient vector. Construct the feature matrix after data processing Where n is the sample data size; the target vector is The least squares method is used to solve the intelligent task allocation model. The formula for calculating the sum of squared residuals function using the least squares method is as follows: In the formula, Let be the residual sum of squares function of least squares method. For the target vector, The feature matrix after data processing. Let be the coefficient vector, in order to find such that Minimum coefficient vector ,calculate The derivative of is calculated using the following formula: In the formula, As a partial differential operator, setting the derivative to 0, we obtain the normal equation: From this, we can obtain the solution. Considering the large amount of data, the Mumpy library was used to construct the intelligent task allocation model. This application employs a log-linear model combined with cross-term features to improve prediction accuracy while ensuring computational efficiency. Least squares parameter estimation enhances model stability and meets the real-time response requirements of the workflow engine.

[0041] In some embodiments, the step of integrating the trained task allocation model into the Activiti process engine specifically includes: configuring a custom listener in the BPMN file, and triggering the listener when the process task reaches the approval node.

[0042] Specifically, when the roadshow meeting process reaches the compliance personnel approval node, this explanation takes node_a of process flow_a as an example;

[0043] Add the following configuration to the .bpmn file of the node_a node in the process flow_a:

[0044] <extensionelements>

[0045] <activiti:taskListener

[0046] event="start"class="listener.MyTaskCreateListener" / >

[0047] < / extensionelements>

[0048] When process flow_a reaches node_a, the custom listener MyTaskCreateListener is triggered. This application achieves non-intrusive system integration through standard BPMN extensions, and the listener mechanism seamlessly adds intelligent allocation capabilities while maintaining the original functionality of Activiti.

[0049] In some embodiments, the steps of the custom listener specifically include: inheriting the TaskListener interface of the Activiti engine and overriding the notify method; obtaining the context information of the current task in the notify method; calling the API of the task allocation model, inputting task type characteristics, task urgency characteristics, and compliance approver task load characteristics, obtaining the predicted approval time of candidate approvers, and determining the optimal approver; dynamically setting the approver variable through the runtime service interface of the Activiti engine, and updating the task allocation through the RuntimeService interface.

[0050] Specifically, a custom process listener class MyTaskCreateListener is defined for the process flow_a. This class implements the TaskListener interface of the acitiviti process engine and overrides the notify method. When the approval task reaches node_a, the notify method in the MyTaskCreateListener listener will be triggered.

[0051] The logic in the notify method is as follows: call the trained machine learning model API, and input the current roadshow meeting type. The urgency level of the task is Compliance approval personnel's workload The system obtains predicted approval times for multiple candidates. It selects the approver with the shortest approval time who can approve the task and is chosen as the current node approver. The `setVariable` method of the `RuntimeService` class in the Activiti process engine is called to dynamically assign this optimal approver to the current task. After assignment, the approver receives a notification of the pending task and begins processing it. This application utilizes dynamic variable setting technology based on the runtime interface to achieve seamless real-time switching of approvers, significantly improving task reassignment efficiency in abnormal situations.

[0052] On the other hand, the present invention also provides an intelligent process task allocation system based on Activiti and machine learning, comprising: a data collection module 10 for collecting multi-dimensional roadshow meeting process data and performing data cleaning; a feature extraction module 20 for processing and calculating the multi-dimensional roadshow meeting process data and extracting roadshow meeting type features, task urgency features, and compliance approval personnel task load features; a model building module 30 for building a task allocation model and training the task allocation model based on the extracted features, wherein the model is used to predict the task approval time of different approvers; an integration module 40 for integrating the trained task allocation model into the Activiti process engine, and calling the model to calculate the predicted approval time of candidate approvers when a process task reaches an approval node; an allocation module 50 for dynamically allocating tasks according to the predicted approval time and selecting the compliance approval personnel with the shortest predicted approval time to execute the current approval task; and an optimization module 60 for periodically collecting new approval data and retraining the task allocation model to iteratively optimize the prediction effect.

[0053] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0054] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0055] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention's specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for intelligent process task allocation based on Activiti and machine learning, characterized by the following steps: include: Collect and preprocess multi-dimensional roadshow meeting process data; The multi-dimensional roadshow meeting process data includes: Roadshow meeting type, task submission time, meeting start time, approver information, professional field matching degree, online status, approval time and number of tasks pending approval by approvers; The roadshow meeting type is used to determine the workload based on different types; The task submission time and the meeting start time are used to calculate the task urgency. The professional domain matching degree is used to assess the suitability of the approver and the task type, the online status is used to determine the real-time availability of the approver, and the approval degree and the online status are combined to determine whether the approver is compliant. The approval time and the number of tasks pending processing by the approver are used to calculate the approver's workload. The multi-dimensional roadshow meeting process data is processed and calculated, and roadshow meeting type characteristics, task urgency characteristics, and compliance approval personnel task load characteristics are extracted; A task allocation model is constructed and trained based on the extracted features. The model is used to predict the task approval time for different approvers. The steps for constructing the task allocation model specifically include: A linear regression model was constructed based on task type characteristics, task urgency characteristics, and task load characteristics of compliance approval personnel. Logarithmic transformation is applied to the approval time to ensure that the predicted time is positive, and cross-term features are introduced to optimize the prediction. The model parameters are solved using the least squares method to obtain the optimal parameters; The trained task allocation model is integrated into the Activiti process engine. When a process task reaches an approval node, the model is called to calculate the predicted approval time for candidate approvers. The steps for integrating the trained task allocation model into the Activiti process engine specifically include: Configure a custom listener in the BPMN file to trigger the listener when the workflow task reaches the approval node; The steps for creating a custom listener specifically include: Inherit the TaskListener interface of the Activiti engine and override the notify method; Retrieve the context information of the current task in the notify method; Call the API of the task allocation model, input task type characteristics, task urgency characteristics, and compliance approver task load characteristics, obtain the predicted approval time of candidate approvers, and determine the optimal approver; The approver variable is dynamically set through the runtime service interface of the Activiti engine, and the task assignment is updated through the RuntimeService interface; Tasks are dynamically allocated based on the predicted approval time, and the compliance approval personnel with the shortest predicted approval time are selected to perform the current approval task. Regularly collect newly approved data and retrain the task allocation model to iteratively optimize prediction results.

2. The intelligent process task allocation method according to claim 1, characterized in that, The preprocessing steps specifically include: During the integration of multi-dimensional roadshow meeting process data, the data source and processing logic were recorded, sensitive information was anonymized, business-related features were retained, and missing data values ​​were removed. One-hot encoding is used to transform categorical data into unique binary vectors, and numerical data is normalized to fall within the [0,1] interval.

3. The intelligent process task allocation method according to claim 1, characterized in that, The steps for determining whether an approver is compliant by combining the professional field matching degree and the online status specifically include: Determine whether the approver's professional field matches the roadshow meeting type. If the approver's professional field includes the roadshow meeting type, the professional field match is recorded as 1; otherwise, it is recorded as 0. Determine if the approver is online; if online, record it as 1; otherwise, record it as 0. If the type of professional field matches the type of meeting and the approver is online, the approver is deemed to be compliant.

4. The intelligent process task allocation method according to claim 1, characterized in that, The calculation steps for the task urgency characteristic specifically include: Get the time difference between the task submission time and the meeting start time; The time difference is converted into an urgency value using a time decay function; the smaller the time difference, the higher the urgency value. Random noise conforming to a normal distribution is introduced into the calculation process to simulate the uncertainty in actual approval scenarios.

5. The intelligent process task allocation method according to claim 4, characterized in that, The calculation method for the task load characteristics of compliance approval personnel includes: Count the number of pending tasks for the current approvers; Calculate the urgency level of each pending task and sort them from highest to lowest urgency. Calculate the normalized task load value based on the current task's position in the sort and the historical maximum number of tasks.

6. An intelligent process task allocation system based on Activiti and machine learning, characterized in that, include: The data collection module is used to collect multi-dimensional roadshow meeting process data and perform data cleaning. The multi-dimensional roadshow meeting process data includes: Roadshow meeting type, task submission time, meeting start time, approver information, professional field matching degree, online status, approval time and number of tasks pending approval by approvers; The roadshow meeting type is used to determine the workload based on different types; The task submission time and the meeting start time are used to calculate the task urgency. The professional domain matching degree is used to assess the suitability of the approver and the task type, the online status is used to determine the real-time availability of the approver, and the approval degree and the online status are combined to determine whether the approver is compliant. The approval time and the number of tasks pending processing by the approver are used to calculate the approver's workload. The feature extraction module is used to process and calculate the multi-dimensional roadshow meeting process data, and extract roadshow meeting type features, task urgency features, and compliance approval personnel task load features; The model building module is used to build a task allocation model and train the task allocation model based on the extracted features. The model is used to predict the task approval time for different approvers. The steps for constructing the task allocation model specifically include: A linear regression model was constructed based on task type characteristics, task urgency characteristics, and task load characteristics of compliance approval personnel. Logarithmic transformation is applied to the approval time to ensure that the predicted time is positive, and cross-term features are introduced to optimize the prediction. The model parameters are solved using the least squares method to obtain the optimal parameters; The integration module is used to integrate the trained task allocation model into the Activiti process engine. When a process task reaches the approval node, the model is called to calculate the predicted approval time for candidate approvers. The steps for integrating the trained task allocation model into the Activiti process engine specifically include: Configure a custom listener in the BPMN file to trigger the listener when the workflow task reaches the approval node; The steps for creating a custom listener specifically include: Inherit the TaskListener interface of the Activiti engine and override the notify method; Retrieve the context information of the current task in the notify method; Call the API of the task allocation model, input task type characteristics, task urgency characteristics, and compliance approver task load characteristics, obtain the predicted approval time of candidate approvers, and determine the optimal approver; The approver variable is dynamically set through the runtime service interface of the Activiti engine, and the task assignment is updated through the RuntimeService interface; The allocation module is used to dynamically allocate tasks based on the predicted approval time, and select the compliance approver with the shortest predicted approval time to execute the current approval task; The optimization module is used to periodically collect newly added approval data and retrain the task allocation model to iteratively optimize the prediction effect.

Images