CN118502918A - Workflow intelligent management system based on artificial intelligence - Google Patents

Abstract

The present application provides an intelligent workflow management system based on artificial intelligence. The task management module is used to monitor the real-time workload of the current online workflow, and adaptively adjust the resource elastic allocation strategy of the target task according to the real-time workload; the decision support module is used to cluster the target task through the decision support model to obtain the clustering result; the priority execution order of each group of target tasks is determined through the decision support model based on the clustering result and the correlation; the resource management module is used to perform task allocation and resource scheduling for each group of target tasks according to the priority execution order and the resource elastic allocation strategy; the adaptive optimization learning module is used to dynamically optimize the hyperparameter combination in the decision support model with the quality evaluation value and/or performance index as the reward item to obtain the best hyperparameter combination. The system can efficiently manage and schedule workflow tasks, and provide more comprehensive and intelligent workflow intelligent management services.

Description

Intelligent workflow management system based on artificial intelligence

Technical Field

The present application belongs to the field of data processing technology, and in particular, relates to an intelligent workflow management system based on artificial intelligence.

Background Art

In traditional workflow management systems, administrators usually need to manually adjust resource allocation strategies, such as allocating computing resources to different tasks or nodes, to meet the needs of each task in the workflow.

However, this manual adjustment method has the following problems: The real-time workload changes of the workflow may not be detected in time. In large-scale workflows, the number and type of tasks may change dynamically, resulting in an unbalanced workload. If the administrator cannot adjust the resource allocation strategy in time, some tasks may not get enough resources to be processed, while other tasks may get too many resources, resulting in a waste of resources and reduced efficiency.

Therefore, it is urgent to design a new solution to solve at least one of the above technical problems.

Summary of the invention

The present application provides an artificial intelligence-based workflow intelligent management system for efficiently managing and scheduling workflow tasks and providing more comprehensive and intelligent workflow intelligent management services.

In a first aspect, the present application provides an artificial intelligence-based workflow intelligent management system, the system comprising:

The task management module is used to monitor the real-time workload of the current online workflow; according to the real-time workload, the resource elastic allocation strategy of the target task is adaptively adjusted to achieve elastic allocation of computing power;

A decision support module is used to perform cluster analysis on target tasks through a decision support model to obtain clustering results, wherein the clustering results include each group of target tasks and the correlation between each group of target tasks; based on the clustering results and the correlation, the priority execution order of each group of target tasks is determined through the decision support model; according to the clustering results, the target tasks with abnormalities are marked as abnormal tasks through the decision support model; wherein the abnormal tasks do not conform to the preset expected mode, or the correlation between the abnormal tasks and other target tasks is lower than the preset abnormal value;

The resource management module is used to perform task allocation and resource scheduling for each group of target tasks according to the priority execution order of each group of target tasks and the resource elasticity allocation strategy; and to perform resource reallocation or isolation operations on abnormal tasks;

An adaptive optimization learning module is used to dynamically optimize the hyperparameter combination in the decision support model by optimizing the reward model with the quality evaluation value and/or performance index corresponding to the clustering result as the reward item to obtain the best hyperparameter combination; according to the quality evaluation value and/or performance index corresponding to the clustering result, a cluster merging operation is performed on the decision support model to improve the robustness of the clustering result;

The report display module is used to respond to user operations, provide interactive analysis of the target task, and display the workflow information of the target task obtained through analysis to the user. The workflow information includes task execution status, task success rate, task exception information, and resource allocation status.

In a second aspect, an embodiment of the present application provides an artificial intelligence-based workflow intelligent management method, the method comprising:

Monitor the real-time workload of current online workflows;

Adaptively adjust the resource elastic allocation strategy of the target task according to the real-time workload to achieve elastic allocation of computing power;

Performing cluster analysis on target tasks through a decision support model to obtain cluster results, which include each group of target tasks and the correlation between each group of target tasks;

Based on the clustering results and correlation, the priority execution order of each group of target tasks is determined through the decision support model;

According to the clustering results, the target tasks with abnormalities are marked as abnormal tasks through the decision support model; wherein the abnormal tasks do not conform to the preset expected pattern, or the correlation between the abnormal tasks and other target tasks is lower than the preset abnormal value;

According to the priority execution order of each group of target tasks and the resource elasticity allocation strategy, each group of target tasks is assigned tasks and resources are scheduled;

Perform resource reallocation or isolation operations on abnormal tasks;

By optimizing the reward model, the quality evaluation value and/or performance index corresponding to the clustering result is used as the reward item, and the hyperparameter combination in the decision support model is dynamically optimized to obtain the best hyperparameter combination;

According to the quality evaluation value and/or performance index corresponding to the clustering result, a cluster merging operation is performed on the decision support model to improve the robustness of the clustering result;

In response to user operations, an interactive analysis of the target task is provided, and the workflow information of the target task obtained through the analysis is displayed to the user; the workflow information includes task execution status, task success rate, task exception information, and resource allocation status.

In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the artificial intelligence-based intelligent workflow management system described in any one of the first aspects above.

In a fourth aspect, a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute any of the artificial intelligence-based intelligent workflow management systems described in the first aspect.

In the technical solution provided by the embodiment of the present application, the task management module is used to monitor the real-time workload of the current online workflow; according to the real-time workload, the resource elastic allocation strategy of the target task is adaptively adjusted to achieve elastic allocation of computing power; the decision support module is used to perform cluster analysis on the target task through a decision support model to obtain clustering results, and the clustering results include each group of target tasks and the correlation between each group of target tasks; based on the clustering results and the correlation, the priority execution order of each group of target tasks is determined through the decision support model; according to the clustering results, the target task with abnormalities is marked as an abnormal task through the decision support model; wherein the abnormal task does not conform to the preset expected mode, or the correlation between the abnormal task and other target tasks is lower than the preset abnormal value; the resource management module is used to perform cluster analysis on the target task according to the group of target tasks. The priority execution order of tasks and the resource elasticity allocation strategy are used to allocate tasks and schedule resources for each group of target tasks; resource reallocation operations or isolation operations are performed on abnormal tasks; an adaptive optimization learning module is used to dynamically optimize the hyperparameter combination in the decision support model by optimizing the reward model with the quality evaluation value and/or performance index corresponding to the clustering result as the reward item to obtain the best hyperparameter combination; according to the quality evaluation value and/or performance index corresponding to the clustering result, a cluster merging operation is performed on the decision support model to improve the robustness of the clustering result; a report display module is used to respond to user operations, provide interactive analysis of the target task, and display the workflow information of the target task obtained by the analysis to the user, and the workflow information includes task execution status, task success rate, task abnormality information, and resource allocation. In the embodiment of the present application, the processing efficiency and accuracy of the workflow can be improved, intelligent task management and scheduling can be realized, the performance of the decision support model can be optimized, and comprehensive display and analysis functions can be provided at the same time, and the understanding of the workflow execution status and results can be enhanced, which is helpful to improve the level of automated management.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

FIG1 is a schematic diagram of an artificial intelligence-based intelligent workflow management system according to an embodiment of the present application;

FIG. 2 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

Currently, in traditional workflow management systems, administrators usually need to manually adjust resource allocation strategies, such as allocating computing resources to different tasks or nodes, to meet the needs of each task in the workflow.

However, in the related art, this manual adjustment method has the following problems:

The real-time workload changes of the workflow may not be detected in time. In large-scale workflows, the number and type of tasks may change dynamically, resulting in an unbalanced workload. If the administrator cannot adjust the resource allocation strategy in time, some tasks may not get enough resources to be processed, while other tasks may get too many resources, resulting in a waste of resources and reduced efficiency.

When adjusting resource allocation strategies, it may be affected by personal subjective preferences or experience, which may lead to unfair or unreasonable resource allocation and fail to meet the actual needs of each task in the workflow.

Since manually adjusting resource allocation strategies takes time and effort, it is difficult to achieve efficient resource elasticity in large-scale or high-frequency workflows, which may lead to delays in workflow processing and reduced efficiency.

Therefore, traditional workflow management systems often cannot automatically adjust resource allocation strategies according to real-time workloads, resulting in workload imbalance and slow task processing. However, AI-based workflow intelligent management systems can monitor workloads in real time and adaptively adjust resource elastic allocation strategies through task management modules and decision support modules to solve this problem and improve workflow processing efficiency and performance.

Therefore, it is urgent to design a new solution to solve at least one of the above technical problems.

The intelligent workflow management solution based on artificial intelligence provided in the embodiment of the present application can be executed by an electronic device, which can be a server, a server cluster, or a cloud server. The electronic device can also be a terminal device such as a mobile phone, a computer, a tablet computer, a wearable device (such as a smart watch, etc.). In an optional embodiment, a service program for executing the intelligent workflow management solution based on artificial intelligence can be installed on the electronic device.

FIG1 is a schematic diagram of an artificial intelligence-based workflow intelligent management system provided in an embodiment of the present application. As shown in FIG1 , the workflow intelligent management system includes:

The task management module 11 is used to monitor the real-time workload of the current online workflow; according to the real-time workload, adaptively adjust the resource elastic allocation strategy of the target task to achieve elastic allocation of computing power;

The decision support module 12 is used to perform cluster analysis on the target tasks through a decision support model to obtain a clustering result, wherein the clustering result includes each group of target tasks and the correlation between each group of target tasks; based on the clustering result and the correlation, the priority execution order of each group of target tasks is determined through the decision support model; according to the clustering result, the target tasks with abnormalities are marked as abnormal tasks through the decision support model; wherein the abnormal tasks do not conform to a preset expected mode, or the correlation between the abnormal tasks and other target tasks is lower than a preset abnormal value;

The resource management module 13 is used to perform task allocation and resource scheduling for each group of target tasks according to the priority execution order of each group of target tasks and the resource elastic allocation strategy; and perform resource reallocation operation or isolation operation on the abnormal task;

The adaptive optimization learning module 14 is used to dynamically optimize the hyperparameter combination in the decision support model by optimizing the reward model with the quality evaluation value and/or performance index corresponding to the clustering result as the reward item to obtain the best hyperparameter combination; according to the quality evaluation value and/or performance index corresponding to the clustering result, perform a cluster merging operation on the decision support model to improve the robustness of the clustering result;

The report display module 15 is used to provide interactive analysis of the target task in response to user operations, and display the workflow information of the target task obtained by analysis to the user, wherein the workflow information includes task execution status, task success rate, task exception information, and resource allocation status.

The functional principles of each module are introduced below with reference to specific examples.

The task management module 11 is used to monitor the real-time workload of the current online workflow; according to the real-time workload, the resource elastic allocation strategy of the target task is adaptively adjusted to achieve elastic allocation of computing power.

Here, the task management module 11 is responsible for the creation, allocation, scheduling and monitoring of target tasks. This module can adaptively adjust the task allocation and resource allocation strategy according to the real-time workload to support the elastic allocation of computing power.

As an optional embodiment, the task management module 11, when adaptively adjusting the resource elastic allocation strategy of the target task according to the real-time workload, is specifically used to:

The CPU utilization, memory usage, and network traffic of the current online workflow are obtained; in combination with the seasonal factors and business characteristics of the target task, the CPU utilization, memory usage, and network traffic of the current online workflow are dynamically predicted through a load prediction model to obtain the load prediction timing corresponding to the target task; according to the load prediction timing, the cloud computing platform is used to perform resource elasticity adjustment on the target task to obtain the resource elasticity allocation strategy of the target task.

The resource elastic allocation strategy includes, but is not limited to: increasing or decreasing virtual machine instances, adjusting storage capacity, and adjusting network bandwidth.

In the embodiment of the present application, the task management module 11 can be connected to the application program interface (API) of the corresponding monitoring tool or cloud computing platform to obtain key indicator information such as CPU utilization, memory usage, network traffic, etc. of the current online workflow. Among them, the application program interface is a set of definitions, programs and protocols, and mutual communication between computer software is realized through the application program interface.

As an optional embodiment, the task management module 11, when performing resource elasticity adjustment on the target task using the cloud computing platform according to the load prediction time sequence to obtain the resource elasticity allocation strategy of the target task, is specifically used to:

According to the load prediction timing, the match between the predicted resource demand of the target task and the currently available resources is evaluated; if it is evaluated that the predicted resource demand exceeds the current resource carrying capacity, the elastic resource function of the cloud computing platform is used to initiate a resource expansion request for the target task; by calling the virtual machine resources in the cloud computing platform, a virtual machine instance that matches the predicted resource demand is obtained; based on the predicted resource demand, the storage capacity and network bandwidth of the target task are dynamically adjusted to achieve elastic expansion operations for the target task; a resource elastic allocation strategy is formed based on the resource expansion request, the required matching virtual machine instance, storage capacity and network bandwidth.

Here, the task management module 11, on the one hand, can dynamically match resource requirements: the task management module 11 can evaluate the matching of the predicted resource requirements of the target task with the currently available resources, realize dynamic matching of resources, avoid excessive or insufficient allocation of resources, and thus improve resource utilization efficiency. On the other hand, it can respond in real time and expand elastically: the task management module 11 can use the virtual machine resources and elastic resource functions of the cloud computing platform to quickly respond to the resource requirements of task execution, and timely elastically expand resources to meet the requirements of task execution and improve the processing efficiency and flexibility of the workflow. On the other hand, it can dynamically adjust the storage capacity and network bandwidth: the task management module 11 can also dynamically adjust the storage capacity and network bandwidth of the task according to the resource requirements of the task, while ensuring the execution efficiency of the task, so as to further improve the processing capacity and quality of the task. On the other hand, it can realize the resource elastic allocation strategy: the task management module 11 can generate a flexible resource allocation strategy in real time according to the resource requirements of the task, so that the execution of the target task is more efficient and stable, thereby improving the processing capacity and efficiency of the entire workflow. To sum up, in the process of using the cloud computing platform to perform resource elastic adjustment on the target task, the task management module 11 can achieve resource demand matching, real-time response and elastic expansion, dynamic adjustment of storage capacity and network bandwidth, and ultimately implement resource elastic allocation strategy, thereby improving the processing capacity and efficiency of the workflow.

Assume that there is an AI-based image processing workflow involving a large number of image processing tasks. The following is an example of how to generate a resource elastic allocation strategy based on load prediction and resource expansion request: First, the task management module 11 predicts the resource requirements of image processing tasks in the future period based on historical data and load prediction models. For example, based on the average number of tasks per hour and the average processing time of each task, it is predicted that 15 parallel image processing tasks will be required in the next hour. Secondly, the task management module 11 compares the predicted resource requirements with the current available resources. Assume that the current available resources are 10 virtual machine instances, each of which has sufficient storage capacity and network bandwidth for image processing. Since the predicted resource requirements exceed the current resource carrying capacity, the task management module initiates a resource expansion request to the cloud computing platform. After receiving the resource expansion request, the cloud computing platform provides additional virtual machine instances for the image processing workflow based on the resource requirements provided in the request (such as the number of virtual machines, configuration requirements, etc.). Assume that the cloud computing platform provides 5 new virtual machine instances. The task management module 11 dynamically adjusts the storage capacity and network bandwidth of the target task based on the predicted resource requirements to achieve elastic expansion for image processing tasks. For example, according to the storage capacity and network bandwidth required by each image processing task, the corresponding resources are allocated to the task. Finally, based on the resource expansion request, the obtained virtual machine instance, the storage capacity and the network bandwidth adjustment, the task management module forms a resource elastic allocation strategy to ensure that the image processing task can be carried out in an efficient and stable manner.

In this example, the task management module initiates a resource expansion request to the cloud computing platform based on load forecasting and resource evaluation, and obtains a virtual machine instance that matches the predicted resource demand. At the same time, the storage capacity and network bandwidth are dynamically adjusted to achieve elastic expansion operations for image processing tasks. Finally, based on these operations, a resource elastic allocation strategy for the workflow is formed to ensure the processing capacity and efficiency of the task.

The decision support module 12 is used to perform cluster analysis on the target tasks through a decision support model to obtain clustering results, wherein the clustering results include each group of target tasks and the correlation between each group of target tasks; based on the clustering results and the correlation, the priority execution order of each group of target tasks is determined through a decision support model; according to the clustering results, the target tasks with abnormalities are marked as abnormal tasks through a decision support model; wherein the abnormal tasks do not conform to a preset expected pattern, or the correlation between the abnormal tasks and other target tasks is lower than a preset abnormal value.

Among them, by way of example, a density-based spatial clustering of applications with noise (DBSCAN) is used to construct a decision support model. DBSCAN is a density clustering algorithm that can automatically cluster samples with similar density and classify data points into different categories. DBSCAN divides sample points into core points, boundary points, and noise points. A core point is a sample point whose number of sample points within a certain radius reaches a minimum value, that is, a sample point whose density reaches a certain level. A boundary point is a sample point whose number of sample points within a certain radius is less than the minimum value but whose distance from the core point is within the radius. Noise points do not belong to any category. The DBSCAN algorithm achieves data clustering and noise point filtering by calculating the distance and density between sample points.

Here, the decision support module 12 can improve the efficiency of task classification and management, optimize the priority execution order of tasks, mark abnormal tasks and respond quickly, and realize intelligent decision making through functions such as cluster analysis, priority execution order determination and abnormal task marking. These functions can significantly improve the execution quality and efficiency of workflows and have important application value.

As an optional embodiment, when the decision support module 12 performs cluster analysis on the target task through the decision support model to obtain the clustering result, it is specifically used to:

According to the target clustering features and distance metrics of the target tasks, the data point density of each target task is calculated; if the data point density of the target task within a specific radius exceeds the minimum neighbor density, the target task is taken as the core data feature; the target tasks with similar densities to the core data feature are divided into the same cluster to obtain the clustering results.

The target clustering feature includes at least one of the following: CPU utilization, memory usage, and I/O data. The distance metric is Manhattan distance.

Exemplarily, the clustering analysis process in the decision support module 12 can calculate the data point density according to the target clustering feature and distance metric of the target task, and divide the target tasks with similar density into the same cluster. The following is an example to illustrate the clustering analysis function of the decision support module 12. First, select one of the following target clustering features: CPU utilization. And collect the CPU utilization data of each task. Manhattan distance is used as the distance metric, which is a distance metric method based on the difference between each feature. For each target task, the Manhattan distance between it and other target tasks is calculated, and the number of data points within a specific radius is counted. If the data point density exceeds the minimum neighbor density (the set threshold), the target task is used as the core data feature. Then, all target tasks with similar density to the core data feature are divided into the same cluster. For example, the final clustering result may be to cluster image processing tasks together, text processing tasks together, and audio processing tasks together. Then, based on the clustering results and correlation, the decision support model can determine the priority execution order of each cluster. By analyzing the task characteristics and dependencies of different clusters, it can be determined which cluster's tasks should be executed before other clusters to maximize execution efficiency. In this example, the decision support module 12 uses target clustering features (such as CPU utilization), distance metrics (Manhattan distance), and data point density calculation to implement cluster analysis of target tasks. This cluster analysis can help understand the similarities between different tasks and provide a basis for formulating a priority execution order, thereby improving the efficiency and quality of task execution.

In this example, assume that there are N target tasks, where the CPU utilization of the i-th task is x_i. Using Manhattan distance as the distance metric means calculating the distance between each pair of target tasks, which can be expressed as:

d(i,j) = |x_i - x_j|;

When calculating the density of data points, it is necessary to determine the radius and minimum neighbor density of each core data feature. Assuming R is used as the radius and k is used as the minimum neighbor density, for each target task i, it is necessary to calculate the number of other tasks n_i within the radius R from the task, and then check whether n_i is at least equal to k. If so, then mark task i as a core data feature. The mathematical expression is:

n_i = ∑_{j=1, j ≠ i}^N [d(i,j) ≤ R];

If n_i ≥ k, then task i is a core data feature.

Through these calculations, the target tasks can be divided into different clusters and their priority execution order can be determined to achieve more efficient task management and execution.

As an optional embodiment, when the decision support module 12 determines the priority execution order of each group of target tasks through a decision support model based on the clustering result and the correlation, it is specifically used to:

The cluster category to which each target task belongs is obtained from the clustering result, and the correlation between each target task and other tasks is obtained; the correlation between each target task and other tasks is quantified to obtain a correlation measure; the clustering result and the correlation measure are used as inputs of the decision support model, and the priority score of each group of target tasks is output through the decision support model; the priority score of each group of target tasks is sorted to obtain the priority execution order of each group of target tasks.

Exemplarily, it is assumed that there is a set of target tasks, including image processing tasks, text processing tasks and audio processing tasks. Based on the assumption, the cluster category to which each target task belongs is obtained from the clustering results. For example, the audio processing task is clustered into cluster A, and the text processing and image processing tasks are clustered into cluster B. Then, the correlation measure between each task is obtained, for example, the processing of the text task needs to rely on the image processing task, which shows that the two tasks have a high correlation. These correlations are quantified and processed into numerical values. For example, similarity can be converted into scores using measurement methods such as correlation coefficients or mutual information. Next, the clustering results and the correlation measure are used as inputs to the decision support model. The decision support model can use a variety of algorithms, such as fuzzy logic, neural networks, simulated annealing algorithms, etc., to calculate the task priority scores of each cluster based on the input data. Through this process, the priority scores of the target tasks of each cluster can be obtained. For example, the audio processing task of cluster A can get 80 points, while the text processing task of cluster B can get 70 points. These scores can help determine the task execution priority of each cluster. Finally, the tasks of each cluster can be sorted according to the scores to determine the task execution order of each cluster. In this example, tasks may be executed in the order of cluster A and cluster B, or the execution order of certain tasks may be determined. For example, for cluster A, the audio processing task may be executed first to ensure its completion in time. Through this example, we can see that the decision support module can help understand the correlation between different tasks through cluster analysis and correlation measurement, so as to formulate a better task execution order. This process can improve the efficiency and quality of task execution and improve the accuracy of decision making.

As an optional embodiment, the correlation between each target task and other tasks at least includes the similarity between the target tasks. Based on this, the decision support module 12 quantifies the correlation between each target task and other tasks to obtain the correlation measurement, which is specifically used for:

The Euclidean distance between the clustering features of each target task is calculated to obtain the similarity between each target task; the higher the similarity between each target task, the closer the clustering features of each target task are, and the higher the correlation between each target task.

In the decision support module, when quantifying the correlation between each target task and other tasks, the Euclidean distance calculation can be used to obtain the similarity between each target task. The following is an example of a specific calculation process. Assume that there are three target tasks: Task A, Task B, and Task C. The Euclidean distance calculation will be used to quantify the similarity between them.

First, we need to determine the clustering features of Task A, Task B, and Task C. Assume that CPU utilization is selected as the clustering feature and the CPU utilization data of each task is collected.

Assume that the CPU utilization data of task A is [60, 70, 80], the CPU utilization data of task B is [65,75, 85], and the CPU utilization data of task C is [50, 75, 90].

Then, the Euclidean distance calculation can be used to calculate the similarity between Task A, Task B, and Task C. The Euclidean distance can be expressed as:

d(A, B) = sqrt((60-65)^2 + (70-75)^2 + (80-85)^2);

d(A, C) = sqrt((60-50)^2 + (70-75)^2 + (80-90)^2);

d(B, C) = sqrt((65-50)^2 + (75-75)^2 + (85-90)^2);

By calculation, we can obtain the Euclidean distance d(A, B) between task A and task B, the Euclidean distance d(A, C) between task A and task C, and the Euclidean distance d(B, C) between task B and task C.

Next, we can use these Euclidean distance calculations to quantify the similarity between tasks. A common approach is to use a similarity metric, which can be expressed as:

Similarity = 1 / (1 + distance);

Therefore, the similarity between task A and task B, task A and task C, and task B and task C can be calculated as follows:

similarity(A, B) = 1 / (1 + d(A, B));

similarity(A, C) = 1 / (1 + d(A, C));

similarity(B, C) = 1 / (1 + d(B, C));

In this example, assume the following calculations:

Similarity(A, B) ≈ 0.098;

Similarity(A, C) ≈ 0.071;

Similarity(B, C) ≈ 0.071;

Through this calculation process, the similarities between Task A, Task B, and Task C can be quantified. These similarity values can be used for further analysis and calculation in the decision support module to determine the priority execution order of each target task.

As an optional embodiment, the correlation between each target task and other tasks at least includes resource competition. Based on this, the decision support module 12 quantifies the correlation between each target task and other tasks to obtain the correlation measurement, which is specifically used for:

If there is a resource competition relationship between different target tasks, the demand of different target tasks for each computing resource is quantified, and the correlation between each target task and other tasks is determined based on the demand; among which, the higher the demand of multiple target tasks for the same computing resource, the higher the correlation between the multiple target tasks.

In the decision support module, the demand for each computing resource of different target tasks can be quantified for the resource competition relationship between each target task and other tasks, and the correlation between tasks can be determined based on the demand. The following is an example of a specific calculation process, assuming that there are three target tasks: Task A, Task B, and Task C. Based on this assumption, computing resources will be used as resources that need to be competed. First, the demand for computing resources of each task needs to be determined. Assume that Task A requires 70% of computing resources, Task B requires 80% of computing resources, and Task C requires 65% of computing resources. Then, the correlation between tasks can be calculated based on the demand. Linear correlation is used, where the correlation can be expressed as the weighted average of the demand for the same computing resource of two tasks. In this example, the correlation between Task A and Task B, Task A and Task C, and Task B and Task C can be calculated as follows:

Correlation(A, B) = (Demand(A) + Demand(B)) / 2;

Correlation(A, C) = (Demand(A) + DemandC) / 2;

Correlation(B, C) = (Demand(B) + DemandC) / 2;

Through calculation, we can get the correlation between task A and task B, task A and task C, and task B and task C:

Correlation(A, B) = (70 + 80) / 2 = 75;

Correlation(A, C) = (70 + 65) / 2 = 67.5;

Correlation(B, C) = (80 + 65) / 2 = 72.5;

In this way, the correlation between Task A, Task B, and Task C can be quantified. These correlation values can be used for further analysis and calculation in the decision support module to determine the priority execution order of each target task.

As an optional embodiment, the correlation between each target task and other tasks at least includes the dependency relationship between previous and next tasks. Based on this, the decision support module 12 quantifies the correlation between each target task and other tasks to obtain the correlation measurement, which is specifically used for:

The dependency relationship between each target task is converted into a directed graph model; the number of dependent nodes and the number of sub-nodes of each target task are obtained, and the degree of dependency and the direction of dependency between each target task are used as the task dependency weight; the correlation measurement between each target task is obtained by calculating the weight of each task dependency.

For example, in the decision support module 12, for the dependencies between each target task and other tasks, the dependencies can be converted into a directed graph model, and the correlation between the tasks can be quantified by the number of nodes and the dependency direction. The following is an example of a specific calculation process, assuming that there are three target tasks: Task A, Task B, and Task C. A directed graph model will be used to represent the dependencies between tasks. First, according to the dependencies between tasks, the following directed graph model can be constructed:

Task A → Task B

↓

Task C

Then, the number of dependent nodes and the number of child nodes of each task can be obtained, and the degree of dependency and the direction of dependency can be used as the weight of the task dependency relationship. In this example, Task A has 1 dependent node (Task A depends on Task B) and 2 child nodes (Task B and Task C); Task B has 0 dependent nodes and 0 child nodes; Task C has 1 dependent node (Task A) and 0 child nodes. Next, the correlation measure between tasks can be calculated based on the number of dependent nodes and the number of child nodes. A commonly used calculation method is to use a weighted measure, where the correlation measure between tasks can be expressed as:

Relevance measure = α * number of dependent nodes + β * number of child nodes

Among them, α and β are factors that weigh the degree and direction of dependence.

In this example, we assume that α = 0.6 and β = 0.4 are chosen for calculation:

Correlation measure (A, B) = 0.6 * 1 + 0.4 * 2 = 1.4;

Correlation measure (A, C) = 0.6 * 1 + 0.4 * 0 = 0.6;

Correlation measure (B, C) = 0.6 * 0 + 0.4 * 0 = 0;

In this way, the correlation measures between Task A, Task B and Task C are obtained. These correlation measures can be used for further analysis and calculation in the decision support module to determine the priority execution order of each target task.

As an optional embodiment, the decision support module 12, when marking the target task with abnormalities as an abnormal task through the decision support model according to the clustering result, is specifically used to:

Determine the preset expected pattern of each target task: the preset expected pattern includes the expected pattern between each target task in different clusters, or the expected pattern of each target task in the same cluster; obtain the outlier between the target task to be determined and other target tasks according to the similarity, resource competition relationship and/or previous and subsequent task dependency relationship between each target task; locate and mark the abnormal task from each target task based on the outlier value through a decision support model.

In another optional embodiment, the decision support module 12 is also used to regularly evaluate the labeling results, check the accuracy of the classification results, the false alarm rate, etc. According to the actual situation, it is necessary to continuously optimize the parameters and algorithms of the decision support model to improve the recognition accuracy of abnormal tasks.

For example, suppose there is a set of target tasks, and use a clustering algorithm to divide them into different clusters. In each cluster, a preset expected mode of each target task can be determined.

For target tasks in different clusters, their expected patterns can be observed to understand their typical behaviors in the same cluster. These expected patterns can be defined based on the indicators, characteristics, or behaviors of the target tasks. For example, for a target task in a cluster, the expected pattern may be to produce a specific number of outputs per day. For target tasks in the same cluster, their expected patterns can be compared to detect anomalies. By calculating the similarity, resource competition relationship, and/or previous and next task dependencies between target tasks, outliers between the target task to be determined and other target tasks can be obtained.

Through decision support models, these outliers can be used to locate and mark abnormal tasks. Specific methods may include setting anomaly thresholds or applying anomaly detection algorithms to determine whether the target task deviates from the preset expected pattern. If the indicators or behaviors of the target task are obviously inconsistent with the preset expected pattern, then it can be marked as an abnormal task.

The benefit of marking abnormal tasks is that in the subsequent decision-making process, special treatment can be given to abnormal tasks, such as allocating additional resources, adjusting task priorities, or taking other intervention measures to ensure the normal operation of the entire system.

This method of marking abnormal tasks based on clustering results and decision support models can help operators discover and solve potential problems in a timely manner and improve the reliability and efficiency of the system.

The adaptive optimization learning module 14 is used to dynamically optimize the hyperparameter combination in the decision support model by optimizing the reward model with the quality evaluation value and/or performance index corresponding to the clustering result as the reward item to obtain the best hyperparameter combination; according to the quality evaluation value and/or performance index corresponding to the clustering result, perform a cluster merging operation on the decision support model to improve the robustness of the clustering result.

Through the adaptive optimization learning module 14, on the one hand, by optimizing the reward model, combining the quality evaluation value and/or performance index corresponding to the clustering result as a reward item, the hyperparameter combination can be automatically optimized in the decision support model. This can improve the performance of the decision support model and meet the system's requirements and constraints to the greatest extent. By dynamically optimizing the hyperparameters, the system can automatically learn and adjust the hyperparameters to adapt to different scenarios and task requirements. On the other hand, according to the quality evaluation value and/or performance index corresponding to the clustering result, performing a cluster merging operation in the decision support model can improve the robustness of the clustering result. The cluster merging operation can classify similar tasks or targets into one category, reduce the impact of noise or abnormal conditions in the clustering results, and improve the accuracy and reliability of the clustering results. This can help the system better understand the relationship between the target tasks and provide more accurate and reliable decision support. In short, through the application of the adaptive optimization learning module 14, the clustering results and quality evaluation indicators can be combined to intelligently adjust and optimize the decision support model. This dynamic optimization method can improve the performance and robustness of the system, help the system better adapt to and respond to different tasks and environmental changes, and achieve more efficient and reliable decision support.

As an optional embodiment, the adaptive optimization learning module 14 performs a cluster merging operation on the decision support model according to the quality evaluation value and/or performance index corresponding to the clustering result to improve the robustness of the clustering result, specifically for:

According to the quality evaluation value and/or performance index of the clustering result, the performance of each cluster is evaluated to obtain the evaluation index of each cluster; the evaluation index includes at least one of the following: silhouette coefficient, internal distance, and inter-class distance; the evaluation index is used to measure the compactness and separation of the cluster; the clustering results that meet the merging conditions in the decision support model are centroidally merged; the merged result is updated to the clustering result to obtain the updated clustering result, and the label of each updated cluster is updated to perform subsequent adaptive optimization and processing on the decision support model.

Specifically, the following is a specific calculation process of an optional embodiment: First, the performance of each cluster is evaluated according to the quality evaluation value and/or performance index of the clustering result to obtain the evaluation index of each cluster. These evaluation indexes may include silhouette coefficient, internal distance, inter-class distance, etc. These indexes can be used to measure the compactness and separation of the cluster, that is, the distance between the data in the cluster should be as small as possible, and the distance between the clusters should be as large as possible. Then, according to the evaluation index, the clusters that meet the merging conditions in the decision support model can be merged by centroid. Specifically, clusters with high similarity, close evaluation indexes, and close distances can be merged to reduce noise and abnormalities in the clustering results and improve the accuracy and reliability of the clustering results. After merging, the merged results are updated to the clustering results to obtain the updated clustering results. In order to maintain consistency, it is also necessary to update the labels of each updated cluster for subsequent adaptive optimization and processing. The update of the label can be calculated based on the new cluster centroid. Through this adaptive optimization learning module 14, the cluster merging operation can be performed in combination with the quality evaluation value and/or performance index of the clustering result, thereby improving the accuracy and reliability of the clustering result. This adaptive optimization learning module 14 can help the system better understand the relationship between the target tasks and provide more accurate and reliable decision support.

The report display module 15 is used to provide interactive analysis of the target task in response to user operations, and display the workflow information of the target task obtained by analysis to the user, wherein the workflow information includes task execution status, task success rate, task exception information, and resource allocation status.

In the report display module, interactive analysis of the target task can be provided according to the user's operation request, and the workflow information obtained from the analysis can be displayed to the user, including task execution status, task success rate, task exception information, resource allocation, etc. The following is an example:

Assume that an enterprise has used the WeTab platform to manage its business processes, and has obtained the clustering results and quality assessment information about the target tasks through the previous modules. At this point, the enterprise can use the report display module to conduct in-depth analysis and monitoring of its business processes to monitor and improve the execution efficiency and accuracy of the entire business process. First, you can select the target tasks that need to be analyzed and query them through the report display module. For example, you can choose to query and analyze the target tasks in a cluster. Secondly, the report display module will display the workflow information of each target task in the cluster, including task execution status, task success rate, task exception information, resource allocation, etc. For example, you can display the execution of a task in the past period of time, including task completion time, resource usage, etc. And, if there is an exception or failure in this task, you can display the relevant exception information and processing records. On this basis, interactive analysis and comparison can be carried out, such as comparing the execution of different tasks in the same cluster, or comparing the execution efficiency and accuracy of tasks between different clusters. In addition, administrators can make relevant decisions or adjustments, such as increasing or decreasing the allocation of a certain resource, changing the running order of tasks, etc., to improve the efficiency and accuracy of the entire business process.

In this way, the report display module can provide enterprises with timely, accurate and useful business process information, playing an important role in monitoring and improving business processes.

In the embodiment of the present application, by dynamically adjusting the resource elastic allocation strategy in real time, the system can perform elastic allocation of computing power according to the actual workload, avoid workload imbalance, and thus improve the processing efficiency and response speed of tasks. Through the decision support model, the system can cluster the target tasks and determine the priority execution order, thereby realizing intelligent task management and scheduling, improving the processing accuracy and priority of tasks in the workflow, and reducing the task processing time. Through the adaptive optimization learning module, the system can dynamically optimize the hyperparameter combination in the decision support model according to the actual workload and quality evaluation value/performance index, improve the prediction and decision accuracy of the model, and further improve the intelligence level and performance of the decision support system. The report display module provides interactive analysis and display functions, and users can intuitively understand the execution status, success rate, abnormal information and resource allocation of the workflow through the interface, which is convenient for rapid problem discovery, optimization adjustment and decision making. In summary, the system can improve the processing efficiency and accuracy of the workflow, realize intelligent task management and scheduling, optimize the performance of the decision support model, and provide comprehensive display and analysis functions at the same time, enhance the user's understanding of the execution status and results of the workflow, and help improve the production efficiency and decision-making level of the enterprise.

In another embodiment of the present application, a workflow intelligent management method based on artificial intelligence is also provided, the method comprising:

Monitor the real-time workload of current online workflows;

Adaptively adjust the resource elastic allocation strategy of the target task according to the real-time workload to achieve elastic allocation of computing power;

Performing cluster analysis on target tasks through a decision support model to obtain cluster results, which include each group of target tasks and the correlation between each group of target tasks;

Based on the clustering results and correlation, the priority execution order of each group of target tasks is determined through the decision support model;

According to the clustering results, the target tasks with abnormalities are marked as abnormal tasks through the decision support model; wherein the abnormal tasks do not conform to the preset expected pattern, or the correlation between the abnormal tasks and other target tasks is lower than the preset abnormal value;

According to the priority execution order of each group of target tasks and the resource elasticity allocation strategy, each group of target tasks is assigned tasks and resources are scheduled;

Perform resource reallocation or isolation operations on abnormal tasks;

By optimizing the reward model, the quality evaluation value and/or performance index corresponding to the clustering result is used as the reward item, and the hyperparameter combination in the decision support model is dynamically optimized to obtain the best hyperparameter combination;

According to the quality evaluation value and/or performance index corresponding to the clustering result, a cluster merging operation is performed on the decision support model to improve the robustness of the clustering result;

In response to user operations, an interactive analysis of the target task is provided, and the workflow information of the target task obtained through the analysis is displayed to the user; the workflow information includes task execution status, task success rate, task exception information, and resource allocation status.

The method embodiment proposed in the present application also includes the various method steps in the above embodiments that can be executed by an artificial intelligence-based workflow intelligent management system.

In another embodiment of the present application, an electronic device is also provided, as shown in FIG2 , including: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

Memory, used to store computer programs;

The processor is used to implement the intelligent test management system described in the method embodiment when executing the program stored in the memory.

The communication bus 1140 mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 1140 may be divided into an address bus, a data bus, a control bus, etc.

For ease of representation, FIG2 shows only one thick line, but this does not mean that there is only one bus or one type of bus.

The communication interface 1120 is used for communication between the above electronic device and other devices.

The memory 1130 may include a random access memory (RAM) or a non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

The processor 1110 may be a general purpose processor, including a central processing unit (Central Processor).

It can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement the various steps that can be executed by the electronic device in the above system embodiment.

Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

The present application is described with reference to the flowcharts and/or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram, as well as the combination of the processes and/or boxes in the flowchart and/or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

In a typical configuration, a computing device includes one or more processors (CPU), input/output interfaces, network interfaces, and memory.

Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims (10)

1. An artificial intelligence based workflow intelligent management system, the system comprising:

The task management module is used for monitoring the real-time workload of the current online workflow; according to the real-time workload, the resource elastic allocation strategy of the target task is adaptively adjusted so as to realize the computational power elastic allocation;

The decision support module is used for carrying out cluster analysis on the target tasks through a decision support model so as to obtain a cluster result, wherein the cluster result comprises all groups of target tasks and the correlation among all groups of target tasks; determining the priority execution sequence of each group of target tasks through a decision support model based on the clustering result and the correlation; marking the abnormal target task as an abnormal task through a decision support model according to the clustering result; wherein the abnormal task does not conform to a preset expected mode, or the correlation between the abnormal task and other target tasks is lower than a preset abnormal value;

the resource management module is used for carrying out task allocation and resource scheduling on each group of target tasks according to the priority execution sequence of each group of target tasks and the resource elastic allocation strategy; performing resource reallocation operation or isolation operation on the abnormal task;

The self-adaptive optimization learning module is used for dynamically optimizing the super-parameter combination in the decision support model by optimizing the reward model by taking the quality evaluation value and/or the performance index corresponding to the clustering result as a reward item so as to obtain the optimal super-parameter combination; according to the quality evaluation value and/or the performance index corresponding to the clustering result, cluster merging operation is carried out on the decision support model so as to improve the robustness of the clustering result;

The report display module is used for responding to user operation, providing interactive analysis of the target task, and displaying workflow information of the target task obtained by analysis to a user, wherein the workflow information comprises task execution state, task success rate, task abnormality information and resource allocation condition.

2. The workflow intelligent management system based on artificial intelligence according to claim 1, wherein the task management module is specifically configured to, when adaptively adjusting a resource elastic allocation policy of a target task according to the real-time workload:

Acquiring the CPU utilization rate, the memory usage amount and the network flow of the current online workflow;

Combining seasonal factors and business characteristics of the target task, and dynamically predicting the CPU utilization rate, the memory usage amount and the network flow of the current online workflow through a load prediction model to obtain a load prediction time sequence corresponding to the target task;

According to the load prediction time sequence, executing resource elastic adjustment on the target task by utilizing a cloud computing platform to obtain a resource elastic allocation strategy of the target task;

the resource elastic allocation strategy comprises the steps of increasing or decreasing virtual machine instances, adjusting storage capacity and adjusting network bandwidth.

3. The workflow intelligent management system based on artificial intelligence according to claim 2, wherein the task management module, according to the load prediction time sequence, is specifically configured to, when performing resource elastic adjustment on the target task by using a cloud computing platform to obtain a resource elastic allocation policy of the target task:

According to the load prediction time sequence, evaluating the matching condition of the predicted resource demand of the target task and the current available resources;

If the predicted resource requirement is estimated to exceed the current resource bearing capacity, initiating a resource expansion request aiming at the target task by utilizing the elastic resource function of the cloud computing platform;

virtual machine resources in the cloud computing platform are called to obtain virtual machine instances matched with predicted resource requirements;

Dynamically adjusting the storage capacity and the network bandwidth of the target task based on the predicted resource demand so as to realize the elastic capacity expansion operation of the target task;

and forming the resource elastic allocation strategy according to the resource expansion request, the virtual machine instance to be matched, the storage capacity and the network bandwidth.

4. The workflow intelligent management system based on artificial intelligence of claim 1, wherein the decision support module is configured to, when performing cluster analysis on the target task by using a decision support model to obtain a cluster result:

Calculating the data point density of each target task according to the target cluster characteristics and the distance measurement of the target task; the target cluster feature comprises at least one of the following: CPU utilization, memory usage, I/O data; the distance measure is Manhattan distance;

If the data point density of the target task in the specific radius exceeds the minimum neighbor density, taking the target task as a core data characteristic;

And dividing target tasks similar to the characteristic density of the core data into the same cluster to obtain the clustering result.

5. The workflow intelligent management system based on artificial intelligence of claim 1, wherein the decision support module, when determining the priority execution order of each group of target tasks by a decision support model based on the clustering result and the correlation, is specifically configured to:

Acquiring cluster categories to which each target task belongs from the clustering result, and acquiring correlations between each target task and other tasks;

carrying out quantization processing on the relevance between each target task and other tasks to obtain a relevance measurement;

Taking the clustering result and the correlation measure as the input of the decision support model, and outputting the priority scores of the target tasks of each group through the decision support model;

And sequencing the priority scores of the target tasks of each group to obtain the priority execution sequence of the target tasks of each group.

6. The intelligent workflow management system of claim 5, wherein the relevance between each target task and other tasks comprises at least a similarity between target tasks;

The decision support module is specifically configured to, when performing quantization processing on the correlation between each target task and other tasks to obtain a correlation metric:

performing Euclidean distance calculation on the clustering features among the target tasks to obtain the similarity among the target tasks; the higher the similarity among the target tasks is, the closer the clustering features of the target tasks are, and the relevance among the target tasks is.

7. The intelligent workflow management system based on artificial intelligence of claim 5, wherein the correlation between each target task and other tasks comprises at least a resource competition relationship;

The decision support module is specifically configured to, when performing quantization processing on the correlation between each target task and other tasks to obtain a correlation metric:

If the resource competition relationship exists among different target tasks, quantifying the demand degree of the different target tasks for each computing power resource, and determining the correlation between each target task and other tasks based on the demand degree; the higher the demand of a plurality of target tasks on the same computing power resource, the higher the correlation among the plurality of target tasks.

8. The intelligent workflow management system based on artificial intelligence of claim 5, wherein the correlation between each target task and other tasks comprises at least a front-back task dependency;

The decision support module is specifically configured to, when performing quantization processing on the correlation between each target task and other tasks to obtain a correlation metric:

Converting the dependency relationship among all target tasks into a directed graph model; the method comprises the steps of obtaining the number of dependent nodes and the number of child nodes of each target task, and taking the degree of dependence and the dependence direction between each target task as task dependence relation weights; and calculating the correlation measurement between the target tasks by calculating the weight value of each task dependency relation.

9. The workflow intelligent management system based on artificial intelligence according to claim 1, wherein the decision support module is specifically configured to, when marking, according to the clustering result, a target task with an abnormality as an abnormal task through a decision support model:

Determining preset expected modes of each target task: the preset expected modes comprise expected modes among all target tasks in different clusters or expected modes of all target tasks in the same cluster;

Acquiring abnormal values between the target tasks to be judged and other target tasks according to the similarity among the target tasks, the resource competition relationship and/or the front-back task dependency relationship;

And positioning and marking the abnormal tasks from the target tasks based on the abnormal values through a decision support model.

10. The workflow intelligent management system based on artificial intelligence according to claim 1, wherein the adaptive optimization learning module is specifically configured to, when performing a cluster merging operation on a decision support model according to a quality evaluation value and/or a performance index corresponding to the clustering result to improve the robustness of the clustering result:

Evaluating the performance of each cluster according to the quality evaluation value and/or the performance index of the clustering result to obtain an evaluation index of each cluster; the evaluation index includes at least one of: profile coefficient, internal distance, inter-class distance; the evaluation index is used for measuring the compactness and the separation degree of the clusters;

centroid merging is carried out on clustering results meeting merging conditions in the decision support model;

Updating the merging result into the clustering result to obtain an updated clustering result, and updating the label of each updated cluster to perform subsequent self-adaptive optimization and processing on the decision support model.