A service processing method, device, medium and product

Abstract

The application relates to the technical field of agent platform architecture, and in particular to a business processing method, equipment, medium and product. The method comprises the following steps: linking a pre-built agent platform and an enterprise system of a target enterprise, and constructing a template task tree based on historical system operation data of the enterprise system; receiving a user demand, decomposing the user demand into a task sequence based on the template task tree, and determining a task type of each task in the task sequence; selecting an agent from an agent candidate pool to form an agent list based on the capability metadata of the agent candidate pool and the task type of the task sequence, and assigning the task sequence to the agent list for execution; determining a plurality of intervention nodes, determining a task execution deviation degree at each intervention node, and adjusting an execution strategy of a subsequent task of the intervention node in the task sequence when the task execution deviation degree of any intervention node exceeds a preset intervention threshold. The application can realize accurate generation and adaptive execution of enterprise tasks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agent platform architecture, in particular to a business processing method, device, medium and product. BACKGROUND

[0002] In the process of enterprise digital transformation, more and more enterprises introduce AI agent platforms to assist in business processing in order to improve business efficiency and automation level. However, the current related technology still has many shortcomings in multiple dimensions, which is difficult to meet the needs of fine business processing of enterprises. On the one hand, the AI agent platform and the enterprise system are often in a relatively independent state, and the two lack a deep cooperation mechanism. Even if some solutions realize coupling through a simple interface, they do not realize systematic integration and value mining of the enterprise system, and cannot form a standardized task decomposition framework that fits the actual business logic of the enterprise, so that the system cannot convert abstract requirements into structured and executable specific tasks when the user publishes tasks through the enterprise system. On the other hand, in the task allocation link, the existing technology generally adopts a single-dimensional screening logic of random selection or manual selection based on experience, which does not take into account the ability information of different agents, nor the individualized task needs of users, resulting in insufficient adaptability of agents and tasks, and ultimately causing the task execution result to not meet the user's expectations, increasing the business rework cost, and seriously affecting the efficiency and quality of enterprise business processing. SUMMARY

[0003] In order to solve the problem that the AI agent platform of the prior art cannot realize accurate generation and adaptive execution of enterprise tasks, the present application provides a business processing method, device, medium and product.

[0004] In a first aspect, the present application provides a business processing method, which adopts the following technical solution:

[0005] A business processing method, comprising:

[0006] When receiving an access request of a target enterprise, link a pre-built agent platform and an enterprise system of the target enterprise, obtain historical system operation data of the enterprise system, and construct a template task tree based on the historical system operation data;

[0007] Receive a user demand uploaded by a target personnel of the target enterprise through the enterprise system, decompose the user demand into a task sequence based on the template task tree, and determine the task type of each task in the task sequence;

[0008] Obtain the ability metadata of an agent candidate pool of the agent platform, select agents from the agent candidate pool to form an agent list based on the ability metadata and the task type of the task sequence, and assign the task sequence to the agent list for execution;

[0009] According to the depth of the task sequence, a plurality of intervention nodes are determined, a task execution deviation degree is determined at each intervention node, and when the task execution deviation degree of any intervention node exceeds a preset intervention threshold, an execution strategy of a subsequent task of the intervention node in the task sequence is adjusted.

[0010] By adopting the technical solution, the intelligent agent platform is linked with the enterprise system, the template task tree is constructed through historical data, the standardization and structuring of the business process are realized, the user demand is decomposed into a task sequence and the task type is determined, the intelligent agent is selected based on the capability metadata, the adaptability and efficiency of task execution are ensured, the intervention nodes are set and the execution strategy is dynamically adjusted, the task deviation can be corrected in time, the accuracy and stability of business processing are ensured, and the accurate generation and adaptive execution of enterprise tasks are realized.

[0011] In a preferred example, the application can be further configured to: constructing the template task tree based on the historical system operation data, comprising:

[0012] The historical system operation data is cut based on a session timeout rule to obtain business activity data, and the business activity data is clustered to obtain a plurality of clustering clusters;

[0013] For each clustering cluster in the plurality of clustering clusters, sequence pattern mining is performed on the cluster members of the clustering cluster to obtain the support degree of each activity mode, the target activity mode is selected from the cluster members based on the support degree, and the dependency relationship and mapping strength coefficient between the target activity modes are identified;

[0014] Each target activity mode obtained by the plurality of clustering clusters is constructed as a node based on the dependency relationship to construct an initial template task tree;

[0015] For each node in the initial template task tree, the cluster-in frequency of the node in the clustering cluster and the global frequency of the node in the plurality of clustering clusters are determined, and the execution tendency weight between the node and the parent node of the node is determined based on the cluster-in frequency and the global frequency;

[0016] The execution tendency weight and the mapping strength coefficient between each two nodes obtained are marked at the corresponding positions of the initial template task tree to obtain the template task tree.

[0017] By adopting the technical solutions, the session cutting and clustering analysis are used to realize effective classification of business activities, accurately distinguish different business scenarios, mine sequence patterns and filter target activity patterns, extract core business steps, identify dependency relationships and mapping strength coefficients, and determine the correlation tightness between business steps, calculate execution tendency weights, quantify the execution preference between steps, and finally construct a template task tree with business logic and quantitative guidance significance, thereby providing precise process support for subsequent requirement disassembly and task execution.

[0018] In a preferred example, the application can be further configured to: the method further comprises:

[0019] creating a test environment, and generating a plurality of test cases containing different types of test tasks and network environment parameter combinations;

[0020] For each agent in the agent candidate pool, each test case is executed to obtain the comprehensive test results of the agent for each task type under each network environment, and the comprehensive test results include task performance indicators and environment sensitivity indicators;

[0021] Based on the comprehensive test results of each agent and a preset scoring rule, a performance score matrix is constructed for each agent, the rows of the performance score matrix represent network environments, the columns represent task types, and the elements represent index score vectors, and the index score vectors include index scores of each task performance indicator and each environment sensitivity indicator;

[0022] The comprehensive test results and the performance score matrix of each agent in the agent candidate pool are stored as the capability metadata.

[0023] By adopting the technical solutions, multi-dimensional test cases are designed to comprehensively cover different task types and network environments, ensuring the integrity of the performance evaluation of the agent; through the comprehensive test results and the performance score matrix, the capability of the agent is quantified and structured, and the performance of the agent in different scenarios is intuitively presented; the capability metadata is stored to provide objective and accurate data basis for subsequent selection of the agent, and subjective selection bias is avoided.

[0024] In a preferred example, the application can be further configured to: selecting an agent from the agent candidate pool based on the capability metadata and the task type of the task sequence to form an agent list, comprising:

[0025] obtaining the current network environment and the lower limit of the task performance indicator of each task type;

[0026] For each task in the task sequence, based on the capability metadata, an agent group capable of meeting all task performance index lower limits of the task type of the task in a network environment matching the current network environment is screened from the agent candidate pool;

[0027] Based on the index score of the dependency constraint relationship and the environment sensitivity index of each agent, one agent is selected from the agent group of each task to form the agent list, and the dependency constraint relationship includes that the agents selected by any two tasks without dependency relationship are not the same.

[0028] By adopting the above technical solution, the agent is screened in combination with the current network environment, the adaptability of the agent in the actual running environment is ensured, and performance loss caused by environmental inadaptability is avoided; the performance index lower limit is set to ensure that the selected agent has basic execution capability; dependency constraints are considered to avoid conflicts between agents of non-dependent tasks, and more stable agents are selected in combination with the environment sensitivity index score.

[0029] In a preferred example, the application can be further configured to determine the current task execution deviation degree of the current intervention node, including:

[0030] Determine the task type of each preceding task before the current intervention node in the task sequence, and determine the task type composition of the executed task sequence based on the task type of each preceding task;

[0031] Determine the dynamic weight of each deviation dimension based on the task type composition;

[0032] Obtain the current task execution result of the current intervention node, and calculate the first task execution deviation degree of the current task execution result in each deviation dimension based on the current task execution result and the dynamic weight of each deviation dimension;

[0033] Determine the preceding task and the subsequent task of the current intervention node in the task sequence, and update the first task execution deviation degree based on the mapping intensity coefficient between the preceding task and the subsequent task in the template task tree to obtain the second task execution deviation degree;

[0034] Determine whether the current intervention node is the first intervention node in the plurality of intervention nodes; if the current intervention node is the first intervention node, the second task execution deviation degree is taken as the current task execution deviation degree;

[0035] If the current intervention node is not the first intervention node, then the time-series cumulative deviation is calculated based on the task execution deviation of all intervention nodes between the current intervention node, and the second task execution deviation is updated based on the time-series cumulative deviation to obtain the current task execution deviation.

[0036] By adopting the above technical solutions, dynamic weights are determined based on the types of preceding tasks, making deviation assessment more aligned with business scenarios and avoiding the limitations of a single weight. The influence of inter-task correlation on deviation is considered by combining the mapping strength coefficient, improving the comprehensiveness of deviation assessment. The introduction of time-series cumulative deviation takes into account the superposition effect of historical deviations, resulting in a more accurate calculation of the degree of task execution deviation.

[0037] In a preferred embodiment, this application can be further configured such that: adjusting the execution strategy of subsequent tasks of the intervention node in the task sequence includes:

[0038] Starting from the first task in the task sequence, the intermediate task execution results after each task is completed are obtained sequentially, and the degree of intermediate task execution deviation of the intermediate task execution results in each deviation dimension is calculated.

[0039] The intermediate task execution deviation obtained after each task is completed is compared with the preset alarm threshold, and the task corresponding to the intermediate task execution deviation that exceeds the preset alarm threshold is taken as the target task.

[0040] Adjust the agent corresponding to the target task in the agent list, and execute the target task and subsequent tasks in the task sequence according to the adjusted agent list.

[0041] By adopting the above technical solutions, the intermediate results and deviation of tasks can be monitored in real time, enabling dynamic tracking of the task execution process. This allows for the rapid identification of target tasks that exceed the threshold; timely adjustment of the intelligent agent corresponding to the target task to address task deviation issues and prevent deviations from propagating to subsequent tasks; and re-execution of the target task and subsequent tasks to ensure that business processing returns to the correct track. This effectively reduces the impact of deviations on the overall business results and guarantees the quality of business processing.

[0042] In a preferred embodiment, this application can be further configured such that the method also includes:

[0043] When the task sequence is completed, determine the lowest-level node of each task in the task sequence in the template task tree;

[0044] Determine the child nodes of the lowest-level node in the template task tree, and arrange each child node in descending order of execution tendency weight with the lowest-level node to obtain the recommended task sequence;

[0045] The recommended task sequence is fed back to the target personnel so that the corresponding task can be executed based on the target personnel's selection of the recommended task sequence.

[0046] By adopting the above technical solution, recommended task sequences are generated based on template task trees and execution tendency weights, which are in line with the enterprise's historical business execution habits, improving the rationality and practicality of the recommendations; the recommended sequences are fed back to the target personnel, providing clear guidance for subsequent business processing and reducing personnel decision-making costs; and personnel can make their own choices, balancing the flexibility of intelligent recommendations with manual intervention.

[0047] Secondly, this application provides an electronic device that adopts the following technical solution:

[0048] One or more processors;

[0049] Memory;

[0050] At least one application, wherein the at least one application is stored in memory and configured to be executed by at least one processor, the at least one application being configured to: execute the business processing method as described in any of the first aspects.

[0051] Thirdly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0052] A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the business processing method as described in any of the first aspects.

[0053] Fourthly, this application provides a computer program product, which adopts the following technical solution:

[0054] A computer program product includes a computer program that, when executed by a processor, implements the business processing method as described in any of the first aspects.

[0055] In summary, this application includes the following beneficial technical effects:

[0056] This application achieves standardization and structuring of business processes by linking an intelligent agent platform with enterprise systems and constructing template task trees using historical data; it decomposes user requirements into task sequences and determines task types; it selects intelligent agents based on capability metadata to ensure the adaptability and efficiency of task execution; and it sets intervention nodes and dynamically adjusts execution strategies to promptly correct task deviations, ensuring the accuracy and stability of business processing, thereby realizing the precise generation and adaptable execution of enterprise tasks. Attached Figure Description

[0057] Figure 1This is a flowchart illustrating a business processing method provided in an embodiment of this application;

[0058] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0059] The following is in conjunction with the appendix Figure 1 To be continued Figure 2 This application will be described in further detail.

[0060] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

[0061] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0063] It should be noted that, in the optional embodiments of this application, the data related to object information, when applied to specific products or technologies, requires the permission or consent of the object. Furthermore, the collection, use, and processing of this data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. In other words, if the embodiments of this application involve data related to an object, it must be obtained with the object's authorization and consent, the authorization and consent of relevant departments, and in accordance with the relevant laws, regulations, and standards of the country and region. If the embodiments involve personal information, the acquisition of all personal information requires the individual's consent. If sensitive information is involved, the separate consent of the information subject is required. The embodiments also need to be implemented with the object's authorization and consent.

[0064] This application provides a business processing method, such as... Figure 1As shown, the method provided in this application embodiment is executed by an electronic device, which can be a server or a terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. The terminal device and the server can be directly or indirectly connected via wired or wireless communication. This application embodiment does not impose any limitations on this connection. The method includes steps S101-S104, wherein:

[0065] S101. When an access request from a target enterprise is received, the pre-built intelligent agent platform is linked with the target enterprise's enterprise system to obtain historical system operation data of the enterprise system, and a template task tree is constructed based on the historical system operation data.

[0066] Specifically, the intelligent agent platform is a pre-built management system integrating multiple intelligent agents with specific functions, enabling agent registration, scheduling, and monitoring. The specific types of intelligent agents can be set according to actual needs; optional intelligent agents may include data extraction agents, text analysis agents, image recognition agents, and decision recommendation agents. Standardized interfaces are developed to support data interaction between the intelligent agent platform and enterprise systems, configure data synchronization rules, and obtain historical system operation data from the enterprise system, such as business documents, approval records, and operation logs from the past three years.

[0067] The acquired historical system operation data is sequentially segmented into business activities, clustered, and the activity pattern is mined to obtain target activity patterns. Clustering analysis yields multiple clusters, each representing a business scenario. Target activity patterns represent effective, high-frequency activity patterns, and each target activity pattern represents a single task, such as creating an order and approving an order as separate tasks. For each cluster, the dependencies between target activity patterns are identified, and a sub-task tree is constructed based on these dependencies. Nodes in the sub-task tree represent target activity patterns. After obtaining the sub-task trees for each cluster, all sub-task trees are integrated to obtain a template task tree. The integration method can be to directly connect the root nodes of each sub-task tree to the same node, with the same node serving as the root node of the template task tree.

[0068] S102. Receive user requirements uploaded by target personnel of the target enterprise through the enterprise system, break down the user requirements into a task sequence based on the template task tree, and determine the task type of each task in the task sequence.

[0069] Specifically, the target personnel are any employee within the target company. A requirement submission interface is developed within the company system, supporting methods such as text input, file upload, or voice-to-text submission. A similarity algorithm is used to calculate the similarity between the user's requirement and each subtask tree. The subtask tree with the highest similarity is selected as the business scenario for matching the user's requirement, and the subtask tree where the user's requirement is successfully matched is recorded as the target subtask tree. Optionally, the similarity algorithm can be a Word2Vec or BERT model.

[0070] During the construction of the template task tree, a structured feature library is built for all nodes in the template task tree. For each node, its node features include: node attributes (node ​​ID, parent node ID, child node ID), functional description (the node's business objective), input requirements (the prerequisite data required before the node execution), and output results (the data generated after the node execution). Taking the sub-task tree corresponding to the business scenario of generating financial statements as an example, the task flow of each node in this sub-task tree is: collecting financial data, data validation, data analysis, generating reports, and report review. Taking the data validation node as an example, its business objective is to verify the integrity and format correctness of the data, the input requirement is to obtain financial data, and the output result is the structured data after successful validation.

[0071] User requirements are structured using features. User objectives and preconditions are extracted via NPL and converted into vector form. For example, a user requirement is to directly generate a financial statement based on the current input data; the core objective is generating the financial statement, and the precondition is using the current input data. When no preconditions exist in the user requirement, the root node of the target subtask tree is directly used as the mapping node. When preconditions exist, vector similarity matching is performed between the preconditions and each node in the target subtask tree. Nodes with successful matches are used as mapping nodes. The matching method compares the similarity between the preconditions and the historical input data of each node in the target subtask tree, and the node with the highest similarity is used as the mapping node. The preconditions serve as constraints for the mapping node. The user objective is matched with the output content of each node, and the node with the highest similarity is used as the terminal node. After determining the mapping nodes and terminal nodes, the path from the mapping node to the terminal node is extracted from the target subtask tree as the task sequence corresponding to the user requirement.

[0072] For task types, a classification model can be pre-trained, using the task sequence as the training set. The model then outputs the task type for each task in the sequence. For example, acquiring monthly financial data corresponds to the task type "data collection," verifying financial data corresponds to the task type "data verification," calculating profit corresponds to the task type "data analysis," generating reports corresponds to the task type "document generation," and reviewing reports corresponds to the task type "process approval."

[0073] S103. Obtain the capability metadata of the agent candidate pool of the agent platform, select agents from the agent candidate pool based on the capability metadata and the task type of the task sequence to form an agent list, and assign the task sequence to the agent list for execution.

[0074] Specifically, a test environment is created, generating test cases containing different types of test tasks and combinations of network environment parameters. All test cases are executed for each agent in the agent candidate pool, and the comprehensive test results for each agent executing each test case are recorded. The comprehensive test results include task performance metrics and environment sensitivity metrics. Task performance metrics indicate the quality of task execution, including accuracy, processing time, and success rate; environment sensitivity metrics indicate the impact on network environment parameters, including stability coefficient, decay rate, and environment adaptability.

[0075] The preset scoring rules include the weighting of each indicator and the scoring calculation method. Based on the preset scoring rules, an indicator score vector is calculated for each agent executing each test case. For each agent, the indicator score vectors obtained from executing all test cases form a performance score matrix. The rows of the performance score matrix represent the network environment, with different combinations of network environment parameters for different network environments. The list represents the task type, and the elements represent the indicator score vectors obtained from executing test cases for the corresponding network environment and task type. The comprehensive test results and performance score matrix of each agent in the agent candidate pool are stored as capability metadata.

[0076] Based on capability metadata and task type, a corresponding agent is selected for each task in the task sequence. The selected agents are arranged in the order of their corresponding tasks in the task sequence to obtain an agent list. A corresponding task is assigned to each agent in the agent list. Through the scheduling engine of the agent platform, execution instructions are sent to the agents in the order of the task sequence, and the execution status is monitored in real time.

[0077] S104. Determine several intervention nodes based on the depth of the task sequence, determine the degree of task execution deviation at each intervention node, and adjust the execution strategy of subsequent tasks of the intervention node in the task sequence when the degree of task execution deviation of any intervention node exceeds the preset intervention threshold.

[0078] Specifically, the depth of the task sequence is determined, where depth represents the total number of tasks 'a' in the sequence. Intervention nodes are set based on this depth according to the principle of equal intervals. This principle means that an intervention node is defined at every interval step, which can be preset and stored in the electronic device. For example, if the interval step of the intervention node is 3, then when a=10, intervention nodes are set after the 3rd, 6th, and 9th tasks.

[0079] When the task execution deviation of any intervention node exceeds the preset intervention threshold, the target task with the largest deviation is determined from the tasks previously completed by that intervention node, the agent used for the target task is adjusted, and the target task and its subsequent tasks in the task sequence are executed according to the adjusted agent list.

[0080] This embodiment links the intelligent agent platform with the enterprise system, constructs a template task tree using historical data, and achieves standardization and structuring of business processes; it decomposes user requirements into task sequences and determines task types; it selects intelligent agents based on capability metadata to ensure the adaptability and efficiency of task execution; and it sets intervention nodes and dynamically adjusts execution strategies to correct task deviations in a timely manner, ensuring the accuracy and stability of business processing, thus realizing the precise generation and adaptable execution of enterprise tasks.

[0081] One possible implementation of this application embodiment involves constructing a template task tree based on historical system operation data, including:

[0082] Based on session timeout rules, historical system operation data is segmented to obtain business activity data, and cluster analysis is performed on the business activity data to obtain multiple clusters.

[0083] For each cluster in multiple clusters, sequence pattern mining is performed on the cluster members to obtain the support of each activity pattern. Based on the support, target activity patterns are selected from the cluster members, and the dependencies and mapping strength coefficients between target activity patterns are identified.

[0084] Based on dependencies, each target activity pattern obtained from multiple clusters is used as a node to construct an initial template task tree;

[0085] For each node in the initial template task tree, determine the intra-cluster frequency of the node in its own cluster and the global frequency of the node in multiple clusters. Based on the intra-cluster frequency and the global frequency, determine the execution tendency weight between the node and its parent node.

[0086] The execution tendency weight and mapping strength coefficient between each pair of nodes are labeled at the corresponding positions in the initial template task tree to obtain the template task tree.

[0087] In this embodiment, historical system operation data consists of logs of all operational behaviors recorded by enterprise employees when using various business systems. Each record includes information such as user ID, operation time, operation module, operation type, input data, and output result. The session timeout rule is a data segmentation strategy. If the time interval between two consecutive operations exceeds a preset time threshold (e.g., 30 minutes), the two operations are considered to belong to different business sessions or activities. The historical system operation data is grouped by user ID, and for each user ID, the operations are sorted by timestamp to obtain an operation series. The time difference between two adjacent operations is calculated. If the time difference exceeds the time threshold set by the session timeout rule, the operation is segmented at the time difference. Ultimately, the user's operation series is segmented into multiple business activity data.

[0088] Furthermore, features are extracted from the business activity data. Feature types may include: operation sequence encoding (converting operation names into numerical sequences through encoding), module distribution (statistically counting the frequency of access to each module in each business activity data, represented in vector form), and key operation flags (whether a key operation is triggered, such as payment or issuance). The features extracted from the business activity data are input into a clustering algorithm. The type of clustering algorithm is not limited in this embodiment. Optionally, if the number of core business processes of the target enterprise can be obtained, K-Means can be used, requiring the number of clusters K to be specified in advance. If the types of business processes are unknown, or the number of business processes is not fixed, the DBSCAN algorithm can be used, without requiring the number of clusters K to be specified; the algorithm can automatically discover clusters of arbitrary shapes. The clustering algorithm automatically assigns business activity data to different clusters based on feature similarity. For example, all business activities related to expense reimbursement are assigned to cluster 1, and all business activities related to employee onboarding are assigned to cluster 2.

[0089] Furthermore, a sub-task tree is constructed for each cluster. Taking any cluster as an example, the process of constructing the sub-task tree for that cluster includes: pre-setting a minimum support threshold; the sequence pattern mining algorithm automatically identifies all activity patterns whose frequency (support) is greater than or equal to the minimum support threshold. Among the mined activity patterns, those with support below the minimum support threshold are discarded, and the remaining ones are taken as target activity patterns. A target activity pattern may contain one or more steps. The steps of all target activity patterns obtained for that cluster are traversed, and target steps without preceding steps are identified. Preceding steps are steps that precede and are adjacent to the target steps. If there is only one target step, it is taken as the root node of the sub-task tree for that cluster; if there are multiple target steps, each step is taken as a child node of the root node.

[0090] For each target step, its subsequent steps are determined from the target activity pattern. The subsequent steps are the steps that follow and are adjacent to the target step. The frequency of each subsequent step in all subsequent steps of the target step is determined. The subsequent steps that exceed the effective frequency threshold are selected as steps that have a dependency relationship with the target step. Each subsequent step with a dependency relationship is set as a node and set as a child node of the node where the target step is located.

[0091] For each child node of the target step, referring to the above embodiment, the subsequent steps of that child node are determined from the target activity pattern, and the frequency of each subsequent step in all subsequent steps of that child node is determined. Subsequent steps exceeding the effective frequency threshold are selected as steps that have a dependency relationship with that child node, and each dependent subsequent step is set as a node and designated as a child node of that child node (i.e., a grandchild node of the target step). This process is repeated until all steps in the target activity pattern have been traversed, resulting in the subtask tree of that cluster. Finally, the subtask trees of each cluster are integrated to obtain the initial template task tree.

[0092] For every two nodes with a parent-child relationship in the initial template task tree, denoted as parent node A and child node B, the output fields of parent node A and the input fields of child node B are extracted from the node feature library. The overlap ratio of the output and input fields is calculated, i.e., the proportion of overlapping fields to the number of input fields in child node B. Then, from the business activity data obtained by segmenting historical system operation data, samples of business activities where child node B is executed directly after parent node A is executed are selected. For each sample, the output data of parent node A and the execution result of child node B are recorded. Randomly constructed perturbation data for parent node A's output is added to the output data of parent node A, and child node B is re-executed to obtain the perturbed execution result. The change magnitude of the output result = |the previous output result of child node B - the next output result of child node B| / the previous output result of child node B. The change magnitude of the input data perturbation = the output perturbation data / the previous output result of parent node A; the single-sample sensitivity = the change magnitude of the output result of child node B / the change magnitude of the input data perturbation. The average of the single-sample sensitivities of all samples is calculated to obtain the correlation sensitivity. The mapping strength coefficient between parent node A and child node B is obtained by weighting and summing the field overlap ratio and association sensitivity based on predefined weights. Following the steps above, the mapping strength coefficient between every two parent and child nodes in the initial template task tree is determined.

[0093] For each node in the initial template task tree, the total number of times the step represented by that node appears in its original cluster is determined as the intra-cluster frequency, and the total number of times the step represented by that node appears in all clusters is determined as the global frequency. The result of a weighted sum of the intra-cluster frequency and the global frequency based on predefined weights is used as the execution preference weight, representing the degree of preference for choosing the path from the parent step to the child step, and is used for task recommendation. The execution preference weight of each node is labeled on the edge between that node and its parent node.

[0094] This embodiment achieves effective classification of business activities and accurately distinguishes different business scenarios through session segmentation and cluster analysis; it mines sequence patterns and filters target activity patterns to extract core business steps; it identifies dependency relationships and mapping strength coefficients to clarify the degree of correlation between business steps; it calculates execution tendency weights to quantify the execution preferences between steps, and the final constructed template task tree has both business logic and quantitative guidance significance, providing accurate process support for subsequent requirement decomposition and task execution.

[0095] One possible implementation of this application embodiment includes:

[0096] Create a test environment and generate multiple test cases containing different types of test tasks and network environment parameter combinations;

[0097] For each agent in the agent candidate pool, execute each test case to obtain the comprehensive test results of the agent for each task type in each network environment. The comprehensive test results include task performance indicators and environmental sensitivity indicators.

[0098] Based on the comprehensive test results of each agent and the preset scoring rules, a performance scoring matrix is ​​constructed for each agent. The rows of the performance scoring matrix represent the network environment, the columns represent the task type, and the elements represent the indicator scoring vector. The indicator scoring vector includes the indicator scores of each task performance indicator and each environment sensitivity indicator.

[0099] The comprehensive test results and performance score matrix of each agent in the agent candidate pool are stored as capability metadata.

[0100] In this embodiment, network environment parameters include bandwidth, latency, packet loss rate, jitter, etc. Task performance metrics may include: accuracy, processing time, and throughput. Accuracy indicates whether the output meets expectations; it can be calculated by predefining a standard output and determining the match between the standard output and the actual output. Processing time represents the difference between total processing time and network latency. Throughput represents the amount of data processed per unit time. Based on the target enterprise's business needs, enterprise managers can pre-set lower limits for task performance metrics for each task type, including a pass / fail threshold for each task's performance metrics.

[0101] Let any task type be the target task type, and any network environment be the target network environment. Each agent in the agent candidate pool executes the target test cases obtained from the combination of the target network environment and the target task type. Let any task performance metric be the target task performance metric. From the comprehensive test results obtained by each agent executing the target test cases, determine the optimal test result for the target task performance metric, and use this optimal test result as the upper limit of the range. Let the lower limit of the target task performance metric corresponding to the target task type be the lower limit of the range. The range formed by the lower limit and the upper limit of the range is the scoring benchmark range for the target task performance metric. Following the above process, the scoring benchmark range for each task performance metric of each test case can be obtained.

[0102] The process of selecting any intelligent agent as the target intelligent agent, obtaining the comprehensive test results of the target intelligent agent executing the target test cases, and determining the indicator scores of the task performance indicators and environmental sensitivity indicators based on the comprehensive test results includes: if the indicator test results of the target task performance indicators in the comprehensive test results are within the scoring benchmark range, then the first difference between the indicator test results of the target task performance indicators and the lower limit of the scoring benchmark range, and the second difference between the upper limit and the lower limit of the scoring benchmark range are calculated. The indicator score of the target intelligent agent on the target task performance indicators = first difference / second difference × 100.

[0103] If the target task performance index in the comprehensive test results is not within the corresponding scoring benchmark range, the target agent is determined to be unusable in executing the target test case. The elements of the target test case in the performance scoring matrix are set to null values. That is, if any task performance index in the test results does not meet the corresponding performance index lower limit, it indicates that the agent is unusable in that test case.

[0104] Environmental sensitivity metrics include stability coefficient, decay rate, and environmental adaptability. For each agent, its performance on the same task type under different network environments is analyzed to obtain an environmental sensitivity metric score. Taking the target agent and target task type as an example, the reciprocal of the standard deviation of the accuracy of the target agent in all network environments performing the target task type is calculated; the smaller the standard deviation, the higher the stability coefficient. For the decay rate, based on the throughput metric under different network environments, the difference between the best network environment (highest throughput) and the worst network environment (lowest throughput) is determined, and the ratio of the difference to the throughput of the best network environment is calculated as the percentage of performance degradation. The accuracy and throughput of the test results can be replaced by a comprehensive performance index obtained by weighted summation of multiple task performance metrics; this embodiment does not impose limitations. Environmental adaptability represents the comprehensive score and can be calculated as the weighted sum of the stability coefficient and (1 - decay rate). After obtaining the values ​​of the environmental sensitivity indicators of the target agent for the target task type, the product of the value of each environmental sensitivity indicator and 100 is used as the corresponding indicator score, thereby mapping each indicator score to the score range of 0-100.

[0105] By utilizing the process of scoring the target agent's performance indicators and environmental sensitivity indicators under any combination of network environment and task type, the target agent's performance indicators are obtained for each network environment and task type, thus constructing a performance scoring matrix for the target agent. Following the same process, a performance scoring matrix is ​​constructed for each agent in the candidate pool.

[0106] This embodiment designs multi-dimensional test cases to comprehensively cover different task types and network environments, ensuring the completeness of agent performance evaluation; by integrating test results and performance scoring matrices, the agent's capabilities are quantified and structured, intuitively presenting the agent's performance in different scenarios; stored as capability metadata, it provides objective and accurate data basis for subsequent agent selection, avoiding subjective selection bias.

[0107] One possible implementation of this application embodiment involves selecting agents from an agent candidate pool to form an agent list based on capability metadata and task types in the task sequence, including:

[0108] Obtain the current network environment and the lower limit of task performance metrics for each task type;

[0109] For each task in the task sequence, based on the capability metadata, select from the agent candidate pool a group of agents that can meet the lower limit of all task performance indicators of the task type in a network environment that matches the current network environment.

[0110] Based on the dependency constraints and the environmental sensitivity index of each agent, an agent is selected from the agent group of each task to form an agent list. The dependency constraints include that the agents selected for any two tasks that do not have a dependency relationship are not the same.

[0111] In this embodiment, any task in the task sequence is taken as the target task, and the task type and lower limit of the task performance index are determined. The parameters of the current network environment and each network environment in the test cases are compared, and the most matching network environment is determined as the best matching network environment. Taking the target agent as an example, the index score vector of the target agent on the task type of the target task under the best matching network environment is extracted from the performance score matrix of the target agent. If it is null, the target agent is unusable. All agents are traversed, and agents whose index score vector on the task type of the target task under the best matching network environment is not null are selected to form an agent group that can meet the lower limit of all task performance indices of the task type of the target task.

[0112] For the agent group targeting the target task, calculate the indicator score vector for each agent in the agent group corresponding to the best-matching network environment and the task type of the target task. Extract the indicator scores of the environment sensitivity indicators from the vectors, and calculate the average indicator score of each environment sensitivity indicator based on predefined weights. Select the agent with the highest average indicator score from the agent group targeting the target task and place it in the agent list.

[0113] If two tasks in a task sequence have a dependency relationship (are adjacent in the task sequence), they can choose the same agent to achieve resource reuse. If two tasks in a task sequence do not have a dependency relationship, the agents for the two tasks are different to avoid resource conflicts. For two tasks without a dependency relationship, the agent for the task whose agent is determined first remains unchanged, while the agent for the task whose agent is determined later automatically selects the agent with the second-highest average score to avoid conflict.

[0114] This embodiment selects agents based on the current network environment to ensure the adaptability of agents in the actual operating environment and avoid performance loss due to environmental incompatibility; it sets a lower limit for performance indicators to ensure that the selected agents have basic execution capabilities; it considers dependency constraints to avoid conflicts between agents with no dependencies, and selects more stable agents based on environmental sensitivity index scores.

[0115] One possible implementation of this application embodiment, determining the degree of current task execution deviation of the current intervention node, includes:

[0116] Determine the task type of each preceding task in the task sequence before the current intervention node, and determine the task type composition of the executed task sequence based on the task type of each preceding task;

[0117] The dynamic weight of each deviation dimension is determined based on the task type.

[0118] Obtain the current task execution result of the current intervention node, and calculate the degree of first task execution deviation of the current task execution result in each deviation dimension based on the current task execution result and the dynamic weight of each deviation dimension.

[0119] Determine the preceding and succeeding tasks of the current intervention node in the task sequence, and update the execution deviation of the first task based on the mapping strength coefficient between the preceding and succeeding tasks in the template task tree to obtain the execution deviation of the second task;

[0120] Determine whether the current intervention node is the first intervention node among several intervention nodes; if the current intervention node is the first intervention node, then use the degree of deviation of the second task execution as the degree of deviation of the current task execution.

[0121] If the current intervention node is not the first intervention node, the time-series cumulative deviation is calculated based on the task execution deviation of all intervention nodes between the current intervention node, and the second task execution deviation is updated based on the time-series cumulative deviation to obtain the current task execution deviation.

[0122] In this embodiment, all preceding tasks before the current intervention node are traversed, and the task type of each task is recorded to form a list of executed task types. The proportion of each task type in the preceding tasks is recorded, such as data collection tasks accounting for 60% and data verification tasks accounting for 40%, thus obtaining the task type composition distribution.

[0123] Multiple deviation dimensions are predefined based on actual business scenarios. Optional deviation dimensions include: accuracy deviation (the degree of matching between the actual output result and the expected result), completeness deviation (the proportion of missing fields in the output data), and format deviation (the degree of non-compliance between the output data format and the format requirements). Basic weights (the sum of weights is 1) are preset for each deviation dimension for each task type. For example, the format dimension has a higher weight in data acquisition tasks, while the accuracy dimension has a higher weight in data validation tasks. The specific distribution of basic weights for each dimension under each task type is not limited in this embodiment.

[0124] Based on the task type composition, the basic weights of each deviation dimension are weighted and summed to obtain the dynamic weight of each deviation dimension. For any deviation dimension k, the proportion of each task type in the task type composition and the basic weight of that task type on dimension k are calculated as the deviation of that task type. The sum of the deviations of all task types in the task type composition is calculated as the dynamic weight of deviation dimension k. Referring to the process for determining the dynamic weight of dimension k, the dynamic weight of each deviation dimension is determined.

[0125] The current task execution result of the current intervention node is collected. The current task execution result is the actual output result, which is compared with the preset expected result. The expected result can be the historical output of the current intervention node in the template task tree, or it can be the expected result predicted by a pre-trained neural network. This embodiment does not limit this. For each deviation dimension, the deviation value between the actual result and the expected result in the current task execution result is calculated. Accuracy deviation value = 1 - number of matching fields / total number of fields. When the output result is numerical data, the accuracy deviation value = 1 - actual result / expected result. Integrity deviation value = number of missing fields / total number of fields. Format deviation value = number of fields with format errors / total number of fields. The obtained deviation values ​​of each dimension are weighted and summed according to the determined dynamic weights to obtain the first task execution deviation degree.

[0126] In the task sequence, the preceding task of the current intervention node is the immediately preceding task, and the following task is the immediately following task. From the template task tree, the mapping strength coefficient between the preceding and following task nodes is found, representing the degree of influence of the preceding task's output on the following task's execution result. This mapping strength coefficient is used as the deviation propagation coefficient, representing the proportion of the preceding task's deviation on the following task. The product of the mapping strength coefficient and the first task's execution deviation is calculated as the second task's execution deviation.

[0127] If the current intervention node is not the first intervention node, then the cumulative effect of the bias of historical intervention nodes is considered.

[0128] Where n is the number of all intervention nodes before the current intervention node, the attenuation coefficient represents the degree of deviation propagation between every two adjacent nodes in the task sequence, and m is the interval step number, representing the difference in task sequence number between every two adjacent intervention nodes.

[0129] Furthermore, the current task execution deviation level = the second task execution deviation level + the time-series cumulative deviation. Specifically, based on actual business needs, the product of the time-series cumulative deviation and the adjustment coefficient can be used as the updated time-series cumulative deviation. Then, the sum of the second task execution deviation level and the updated time-series cumulative deviation is calculated as the current task execution deviation level. This allows for the control of the impact weight of the time-series cumulative deviation through the adjustment coefficient, preventing excessive amplification.

[0130] This embodiment determines dynamic weights based on the types of preceding tasks, making deviation assessment more aligned with business scenarios and avoiding the limitations of a single weight. It also considers the impact of inter-task correlations on deviations by combining mapping strength coefficients, thus improving the comprehensiveness of deviation assessment. Furthermore, it introduces time-series cumulative deviations to take into account the cumulative effect of historical deviations, resulting in a more accurate calculation of the degree of task execution deviation.

[0131] One possible implementation of this application embodiment involves adjusting the execution strategy of subsequent tasks of an intervention node in a task sequence, including:

[0132] Starting from the first task in the task sequence, obtain the intermediate task execution results after each task is completed, and calculate the degree of intermediate task execution deviation of the intermediate task execution results in each deviation dimension;

[0133] The intermediate task execution deviation obtained after each task is completed is compared with the preset alarm threshold, and the task corresponding to the intermediate task execution deviation that exceeds the preset alarm threshold is taken as the target task.

[0134] Adjust the agents corresponding to the target task in the agent list, and execute the target task and subsequent tasks in the task sequence according to the adjusted agent list.

[0135] In this embodiment, the preset alarm threshold can be set based on practical experience. Referring to the method described above for calculating the first task execution deviation of the intervention node, the intermediate task execution results after each task is completed are calculated. The task whose first intermediate task execution deviation exceeds the preset alarm threshold is taken as the target task. For the target task, from the agent group corresponding to the target task during the agent list construction process, the agent that satisfies the dependency constraint relationship and has the highest average score, excluding agents already matched with the target task in the agent list, is selected as the adjusted agent, thereby completing the update of the agent list.

[0136] This embodiment achieves dynamic tracking of the task execution process by monitoring intermediate results and deviation levels in real time. It can quickly locate target tasks that exceed the threshold, adjust the corresponding intelligent agent for the target task in a timely manner, address task deviation issues in a targeted manner, and prevent deviations from being transmitted to subsequent tasks. It re-executes the target task and subsequent tasks to ensure that business processing returns to the correct track, effectively reducing the impact of deviations on the overall business results and ensuring the quality of business processing.

[0137] One possible implementation of this application embodiment is to determine the lowest-level node of each task in the template task tree when the task sequence is completed.

[0138] Determine the child nodes of the lowest-level node in the template task tree, and arrange each child node in descending order of execution tendency weight with the lowest-level node to obtain the recommended task sequence;

[0139] The recommended task sequence is fed back to the target personnel so that the corresponding task can be executed based on the target personnel's selection of the recommended task sequence.

[0140] In this embodiment, the lowest-level node is the node corresponding to the last task in the task sequence in the template task tree. Combining the child nodes and execution preference weights of this node, a recommended task sequence that conforms to business habits is generated.

[0141] This embodiment generates recommended task sequences based on template task trees and execution tendency weights, which aligns with the enterprise's historical business execution habits, improving the rationality and practicality of the recommendations; the recommended sequences are fed back to the target personnel, providing clear guidance for subsequent business processing and reducing personnel decision-making costs; and it supports personnel's independent selection, balancing the flexibility of intelligent recommendations with manual intervention.

[0142] This application provides an electronic device, such as... Figure 2 As shown, Figure 2 The illustrated electronic device 200 includes a processor 201 and a memory 203. The processor 201 and the memory 203 are connected, for example, via a bus 202. Optionally, the electronic device 200 may also include a transceiver 204. It should be noted that in practical applications, the transceiver 204 is not limited to one type, and the structure of this electronic device 200 does not constitute a limitation on the embodiments of this application.

[0143] Processor 201 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 201 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0144] Bus 202 may include a pathway for transmitting information between the aforementioned components. Bus 202 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 202 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 2 The symbol is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0145] The memory 203 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0146] The memory 203 is used to store the application code that executes the solution of this application, and its execution is controlled by the processor 201. The processor 201 is used to execute the application code stored in the memory 203 to implement the content shown in the aforementioned business processing method embodiment.

[0147] Figure 2 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0148] This application provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the content shown in the aforementioned business processing method embodiments.

[0149] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0150] This application provides a computer program product, including a computer program, which, when executed by a processor, implements the content shown in the aforementioned business processing method embodiment.

[0151] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A business processing method, characterized in that, include: When an access request is received from a target enterprise, the pre-built intelligent agent platform is linked with the target enterprise's enterprise system to obtain historical system operation data of the enterprise system, and a template task tree is constructed based on the historical system operation data. Receive user requirements uploaded by target personnel of the target enterprise through the enterprise system, break down the user requirements into a task sequence based on the template task tree, and determine the task type of each task in the task sequence; Obtain capability metadata of the agent candidate pool of the agent platform, select agents from the agent candidate pool based on the capability metadata and the task type of the task sequence to form an agent list, and assign the task sequence to the agent list for execution; Based on the depth of the task sequence, several intervention nodes are determined, and the degree of task execution deviation is determined at each intervention node. When the degree of task execution deviation of any intervention node exceeds a preset intervention threshold, the execution strategy of subsequent tasks of the intervention node in the task sequence is adjusted. The construction of the template task tree based on the historical system operation data includes: Based on session timeout rules, the historical system operation data is segmented to obtain business activity data, and the business activity data is then subjected to cluster analysis to obtain multiple clusters. For each of the multiple clusters, sequence pattern mining is performed on the cluster members to obtain the support of each activity pattern. Based on the support, target activity patterns are selected from the cluster members, and the dependency relationships and mapping strength coefficients between the target activity patterns are identified. Based on the dependencies, each target activity pattern obtained from the multiple clusters is used as a node to construct an initial template task tree; For each node in the initial template task tree, determine the intra-cluster frequency of the node in its cluster and the global frequency of the node in the multiple clusters. Based on the intra-cluster frequency and the global frequency, determine the execution tendency weight between the node and its parent node. The execution tendency weight and mapping strength coefficient between each pair of nodes are labeled at the corresponding positions in the initial template task tree to obtain the template task tree.

2. The business processing method according to claim 1, characterized in that, The method further includes: Create a test environment and generate multiple test cases containing different types of test tasks and network environment parameter combinations; For each agent in the agent candidate pool, each test case is executed to obtain the comprehensive test results of the agent for each task type in each network environment. The comprehensive test results include task performance indicators and environmental sensitivity indicators. Based on the comprehensive test results of each agent and the preset scoring rules, a performance scoring matrix is ​​constructed for each agent. The rows of the performance scoring matrix represent the network environment, the columns represent the task type, and the elements represent the indicator scoring vector. The indicator scoring vector includes the indicator scores of each task performance indicator and each environment sensitivity indicator. The comprehensive test results and performance score matrix of each agent in the agent candidate pool are stored as the capability metadata.

3. The business processing method according to claim 2, characterized in that, The process of selecting agents from the agent candidate pool to form an agent list based on the capability metadata and the task type of the task sequence includes: Obtain the current network environment and the lower limit of task performance metrics for each task type; For each task in the task sequence, based on the capability metadata, select from the agent candidate pool a group of agents that can meet the lower limit of all task performance indicators of the task type under the network environment that matches the current network environment. Based on the dependency constraints and the environmental sensitivity index of each agent, an agent is selected from the agent group of each task to form the agent list. The dependency constraints include the fact that the agents selected for any two tasks that do not have a dependency relationship are not the same.

4. The business processing method according to claim 1, characterized in that, Determine the degree of deviation in the current task execution of the current intervention node, including: Determine the task type of each preceding task in the task sequence before the current intervention node, and determine the task type composition of the executed task sequence based on the task type of each preceding task; The dynamic weights for each deviation dimension are determined based on the task type. Obtain the current task execution result of the current intervention node, and calculate the degree of first task execution deviation of the current task execution result on each deviation dimension based on the current task execution result and the dynamic weight of each deviation dimension. Determine the preceding and following tasks of the current intervention node in the task sequence, and update the first task execution deviation degree based on the mapping strength coefficient between the preceding and following tasks in the template task tree to obtain the second task execution deviation degree; Determine whether the current intervention node is the first intervention node among the plurality of intervention nodes; if the current intervention node is the first intervention node, then use the degree of deviation of the second task execution as the degree of deviation of the current task execution; If the current intervention node is not the first intervention node, then the time-series cumulative deviation is calculated based on the task execution deviation of all intervention nodes between the current intervention node, and the second task execution deviation is updated based on the time-series cumulative deviation to obtain the current task execution deviation.

5. The business processing method according to claim 1, characterized in that, The adjustment of the execution strategy for subsequent tasks of the intervention node in the task sequence includes: Starting from the first task in the task sequence, the intermediate task execution results after each task is completed are obtained sequentially, and the degree of intermediate task execution deviation of the intermediate task execution results in each deviation dimension is calculated. The intermediate task execution deviation obtained after each task is completed is compared with the preset alarm threshold, and the task corresponding to the intermediate task execution deviation that exceeds the preset alarm threshold is taken as the target task. Adjust the agent corresponding to the target task in the agent list, and execute the target task and subsequent tasks in the task sequence according to the adjusted agent list.

6. The business processing method according to claim 1, characterized in that, The method further includes: When the task sequence is completed, determine the lowest-level node of each task in the task sequence in the template task tree; Determine the child nodes of the lowest-level node in the template task tree, and arrange each child node in descending order of execution tendency weight with the lowest-level node to obtain the recommended task sequence; The recommended task sequence is fed back to the target personnel so that the corresponding task can be executed based on the target personnel's selection of the recommended task sequence.

7. An electronic device, characterized in that, include: At least one processor; Memory; At least one application, wherein the at least one application is stored in memory and configured to be executed by at least one processor, said at least one application being configured to: perform the business processing method according to any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed in the computer, the computer is instructed to perform the business processing method according to any one of claims 1-6.

9. A computer program product, characterized in that, It includes a computer program, which, when executed by a processor, implements the steps of the business processing method according to any one of claims 1-6.

Images