Complex business processing method and system based on distributed intelligent agent system

Abstract

The embodiment of the application provides a complex business processing method and system based on a distributed intelligent agent system, which comprises the following steps: obtaining business data and business processing requirements, analyzing and processing the business processing requirements to obtain a plurality of subtasks, assigning types to the subtasks according to the attributes of the business data to obtain corresponding agent types, determining a first distributed agent node based on a load resource prediction model and the load demand of the subtask associated with a single-step agent, and using the first distributed agent node to process the subtask of the single-step agent to obtain a first result, screening a plurality of candidate node combinations of the subtask of a sequential agent according to the load rate and the communication delay time of each distributed agent node that has not been selected, determining a second node combination based on the matching degree between each candidate node in each candidate node combination, using the second node combination to process the subtask of the sequential agent to obtain a second result, and fusing the first and second results to obtain a final result. The scheme can realize efficient processing of business data.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of electric digital data processing, and in particular to a complex business processing method and system based on a distributed intelligent agent system. BACKGROUND

[0002] With the rapid evolution of artificial intelligence technology, the complexity of its business processing tasks in the fields of finance, medicine, intelligent manufacturing, etc. is experiencing exponential growth. Current artificial intelligence systems can effectively handle simple business scenarios with high standardization and relatively single processes through lightweight model architectures and distributed computing frameworks. A distributed artificial intelligence system is composed of multiple distributed agent nodes, each of which has autonomous decision-making capabilities, local resources (computing / storage / communication), and task execution logic, and collaborates with other nodes through a decentralized communication protocol to complete complex business goals.

[0003] In related technologies, the processing efficiency for single or relatively simple business tasks can basically meet user needs and costs are controllable, such as image classification, text translation, etc. tasks, which can achieve millisecond-level response through distributed computing framework and controllable hardware costs, but the processing efficiency for complex business problems is low and cannot meet the processing needs. SUMMARY

[0004] Embodiments of the present application provide a complex business processing method and system based on a distributed intelligent agent system, which solves the problem of low processing efficiency for complex business problems in the prior art, which cannot meet the processing needs, and can achieve efficient processing of business data, while ensuring processing efficiency and reducing processing costs.

[0005] In a first aspect, embodiments of the present application provide a complex business processing method based on a distributed intelligent agent system, comprising:

[0006] Obtaining business data to be processed and corresponding business processing requirements, performing analysis and processing on the business processing requirements to obtain a plurality of subtasks, and assigning a proxy type to each subtask according to the attributes of the business data to obtain a corresponding proxy type, the proxy type including a single-step proxy type and a sequential proxy type;

[0007] Determine a first distributed agent node based on a load resource prediction model and the load demand associated with the single-step proxy type subtask in a plurality of distributed agent nodes, and send the single-step proxy type subtask to the first distributed agent node for processing to obtain a first processing result;

[0008] screening a plurality of candidate agent node combinations of the sequential agent type subtask from the unselected distributed agent nodes according to the load rate and the communication delay time, determining a second agent node combination based on the matching degree between each candidate agent node in each candidate agent node combination;

[0009] sending the sequential agent type subtask to the second agent node combination for processing to obtain a second processing result, and fusing the first processing result and the second processing result to obtain a final processing result of the service data.

[0010] Optionally, the first distributed agent node is determined in the plurality of distributed agent nodes based on the load resource prediction model and the load demand associated with the single-step agent type subtask, comprising:

[0011] inputting the current comprehensive load information of the plurality of distributed agent nodes and the time period into the load resource prediction model to obtain load resource prediction information of each distributed agent node;

[0012] determining the first distributed agent node corresponding to the single-step agent type subtask according to the load resource prediction information of each distributed agent node and the load demand.

[0013] Optionally, the first distributed agent node corresponding to the single-step agent type subtask is determined according to the load resource prediction information of each distributed agent node and the load demand, comprising:

[0014] in the case that there are a plurality of single-step agent type subtasks, comparing the load demand associated with each single-step agent type subtask, and performing priority sorting according to the comparison result to obtain an allocation priority of each single-step agent type subtask;

[0015] comparing the load demand of the subtask with the lowest allocation priority with the load resource prediction information of each distributed agent node to obtain a candidate agent node sequence combination, and determining the first distributed agent node of each single-step agent type subtask according to the allocation priority and the candidate agent node sequence combination.

[0016] Optionally, the sequential agent type subtask comprises a plurality of sequentially executed minimum division tasks, and the plurality of candidate agent node combinations of the sequential agent type subtask are screened from the unselected distributed agent nodes according to the load rate and the communication delay time, comprising:

[0017] According to each of the unselected distributed agent node, a plurality of candidate agent nodes with a load rate lower than a first threshold and a communication delay time lower than a second threshold are screened out, and a plurality of candidate agent node combinations are obtained by combining processing of the plurality of candidate agent nodes according to the minimum number of divided tasks.

[0018] Optionally, the determining a second agent node combination based on matching degrees between each candidate agent node in each of the candidate agent node combinations comprises:

[0019] Each of the candidate agent nodes in each of the candidate agent node combinations is sequentially arranged to obtain a plurality of sequential agent node sequences corresponding to the candidate agent node combinations.

[0020] The second agent node combination is determined according to the matching degrees between each of the candidate agent nodes and the plurality of sequential agent node sequences of each of the candidate agent node combinations.

[0021] Optionally, the determining a second agent node combination based on matching degrees between each candidate agent node in each of the candidate agent node combinations comprises:

[0022] The matching degrees between sequentially connected candidate agent nodes in the plurality of sequential agent node sequences of the candidate agent node combination are superimposed to obtain a comprehensive matching degree, and a sequential agent node sequence with the highest comprehensive matching degree is determined as a to-be-determined sequential combination of the corresponding candidate agent node combination.

[0023] The to-be-determined sequential combinations are compared according to the comprehensive matching degrees of each of the to-be-determined sequential combinations, and a to-be-determined sequential combination with the highest comprehensive matching degree is determined as the second agent node combination.

[0024] Optionally, the fusing processing of the first processing result and the second processing result to obtain a final processing result of the business data comprises:

[0025] A first weight of the first processing result is determined according to a processing time length corresponding to the first processing result, and a second weight of the second processing result is determined according to a historical processing success rate associated with the second agent node combination corresponding to the second processing result.

[0026] The first processing result and the second processing result are fused and calculated according to the first weight and the second weight to obtain the final processing result of the business data.

[0027] In a second aspect, the embodiments of the present application further provide a complex business processing system based on a distributed intelligent agent system, comprising:

[0028] An acquisition module is configured to acquire service data to be processed and corresponding service processing requirements.

[0029] A task analysis module is configured to analyze and process the service processing requirements to obtain a plurality of subtasks.

[0030] A type allocation module is configured to allocate a proxy type to each of the subtasks according to attributes of the service data to obtain a corresponding proxy type, wherein the proxy type includes a single-step proxy type and a sequential proxy type.

[0031] A proxy node determination module is configured to determine a first distributed proxy node among a plurality of distributed proxy nodes based on a load resource prediction model and a load requirement associated with a subtask of the single-step proxy type.

[0032] A task sending module is configured to send the subtask of the single-step proxy type to the first distributed proxy node for processing to obtain a first processing result.

[0033] A node combination determination module is configured to filter a plurality of candidate proxy node combinations of the subtask of the sequential proxy type according to load rates and communication delay times of each of the distributed proxy nodes that have not been selected, and determine a second proxy node combination based on matching degrees between each of the candidate proxy nodes in each of the candidate proxy node combinations.

[0034] The task sending module is further configured to send the subtask of the sequential proxy type to the second proxy node combination for processing to obtain a second processing result.

[0035] A fusion processing module is configured to fuse the first processing result and the second processing result to obtain a final processing result of the service data.

[0036] In a third aspect, an embodiment of the present application further provides a complex service processing device based on a distributed intelligent proxy system, which comprises:

[0037] One or more processors;

[0038] A storage device configured to store one or more programs,

[0039] When the one or more programs are executed by the one or more processors, the one or more processors implement the complex service processing method based on the distributed intelligent proxy system according to the embodiments of the present application.

[0040] In a fourth aspect, an embodiment of the present application further provides a storage medium storing computer executable instructions, which, when executed by a computer processor, are used to perform the complex service processing method based on the distributed intelligent proxy system according to the embodiments of the present application.

[0041] In the embodiment of the application, by acquiring the to-be-processed service data and the corresponding service processing demand, the service processing demand is analyzed and processed to obtain a plurality of sub-tasks, the attribute of the service data is used to allocate a proxy type to each sub-task to obtain a corresponding proxy type, the proxy type includes a single-step proxy type and a sequential proxy type, a first distributed proxy node is determined in a plurality of distributed proxy nodes based on a load resource prediction model and a load demand associated with the sub-task of the single-step proxy type, and the sub-task of the single-step proxy type is sent to the first distributed proxy node for processing to obtain a first processing result, a plurality of candidate proxy node combinations of the sub-task of the sequential proxy type are screened out according to the load rate and the communication delay time of each distributed proxy node that is not selected, a second proxy node combination is determined based on the matching degree between each candidate proxy node in each candidate proxy node combination, the sub-task of the sequential proxy type is sent to the second proxy node combination for processing to obtain a second processing result, and the first processing result and the second processing result are fused to obtain a final processing result of the service data. The scheme solves the problem that the processing efficiency of the prior art for complex service problems is low and cannot meet the processing demand, and can realize efficient processing of service data while ensuring processing efficiency and reducing processing cost. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 A flowchart of a complex service processing method based on a distributed intelligent proxy system provided by the embodiment of the application;

[0043] Figure 2 A flowchart of a complex service processing method based on a distributed intelligent proxy system provided by the embodiment of the application, which includes determining a first distributed proxy node;

[0044] Figure 3 A flowchart of a complex service processing method based on a distributed intelligent proxy system provided by the embodiment of the application, which includes determining a plurality of candidate proxy node combinations;

[0045] Figure 4 A flowchart of a complex service processing method based on a distributed intelligent proxy system provided by the embodiment of the application, which includes determining a second proxy node combination;

[0046] Figure 5 A flowchart of a complex service processing method based on a distributed intelligent proxy system provided by the embodiment of the application, which includes determining a final processing result;

[0047] Figure 6 A module structure block diagram of a complex service processing system based on a distributed intelligent proxy system provided by the embodiment of the application;

[0048] Figure 7 A structural schematic diagram of a complex business processing device based on a distributed intelligent agent system is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0049] The embodiments of the present application will be further described below in conjunction with the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the embodiments of the present application, but not to limit the embodiments of the present application. In addition, it should be noted that, for the convenience of description, only the parts related to the embodiments of the present application are shown in the drawings, but not all the structures.

[0050] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be exchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are usually a class, and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the front and rear associated objects are in a "or" relationship.

[0051] The complex business processing method based on the distributed intelligent agent system provided by the embodiments of the present application can be applied to the processing scene of complex business such as high-frequency quantitative transaction decision chain, multi-modal financial risk control, supply chain intelligent scheduling, etc. The complex business processing method based on the distributed intelligent agent system provided by the embodiments of the present application, the execution subject of each step is a server.

[0052] Figure 1 A flowchart of a complex business processing method based on a distributed intelligent agent system provided for the embodiments of the present application is shown as Figure 1 , and specifically includes:

[0053] Step S101, obtaining business data to be processed and corresponding business processing requirements, performing analysis and processing on the business processing requirements to obtain a plurality of subtasks, and performing agent type distribution on each subtask according to the attribute of the business data to obtain a corresponding agent type, wherein the agent type includes a single-step agent type and a sequential agent type.

[0054] The business data is used to represent related data of a to-be-processed business. The business processing requirement can be a target to be achieved by processing the business data. The business processing requirement can be a piece of text description, for example, the business processing requirement is to analyze today's order data of an e-commerce platform in real time to identify high-value users and push coupons. A plurality of sub-tasks are obtained by analyzing and processing the business processing requirement. The sub-tasks can be various tasks required to achieve the business processing requirement. The attributes of the business data can be used to assign a proxy type to each sub-task. The proxy type refers to a task processing mode divided according to the structure and execution logic of the sub-task. The proxy type can include a single-step proxy type and a sequential proxy type. The single-step proxy type refers to a type of sub-task that can be completed in a single operation without relying on other steps or intermediate results. The sequential proxy type refers to a type of sub-task that requires multiple steps to cooperate, or the output of the previous step as the input of the next step. In an embodiment, after obtaining the to-be-processed business data and the corresponding business processing requirement, the business processing requirement is subjected to natural language processing to obtain a plurality of corresponding sub-tasks. Whether there is a dependency relationship between each sub-task is determined according to the dependency relationship between each sub-data in the business data. The sub-tasks with a dependency relationship are determined as the sequential proxy type, and the sub-tasks without a dependency relationship are determined as the single-step proxy type. For example, the business processing requirement is to judge financial risk, and the business data includes a transaction amount table, a transaction IP address table, a user login information table, a user browsing information table, and a payment information table. There is a dependency relationship between the user login information table and the user browsing information table, and there is a dependency relationship between the user browsing information table and the payment information table. The plurality of corresponding sub-tasks obtained by subjecting the business processing requirement to natural language processing are detecting transaction amount, detecting transaction IP address, analyzing user login behavior, analyzing user browsing behavior, and analyzing user payment behavior. Then, detecting transaction amount and detecting transaction IP address are single-step proxy type sub-tasks, and analyzing user login behavior→analyzing user browsing behavior→analyzing user payment behavior are sequential proxy type sub-tasks.

[0055] In step S102, a first distributed proxy node is determined in the plurality of distributed proxy nodes based on the load resource prediction model and the load demand associated with the single-step proxy type sub-task, and the single-step proxy type sub-task is sent to the first distributed proxy node for processing to obtain a first processing result.

[0056] The load resource prediction model can be a model for predicting the load resource of the distributed agent node in the next period of time. The load demand refers to the demand of the distributed agent node for the load resource of the single-step agent type subtask. The load resource prediction model and the load demand can be used to determine the distributed agent node that processes the corresponding single-step agent type subtask among the plurality of distributed agent nodes. The distributed agent node refers to an intelligent computing unit deployed in a distributed system and having autonomous decision-making capability, which realizes complex business processing through a cooperative working mechanism. Each distributed agent node can operate independently or cooperate with other nodes in the system through a communication protocol to form a unified processing network. In an embodiment, a manner of determining the first distributed agent node can be that the current comprehensive load information of the plurality of distributed agent nodes and the time period are input into the load resource prediction model to obtain the load resource prediction information of each distributed agent node, the load resource prediction information of each distributed agent node is compared with the load demand associated with the single-step agent type subtask, and the optimal distributed agent node is determined according to the comparison result and the first distributed agent node for the single-step agent type subtask is determined.

[0057] In step S103, a plurality of candidate agent node combinations for the sequential agent type subtask are screened according to the load rate and the communication delay time of each unselected distributed agent node, and a second agent node combination is determined based on the matching degree between each candidate agent node in each candidate agent node combination.

[0058] The load rate is used to represent the ratio of the actual workload of the distributed agent node in processing a request, data or computing task to the theoretical maximum carrying capacity. The communication delay time can be the delay time of the distributed agent node in analyzing, calculating or forwarding a data packet. The load rate and the communication delay time of each unselected distributed agent node can be used to screen a plurality of candidate agent node combinations for the sequential agent type subtask. The candidate agent node combination is used to represent the combination of each distributed agent node to be selected to process the sequential agent type subtask. In an embodiment, the sequential agent type subtask includes a plurality of sequentially executed minimum division tasks. A manner of screening the plurality of candidate agent node combinations can be that the real-time score is calculated according to the load rate and the communication delay time of each unselected distributed agent node, a plurality of candidate agent nodes with a real-time score higher than a preset threshold are screened, and the plurality of candidate agent nodes are combined according to the number of minimum division tasks to obtain the plurality of candidate agent node combinations. In an embodiment, a manner of determining the second agent node combination can be that the average value of the matching degree between each candidate agent node in each candidate agent node combination is calculated, and the candidate agent node combination with the highest average value is determined as the second agent node combination.

[0059] Step S104, sending the subtask of the sequential agent type to the second agent node combination for processing to obtain a second processing result, fusing the first processing result and the second processing result to obtain a final processing result of the service data.

[0060] In one embodiment, a fusing processing manner can be that a first weight is determined according to a processing time length and a historical success rate of the first distributed agent node corresponding to the first processing result, a second weight is determined according to a processing time length and a historical processing success rate of the second agent node combination corresponding to the second processing result, and the first processing result and the second processing result are fused and calculated according to the first weight and the second weight to obtain the final processing result of the service data. By assigning the corresponding weights to each processing result according to the processing time length and the historical processing success rate, and fusing and calculating to obtain the final processing result of the service data, the rationality and the accuracy of the final processing result can be improved.

[0061] As known from the above, by obtaining the service data to be processed and the corresponding service processing demand, the service processing demand is analyzed and processed to obtain a plurality of subtasks, each subtask is assigned to a corresponding agent type according to the attribute of the service data, the agent type includes a single-step agent type and a sequential agent type, a first distributed agent node is determined in the plurality of distributed agent nodes based on a load resource prediction model and a load demand associated with the subtask of the single-step agent type, the subtask of the single-step agent type is sent to the first distributed agent node for processing to obtain a first processing result, a plurality of candidate agent node combinations of the subtask of the sequential agent type are screened out according to the load rate and the communication delay time of each distributed agent node which is not selected, a second agent node combination is determined based on the matching degree between each candidate agent node in each candidate agent node combination, the subtask of the sequential agent type is sent to the second agent node combination for processing to obtain a second processing result, and the first processing result and the second processing result are fused to obtain a final processing result of the service data. The present scheme solves the problem that the processing efficiency of the prior art for complex service problems is low and cannot meet the processing demand, and can realize efficient processing of service data while ensuring processing efficiency and reducing processing cost.

[0062] Figure 2 A flowchart of a complex service processing method based on a distributed intelligent agent system including determining a first distributed agent node provided by an embodiment of the present application is shown in FIG. 1, and specifically includes the following steps. Figure 2 ​

[0063] In step S201, the service data to be processed and the corresponding service processing requirement are acquired, the service processing requirement is parsed to obtain a plurality of subtasks, and the proxy types of the subtasks are allocated according to the attributes of the service data to obtain corresponding proxy types, wherein the proxy types include a single-step proxy type and a sequential proxy type.

[0064] In step S202, the comprehensive load information of the plurality of distributed proxy nodes and the time period in which the distributed proxy nodes are located are input into a load resource prediction model to obtain load resource prediction information of each distributed proxy node, a first distributed proxy node of the subtask of the single-step proxy type is determined according to the load resource prediction information of each distributed proxy node and the load requirement, and the subtask of the single-step proxy type is sent to the first distributed proxy node for processing to obtain a first processing result.

[0065] The comprehensive load information can be comprehensive information of each load resource of the distributed proxy node at present, such as CPU occupancy, queue length, memory usage, and the like. The comprehensive load information can also be a comprehensive load score evaluated by each load resource of the distributed proxy node at present. The time period in which the distributed proxy nodes are located is used to represent the time period in which the plurality of distributed proxy nodes are currently located, such as a weekday time period, a rest day time period, a holiday time period, and the like. By inputting the comprehensive load information of the plurality of distributed proxy nodes at present and the time period in which the distributed proxy nodes are located into the load resource prediction model, the load resource prediction information of each distributed proxy node can be obtained. The load resource prediction information is used to represent the predicted load resource information of the distributed proxy node in the next time period. The first distributed proxy node can be determined by using the load resource prediction information and the load requirement associated with the subtask of the single-step proxy type, and the first processing result can be obtained by sending the subtask of the single-step proxy type to the first distributed proxy node for processing.

[0066] Optionally, a manner of determining the first distributed agent node can be that, in the case that multiple single-step agent type subtasks exist, the load requirements of each single-step agent type subtask are compared, the allocation priority of each single-step agent type subtask is obtained according to the comparison result, the load requirement of the subtask with the lowest allocation priority is compared with the load resource prediction information of each distributed agent node to obtain an optional agent node sequence combination, and the first distributed agent node of each single-step agent type subtask is determined according to the allocation priority and the optional agent node sequence combination. For example, three single-step agent type subtasks are subtask 1, subtask 2 and subtask 3, the load requirement of subtask 1 is CPU occupancy less than 30% and memory usage less than 40%, the load requirement of subtask 2 is CPU occupancy less than 40% and memory usage less than 50%, and the load requirement of subtask 3 is CPU occupancy less than 45% and memory usage less than 50%. The allocation priority of each single-step agent type subtask is subtask 1>subtask 2>subtask 3, and six distributed agent nodes are node a, node b, node c, node d, node e and node f. The load resource prediction information of node a is CPU occupancy 25% and memory usage 35%, the load resource prediction information of node b is CPU occupancy 20% and memory usage 25%, the load resource prediction information of node c is CPU occupancy 35% and memory usage 45%, the load resource prediction information of node d is CPU occupancy 35% and memory usage 40%, the load resource prediction information of node e is CPU occupancy 50% and memory usage 40%, and the load resource prediction information of node d is CPU occupancy 50% and memory usage 55%. The distributed agent nodes meeting the load requirement of subtask 3 are node a, node b, node c and node d. The optional agent node sequence combination is obtained by sorting the distributed agent nodes meeting the requirement according to the degree of meeting the requirement, and the optional agent node sequence combination is node b>node a>node d>node c. The first distributed agent node of subtask 1 is node b, the first distributed agent node of subtask 2 is node a, and the first distributed agent node of subtask 3 is node d according to the allocation priority in the optional agent node sequence combination.

[0067] In another embodiment, a manner of determining the first distributed agent node can be that, the load resource prediction information closest to the load requirement of the single-step agent type subtask is determined from the load resource prediction information of each distributed agent node, and the closest load resource prediction information is determined as the first distributed agent node of the single-step agent type subtask.

[0068] Step S203, screening out a plurality of candidate agent node combinations of the sequential agent type subtask according to the load rate and the communication delay time of each unselected distributed agent node, and determining a second agent node combination based on the matching degree between each candidate agent node in each candidate agent node combination.

[0069] Step S204, sending the sequential agent type subtask to the second agent node combination for processing to obtain a second processing result, and fusing the first processing result and the second processing result to obtain a final processing result of the service data.

[0070] As can be seen from the above, by inputting the current comprehensive load information of the plurality of distributed agent nodes and the time period into the load resource prediction model to obtain the load resource prediction information of each distributed agent node, a first distributed agent node corresponding to the single-step agent type subtask is determined according to the load resource prediction information of each distributed agent node and the load demand, and the single-step agent type subtask is sent to the first distributed agent node for processing to obtain a first processing result. The present scheme can ensure load balancing and improve task processing efficiency by predicting the load resources of each distributed agent node and determining the corresponding distributed agent node according to the load demand of the single-step agent type subtask.

[0071] Figure 3 A flowchart of a complex service processing method based on a distributed intelligent agent system containing determining a plurality of candidate agent node combinations provided by an embodiment of the present application is shown in FIG. 1, which specifically includes the following steps. Figure 3 As shown in FIG. 1, the method comprises the following steps.

[0072] Step S301, obtaining service data to be processed and corresponding service processing demand, performing analysis and processing on the service processing demand to obtain a plurality of subtasks, and performing agent type allocation on each subtask according to the attribute of the service data to obtain a corresponding agent type, wherein the agent type includes a single-step agent type and a sequential agent type.

[0073] Step S302, determining a first distributed agent node in a plurality of distributed agent nodes based on a load resource prediction model and a load demand associated with the single-step agent type subtask, and sending the single-step agent type subtask to the first distributed agent node for processing to obtain a first processing result.

[0074] Step S303, screening out a plurality of candidate agent nodes with a load rate lower than a first threshold value and a communication delay time lower than a second threshold value from each unselected distributed agent node, combining the plurality of candidate agent nodes according to the minimum number of divided tasks to obtain a plurality of candidate agent node combinations, and determining a second agent node combination based on the matching degree between each candidate agent node in each candidate agent node combination.

[0075] The subtask of the sequential agent type includes a plurality of sequentially executed minimum partition tasks. The first threshold is a threshold for comparing the load rate, which is set according to actual needs. The second threshold is a threshold for comparing the communication delay time, which is set according to actual needs. The plurality of candidate agent nodes can be obtained by comparing the load rate and the communication delay time of each unselected distributed agent node with the first threshold and the second threshold. The plurality of candidate agent node combinations can be obtained by combining the plurality of candidate agent nodes according to the number of minimum partition tasks in the subtask of the sequential agent type. For example, the unselected distributed agent nodes are node 1, node 2, node 3, node 4 and node 5, and the minimum partition tasks in the subtask of the sequential agent type are task a→task b→task c in turn. The candidate agent nodes with the load rate lower than the first threshold and the communication delay time lower than the second threshold are node 1, node 2, node 3 and node 4. The plurality of candidate agent node combinations obtained by combining the plurality of candidate agent nodes according to the number of minimum partition tasks are node 1, 2, 3, node 1, 2, 4, node 1, 3, 4 and node 2, 3, 4.

[0076] In step S304, the subtask of the sequential agent type is sent to the second agent node combination for processing to obtain a second processing result. The first processing result and the second processing result are fused to obtain the final processing result of the service data.

[0077] As described above, the plurality of candidate agent nodes with the load rate lower than the first threshold and the communication delay time lower than the second threshold are selected from the unselected distributed agent nodes. The plurality of candidate agent node combinations are obtained by combining the plurality of candidate agent nodes according to the number of minimum partition tasks. The present scheme determines the distributed agent node combination for processing the subtask of the sequential agent type according to the load rate and the communication delay time, which can ensure the rationality of the selected distributed agent node combination and improve the processing efficiency.

[0078] Figure 4 A flowchart of a complex service processing method based on a distributed intelligent agent system according to an embodiment of the present application is shown in FIG. 4. The method includes the following steps. Figure 4

[0079] In step S401, the service data to be processed and the corresponding service processing requirement are obtained. The service processing requirement is analyzed to obtain a plurality of subtasks. The agent type of each subtask is allocated according to the attribute of the service data to obtain the corresponding agent type. The agent type includes a single-step agent type and a sequential agent type.

[0080] ​In step S402, the first distributed agent node is determined in the plurality of distributed agent nodes based on the load resource prediction model and the load demand of the subtask associated with the single-step agent type, and the subtask of the single-step agent type is sent to the first distributed agent node for processing to obtain a first processing result.

[0081] In step S403, a plurality of candidate agent node combinations of the subtask of the sequential agent type are screened according to the load rates and the communication delay times of the unselected distributed agent nodes, each candidate agent node in each candidate agent node combination is sequentially arranged to obtain a plurality of sequential agent node sequences corresponding to the candidate agent node combination, and the second agent node combination is determined according to the matching degrees between the candidate agent nodes and the plurality of sequential agent node sequences of each candidate agent node combination.

[0082] The matching degree can be the compatibility between the distributed agent nodes. Alternatively, a determination method of the second agent node combination can be that the matching degrees between the sequentially connected candidate agent nodes in the plurality of sequential agent node sequences of the candidate agent node combination are superimposed to obtain a comprehensive matching degree, the sequential agent node sequence with the highest comprehensive matching degree is determined as the to-be-determined sequential combination corresponding to the candidate agent node combination, the comprehensive matching degrees of the to-be-determined sequential combinations are compared, and the to-be-determined sequential combination with the highest comprehensive matching degree is determined as the second agent node combination. For example, the candidate agent node combination A is distributed agent node 1, 2 and 3, the sequential agent node sequence a is 1→2→3, the sequential agent node sequence b is 1→3→2, the sequential agent node sequence c is 2→1→3, the sequential agent node sequence d is 2→3→1, the sequential agent node sequence e is 3→1→2, the sequential agent node sequence f is 3→2→1, the matching degree between the distributed agent node 1 and the distributed agent node 2 in the sequential agent node sequence a is 90, the matching degree between the distributed agent node 2 and the distributed agent node 3 is 95, and the comprehensive matching degree is 185. The comprehensive matching degrees of the sequential agent node sequences b, c, d, e and f are 190, 180, 175, 196 and 185 in turn, the sequential agent node sequence with the highest comprehensive matching degree is 3→1→2, and 3→1→2 is determined as the to-be-determined sequential combination of the candidate agent node combination A. In the same way, the to-be-determined sequential combinations of the candidate agent node combinations B, C and D are 1→2→4, 1→3→4 and 2→3→4 in turn, and the corresponding comprehensive matching degrees are 193, 190 and 188 in turn. The comprehensive matching degree of the to-be-determined sequential combination 3→1→2 is the highest, and therefore 3→1→2 is determined as the second agent node combination.

[0083] In another embodiment, the second proxy node combination can be determined in the following manner: average matching degrees between sequentially connected candidate proxy nodes in a plurality of sequential proxy node sequences of a candidate proxy node combination are calculated, a sequential proxy node sequence with the highest average matching degree is determined as a to-be-determined sequential combination corresponding to the candidate proxy node combination, comparison is performed according to the average matching degrees of the to-be-determined sequential combinations, and a to-be-determined sequential combination with the highest average matching degree is determined as the second proxy node combination.

[0084] In step S404, the sequential proxy type subtask is sent to the second proxy node combination for processing to obtain a second processing result, and the first processing result and the second processing result are fused to obtain a final processing result of the service data.

[0085] As described above, the second proxy node combination is determined by sequentially arranging each candidate proxy node in each candidate proxy node combination to obtain a plurality of corresponding sequential proxy node sequences and according to the matching degrees between the candidate proxy nodes and the plurality of sequential proxy node sequences of each candidate proxy node combination. The present scheme determines the distributed proxy node combination for processing the sequential proxy type subtask according to the matching degrees between the candidate proxy nodes, which can improve the rationality of the selected distributed proxy node combination and the task processing efficiency.

[0086] Figure 5 A flowchart of a complex service processing method based on a distributed intelligent agent system for determining a final processing result according to an embodiment of the present application is shown in FIG. 5, which specifically includes the following steps. Figure 5

[0087] In step S501, service data to be processed and corresponding service processing requirements are obtained, the service processing requirements are parsed to obtain a plurality of subtasks, and each subtask is assigned a proxy type according to the attributes of the service data, wherein the proxy type includes a single-step proxy type and a sequential proxy type.

[0088] In step S502, a first distributed proxy node is determined in a plurality of distributed proxy nodes based on a load resource prediction model and a load requirement associated with the single-step proxy type subtask, and the single-step proxy type subtask is sent to the first distributed proxy node for processing to obtain a first processing result.

[0089] In step S503, a plurality of candidate proxy node combinations of the sequential proxy type subtask are screened out according to the load rates and communication delay times of the unselected distributed proxy nodes, and a second proxy node combination is determined based on the matching degrees between the candidate proxy nodes in each candidate proxy node combination.

[0090] ​In step S504, the subtask of the sequential agent type is sent to the second agent node combination for processing to obtain a second processing result, a first weight of the first processing result is determined according to a processing time length corresponding to the first processing result, a second weight of the second processing result is determined according to a historical processing success rate associated with the second agent node combination corresponding to the second processing result, and the first processing result and the second processing result are fused and calculated according to the first weight and the second weight to obtain a final processing result of the business data.

[0091] The processing time length is used to represent a time length for the first distributed agent node to process the corresponding single-step agent type subtask to obtain the first processing result. The processing time length is used to determine the first weight of the first processing result. The historical processing success rate can be a success rate when the second agent node combination processes historical tasks. The historical processing success rate is used to determine the second weight of the second processing result. The first weight and the second weight are fused and calculated to obtain the final processing result of the business data. For example, the single-step agent type subtask is to detect transaction amount risk, the sequential agent type subtask is to analyze user behavior risk (analyze user login behavior, analyze user browsing behavior, and analyze user payment behavior), the first processing result is a transaction risk score 9.0, the second processing result is a behavior risk score 8.0, the processing time length of the first result is 2s, the corresponding weight is 0.4, the historical processing success rate associated with the second agent node combination corresponding to the second processing result is 80%, and the corresponding weight is 0.6. Therefore, the final risk score is 9.0*0.4+8.0*0.6=0.84.

[0092] As can be seen from the above, the first weight of the first processing result is determined according to the processing time length corresponding to the first processing result, the second weight of the second processing result is determined according to the historical processing success rate associated with the second agent node combination corresponding to the second processing result, and the first processing result and the second processing result are fused and calculated to obtain the final processing result of the business data. According to the processing time length and the historical processing success rate, the corresponding weights of the processing results are allocated, and the final processing result of the business data is obtained by fusion calculation. The rationality and accuracy of the final processing result can be improved.

[0093] Figure 6 A module structure block diagram of a complex business processing system based on a distributed intelligent agent system is provided for the embodiments of the present application. The system is used to execute the complex business processing method based on the distributed intelligent agent system provided by the above embodiments, and has the corresponding function modules and beneficial effects of the execution method. As shown in Figure 6 The system specifically includes:

[0094] The acquisition module 101 is configured to acquire business data to be processed and corresponding business processing requirements.

[0095] a task analysis module 102, configured to analyze the business processing requirement to obtain a plurality of subtasks;

[0096] a type allocation module 103, configured to allocate a proxy type to each of the subtasks according to an attribute of the business data, the proxy type including a single-step proxy type and a sequential proxy type;

[0097] a proxy node determination module 104, configured to determine a first distributed proxy node based on a load resource prediction model and a load requirement associated with the subtask of the single-step proxy type among a plurality of distributed proxy nodes;

[0098] a task sending module 105, configured to send the subtask of the single-step proxy type to the first distributed proxy node for processing to obtain a first processing result;

[0099] a node combination determination module 106, configured to filter a plurality of candidate proxy node combinations of the subtask of the sequential proxy type according to a load rate and a communication delay time of each of the distributed proxy nodes that is not selected, and determine a second proxy node combination based on a matching degree between each of the candidate proxy nodes in each of the candidate proxy node combinations;

[0100] the task sending module 105 is further configured to send the subtask of the sequential proxy type to the second proxy node combination for processing to obtain a second processing result;

[0101] a fusion processing module 107, configured to fuse the first processing result and the second processing result to obtain a final processing result of the business data.

[0102] According to the above scheme, by acquiring the to-be-processed business data and the corresponding business processing demand, the business processing demand is analyzed and processed to obtain a plurality of sub-tasks, the attribute of the business data is used to allocate a proxy type for each sub-task to obtain a corresponding proxy type, the proxy type includes a single-step proxy type and a sequential proxy type, based on a load resource prediction model and a load demand associated with a single-step proxy type sub-task, a first distributed proxy node is determined in a plurality of distributed proxy nodes, and the single-step proxy type sub-task is sent to the first distributed proxy node for processing to obtain a first processing result, a plurality of candidate proxy node combinations of the sequential proxy type sub-task are selected according to the load rate and the communication delay time of each unselected distributed proxy node, a second proxy node combination is determined based on the matching degree between each candidate proxy node in each candidate proxy node combination, and the sequential proxy type sub-task is sent to the second proxy node combination for processing to obtain a second processing result. The first processing result and the second processing result are fused to obtain the final processing result of the business data. The scheme solves the problem that the processing efficiency of the prior art for complex business problems is low and cannot meet the processing demand, and can realize efficient processing of business data, while ensuring processing efficiency and reducing processing cost.

[0103] In one possible embodiment, the proxy node determination module 104 is specifically configured to:

[0104] Input the current comprehensive load information of the plurality of distributed proxy nodes and the time period into the load resource prediction model to obtain the load resource prediction information of each distributed proxy node.

[0105] According to the load resource prediction information of each distributed proxy node and the load demand, a first distributed proxy node of the sub-task of the single-step proxy type is determined.

[0106] In one possible embodiment, the proxy node determination module 104 is further configured to:

[0107] In the case where there are a plurality of single-step proxy type sub-tasks, the load demands associated with each single-step proxy type sub-task are compared and processed, and the allocation priority of each single-step proxy type sub-task is obtained according to the comparison processing result.

[0108] According to the load demand of the sub-task with the lowest allocation priority and the load resource prediction information of each distributed proxy node, a selectable proxy node sequence combination is obtained, and the first distributed proxy node of each single-step proxy type sub-task is determined according to the allocation priority and the selectable proxy node sequence combination.

[0109] In a possible embodiment, the node combination determining module 106 is specifically configured to:

[0110] According to the unselected respective distributed agent nodes, a plurality of candidate agent nodes with a load rate lower than a first threshold value and a communication delay time lower than a second threshold value are screened out, and the plurality of candidate agent nodes are processed in combination according to the minimum number of divided tasks to obtain a plurality of candidate agent node combinations.

[0111] In a possible embodiment, the node combination determining module 106 is further configured to:

[0112] Each of the candidate agent nodes in each of the candidate agent node combinations is sequentially arranged to obtain a plurality of sequential agent node sequences corresponding to the candidate agent node combinations.

[0113] A second agent node combination is determined according to a matching degree between the respective candidate agent nodes and the sequential agent node sequences of each of the candidate agent node combinations.

[0114] In a possible embodiment, the node combination determining module 106 is further configured to:

[0115] The matching degrees between sequentially connected respective candidate agent nodes in the sequential agent node sequences of the candidate agent node combinations are superimposed to obtain a comprehensive matching degree, and a sequential agent node sequence with the highest comprehensive matching degree is determined as a to-be-determined sequential combination of a corresponding candidate agent node combination.

[0116] The to-be-determined sequential combinations are compared according to the comprehensive matching degrees of the respective to-be-determined sequential combinations, and a to-be-determined sequential combination with the highest comprehensive matching degree is determined as the second agent node combination.

[0117] In a possible embodiment, the fusion processing module 107 is specifically configured to:

[0118] A first weight of the first processing result is determined according to a processing time length corresponding to the first processing result, and a second weight of the second processing result is determined according to a historical processing success rate associated with the second agent node combination corresponding to the second processing result.

[0119] The first processing result and the second processing result are fused and calculated according to the first weight and the second weight to obtain a final processing result of the service data.

[0120] Figure 7 A structural schematic diagram of a complex service processing device based on a distributed intelligent agent system is provided for the embodiments of the present application, as shown in Figure 7As shown, the device includes a processor 201, a memory 202, an input device 203 and an output device 204; the number of processors 201 in the device can be one or more, Figure 7 The processor 201 in the device is taken as an example in this embodiment. The processor 201, the memory 202, the input device 203 and the output device 204 in the device can be connected through a bus or other means, Figure 7 The memory 202 is taken as an example in this embodiment. The memory 202 is a computer readable storage medium, which can be used to store software programs, computer executable programs and modules, such as program instructions or modules corresponding to the complex business processing method based on a distributed intelligent agent system in the embodiments of the present application. The processor 201 executes the software programs, instructions and modules stored in the memory 202, thereby performing various functional applications and data processing of the device, i.e., implementing the complex business processing method based on a distributed intelligent agent system. The input device 203 can be used to receive input digital or character information, and generate key signal input related to user settings and function control of the device. The output device 204 can include a display device such as a display screen.

[0121] The embodiments of the present application also provide a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute a complex business processing method based on a distributed intelligent agent system, the method comprising:

[0122] Obtaining business data to be processed and corresponding business processing requirements, performing analysis and processing on the business processing requirements to obtain a plurality of subtasks, assigning an agent type to each of the subtasks according to the attributes of the business data to obtain a corresponding agent type, the agent type including a single-step agent type and a sequential agent type;

[0123] Determining a first distributed agent node in a plurality of distributed agent nodes based on a load resource prediction model and a load demand associated with the subtask of the single-step agent type, and sending the subtask of the single-step agent type to the first distributed agent node for processing to obtain a first processing result;

[0124] Filtering a plurality of candidate agent node combinations of the subtask of the sequential agent type according to the load rates and communication delay times of each of the unselected distributed agent nodes, and determining a second agent node combination based on the matching degrees between each of the candidate agent nodes in each of the candidate agent node combinations;

[0125] Sending the subtask of the sequential agent type to the second agent node combination for processing to obtain a second processing result, and performing fusion processing on the first processing result and the second processing result to obtain a final processing result of the business data.

[0126] It is worth noting that the embodiments of the complex business processing method system based on the distributed intelligent agent system described above include various units and modules, which are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy mutual differentiation, and are not used to limit the protection scope of the embodiments of the present application.

[0127] It should be noted that the above are only the preferred embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the embodiments of the present application are not limited to the specific embodiments described herein, and those skilled in the art can make various obvious changes, readjustments and substitutions without departing from the protection scope of the embodiments of the present application. Therefore, although the embodiments of the present application have been described in more detail through the above embodiments, the embodiments of the present application are not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the embodiments of the present application, and the scope of the embodiments of the present application is determined by the scope of the appended claims.

Claims

1. A method for complex business process based on a distributed intelligent agent system, characterized in that, The method comprises: obtaining to-be-processed business data and corresponding business processing requirements, performing analysis and processing on the business processing requirements to obtain a plurality of subtasks, and performing proxy type allocation on each of the subtasks according to the attributes of the business data to obtain corresponding proxy types, wherein the proxy types include a single-step proxy type and a sequential proxy type; determining a first distributed proxy node in a plurality of distributed proxy nodes based on a load resource prediction model and a load demand associated with a subtask of the single-step proxy type, and sending the subtask of the single-step proxy type to the first distributed proxy node for processing to obtain a first processing result; screening a plurality of candidate proxy node combinations of the subtask of the sequential proxy type according to the load rates and communication delay times of unselected distributed proxy nodes, and determining a second proxy node combination based on the matching degrees between each candidate proxy node in each of the candidate proxy node combinations; sending the subtask of the sequential proxy type to the second proxy node combination for processing to obtain a second processing result, and performing fusion processing on the first processing result and the second processing result to obtain a final processing result of the business data.

2. The complex business process handling method based on a distributed intelligent agent system according to claim 1, characterized in that, The determination of the first distributed proxy node in the plurality of distributed proxy nodes based on the load resource prediction model and the load demand associated with the subtask of the single-step proxy type comprises: inputting current comprehensive load information of the plurality of distributed proxy nodes and time periods into the load resource prediction model to obtain load resource prediction information of each of the distributed proxy nodes; determining the first distributed proxy node of the subtask of the corresponding single-step proxy type according to the load resource prediction information of each of the distributed proxy nodes and the load demand.

3. The complex business process handling method based on a distributed intelligent agent system according to claim 2, characterized in that, The determination of the first distributed proxy node of the subtask of the corresponding single-step proxy type according to the load resource prediction information of each of the distributed proxy nodes and the load demand comprises: in the case where there are a plurality of subtasks of the single-step proxy type, performing comparison processing on the load demands associated with each of the subtasks of the single-step proxy type, and performing priority sorting according to the comparison processing result to obtain allocation priorities of each of the subtasks of the single-step proxy type; performing comparison between the load demand of the subtask with the lowest allocation priority and the load resource prediction information of each of the distributed proxy nodes to obtain a selectable proxy node sequence combination, and determining the first distributed proxy node of each of the subtasks of the single-step proxy type according to the allocation priority and the selectable proxy node sequence combination.

4. The complex business process handling method based on a distributed intelligent agent system according to any one of claims 1 to 3, characterized in that, The subtask of the sequential proxy type comprises a plurality of sequentially executed minimum division tasks, and the screening of the plurality of candidate proxy node combinations of the subtask of the sequential proxy type according to the load rates and communication delay times of unselected distributed proxy nodes comprises: According to the load rate and the communication delay time of each of the unselected distributed agent nodes, a plurality of candidate agent nodes with a load rate lower than a first threshold and a communication delay time lower than a second threshold are screened out, and the plurality of candidate agent nodes are combined according to the minimum number of divided tasks to obtain a plurality of candidate agent node combinations.

5. The complex business process handling method based on a distributed intelligent agent system according to any one of claims 1 to 3, characterized in that, The second agent node combination is determined based on the matching degree between each candidate agent node in each candidate agent node combination, comprising: Each candidate agent node in each candidate agent node combination is sequentially arranged to obtain a corresponding plurality of sequential agent node sequences. The second agent node combination is determined based on the matching degree between each candidate agent node and the plurality of sequential agent node sequences of each candidate agent node combination.

6. The complex business process handling method based on a distributed intelligent agent system according to claim 5, characterized by, The second agent node combination is determined based on the matching degree between each candidate agent node in each candidate agent node combination, comprising: The matching degrees between sequentially connected candidate agent nodes in the plurality of sequential agent node sequences of the candidate agent node combination are superimposed to obtain a comprehensive matching degree, and the sequential agent node sequence with the highest comprehensive matching degree is determined as a to-be-determined sequential combination of the corresponding candidate agent node combination. The to-be-determined sequential combinations are compared based on the comprehensive matching degrees, and the to-be-determined sequential combination with the highest comprehensive matching degree is determined as the second agent node combination.

7. The complex business process handling method based on a distributed intelligent agent system according to any one of claims 1 to 3, characterized in that, The first processing result and the second processing result are fused to obtain the final processing result of the service data, comprising: A first weight of the first processing result is determined according to a processing time length corresponding to the first processing result, and a second weight of the second processing result is determined according to a historical processing success rate associated with the second agent node combination corresponding to the second processing result; The first processing result and the second processing result are fused and calculated according to the first weight and the second weight to obtain the final processing result of the service data.

8. A complex business processing system based on a distributed intelligent agent system, characterized in that, Comprising: An acquisition module is configured to acquire service data to be processed and corresponding service processing requirements; A task analysis module is configured to analyze and process the service processing requirements to obtain a plurality of subtasks; A type allocation module is configured to allocate agent types to each subtask according to the attributes of the service data to obtain corresponding agent types, wherein the agent types include a single-step agent type and a sequential agent type; An agent node determination module is configured to determine a first distributed agent node based on a load resource prediction model and a load demand associated with a subtask of the single-step agent type in a plurality of distributed agent nodes; A task sending module is configured to send the subtask of the single-step agent type to the first distributed agent node for processing to obtain a first processing result; A node combination determination module is configured to screen out a plurality of candidate agent node combinations of the subtask of the sequential agent type according to the load rate and the communication delay time of each of the unselected distributed agent nodes, and determine a second agent node combination based on the matching degree between each candidate agent node in each candidate agent node combination. The task sending module is further configured to send the sub-tasks of the sequential agent type to the second agent node combination for processing to obtain a second processing result. The fusion processing module is configured to perform fusion processing on the first processing result and the second processing result to obtain a final processing result of the service data.

9. A complex business processing apparatus based on a distributed intelligent agent system, characterized by, The device includes one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the complex service processing method based on the distributed intelligent agent system as claimed in any one of claims 1-7.

10. A storage medium storing computer-executable instructions, wherein: The computer executable instructions, when executed by a computer processor, are configured to perform the complex service processing method based on the distributed intelligent agent system as claimed in any one of claims 1-7. The computer executable instructions, when executed by a computer processor, are configured to perform the complex service processing method based on the distributed intelligent agent system as claimed in any one of claims 1-7.

Images