Decision support method and system based on large model technology

Abstract

The application discloses a decision support method and system based on large model technology. The method comprises the following steps: acquiring target business process data in an execution target business process and a plurality of target business nodes corresponding to the target business, and constructing a target business graph according to the nodes and the process data; determining the similarity between the nodes through the graph to generate a target similarity matrix; evaluating a target importance score set and an automation feasibility score set of each node from multiple dimensions; using a pre-trained target large model to perform deep analysis on the business process data, the similarity matrix, the importance score set and the automation feasibility score set, and generating an optimization processing strategy for each node, including keeping the original state, deleting redundant nodes or performing node automation configuration; and processing each target business node according to the optimization strategy. The application solves the technical problem of low efficiency caused by cumbersome business processing nodes when processing related businesses.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of financial services, in particular to a decision support method and system based on large model technology. BACKGROUND

[0002] In current business process management, as the scale of enterprises and the complexity of business increase, business nodes become increasingly cumbersome, and a large number of nodes contained in modern enterprise business processes usually involve multiple departments and systems, the dependency relationship between nodes is complex, and at present, many business processes still rely on manual management and adjustment, especially in terms of node optimization and resource allocation, there is a lack of automated tool support, the data of different business nodes is scattered in various systems, which is difficult to integrate and analyze, limiting the global optimization of the overall process, and due to the excessive and complex business nodes, manual processing is prone to delay and error, affecting the overall business efficiency, traditional methods are difficult to effectively identify and eliminate unnecessary business nodes, resulting in resource waste, and the existing technology cannot dynamically adjust node configuration according to business data, and the optimization strategy is mainly based on experience, lacking scientific basis.

[0003] In view of the above problems, no effective solution has been proposed so far. SUMMARY

[0004] The embodiments of the present application provide a decision support method and system based on large model technology, to at least solve the technical problem of low efficiency caused by cumbersome business processing nodes when processing related businesses.

[0005] According to an aspect of an embodiment of the present application, a decision support method based on large model technology is provided, comprising: acquiring target business process data in an execution target business process and a plurality of target business nodes corresponding to the target business, constructing a target business graph according to the plurality of target business nodes and the target business process data; determining a target similarity between each target business node according to the target business graph, and constructing a target similarity matrix according to each target similarity; determining a target importance score set corresponding to each target business node from multiple dimensions according to the target business graph, and determining a target automation feasibility score set corresponding to each target business node from multiple dimensions; analyzing the target business process data, the target similarity matrix, the plurality of target importance score sets and the plurality of target automation feasibility score sets by using a pre-trained target large model, to obtain an optimization processing strategy for each target business node, wherein the optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration; and executing the corresponding optimization processing strategy for each target business node.

[0006] Optionally, the target business process data in the execution of the target business process is acquired, and a plurality of target business nodes corresponding to the target business are acquired, a target business graph is constructed according to the plurality of target business nodes and the target business process data, including: acquiring the target business process data in the execution of the target business process, wherein the target business process data includes at least one of the following: operation log, task record, processing timestamp, resource allocation information; a plurality of target business nodes corresponding to the target business are determined, and multi-dimensional target attribute information of each target business node and a task dependency relationship between each target business node are determined according to the target business process data, wherein the target attribute information includes at least one of the following: node name, execution frequency, single execution time, single input data, single output data, preset node level, execution exception frequency, and single resource consumption; a target business graph corresponding to the target business is constructed according to each target business node and the corresponding target attribute information and the task dependency relationship.

[0007] Optionally, the target similarity between each target business node is determined according to the target business graph, and a target similarity matrix is constructed according to each target similarity, including: for each two target business nodes, the target sub-similarity between the target attribute information of the two target business nodes in each dimension is calculated respectively; the target sub-similarities in the plurality of dimensions are weighted and summed according to a preset weight coefficient to obtain the target similarity between the two target business nodes; and the target similarity matrix is constructed according to the target similarity between each target business node.

[0008] Optionally, a set of target importance scores corresponding to each target business node is determined from multiple dimensions according to the target business graph, including: for each target business node, the execution frequency, single execution time, preset node level, and single resource consumption of the target business node are obtained from the target business graph; the execution frequency score of the target business node is determined according to the execution frequency; the average execution time of the target business node is determined according to the single execution time, and the efficiency score of the target business node is determined according to the average execution time; the average resource consumption of the target business node is determined according to the single resource consumption, and the cost score of the target business node is determined according to the average resource consumption; the importance score of the target business node is determined according to the preset node level; the in-degree and out-degree of the target business node are determined according to the target business graph, and the in-degree score and out-degree score of the target business node are determined according to the in-degree and out-degree respectively; the centrality index of the target business node in the target business graph is determined, wherein the centrality index includes at least one of the following: degree centrality index, closeness centrality index, and betweenness centrality index; the execution frequency score, efficiency score, cost score, importance score, in-degree score, out-degree score, and centrality index are combined to obtain the set of target importance scores of the target business node.

[0009] Optionally, the target automation feasibility score set corresponding to each target business node is determined from multiple dimensions, including: for each target business node, determining the single-time input data, single-time output data, and execution exception number of the target business node according to the target business graph; determining the input standardization score of the target business node according to the single-time input data; determining the processing logic rule of the target business node when executing the target business according to the single-time input data and the single-time output data, and determining the rule score of the target business node according to the processing logic rule; determining the abnormal risk score of the target business node according to the execution exception number; and combining the input standardization score, the rule score, and the abnormal risk score to obtain the target automation feasibility score set of the target business node.

[0010] Optionally, the training process of the target large model includes: obtaining a plurality of sets of business process data corresponding to a plurality of businesses, and determining a plurality of business nodes corresponding to each business to construct a business graph of each business; for each business, determining a similarity matrix corresponding to a plurality of business nodes of the business, an importance score set of each business node, and an automation feasibility score set according to the business graph of the business; taking the business process data, the similarity matrix, the plurality of importance score sets, and the plurality of automation feasibility score sets corresponding to the business as a training sample, and taking the optimization processing strategy for the business as the corresponding sample label; constructing an initial large model, and iteratively training the initial large model using the plurality of training samples and sample labels, and taking the trained large model as the target large model.

[0011] Optionally, in the case where the optimization processing strategy is node automation configuration, the optimization processing strategy further includes an automation script generated by the target large model for the target business node to be subjected to node automation configuration, wherein the automation script includes code instructions and flow control logic required to implement the function of the target business node; and executing the corresponding optimization processing strategy on the target business node includes adding the automation script to the target business node.

[0012] Optionally, according to another aspect of the embodiments of the present application, a decision support system based on large model technology is also provided, comprising: an acquisition module configured to acquire target business process data in a target business process and a plurality of target business nodes corresponding to the target business, and construct a target business graph according to the plurality of target business nodes and the target business process data; a construction module configured to determine target similarities between the target business nodes according to the target business graph, and construct a target similarity matrix according to the target similarities; a determination module configured to determine a target importance score set corresponding to each target business node from multiple dimensions according to the target business graph, and determine a target automation feasibility score set corresponding to each target business node from multiple dimensions; a strategy generation module configured to analyze the target business process data, the target similarity matrix, the plurality of target importance score sets, and the plurality of target automation feasibility score sets by using a pre-trained target large model, to obtain an optimization processing strategy for each target business node, wherein the optimization processing strategy comprises at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration; and an execution module configured to execute the corresponding optimization processing strategy for each target business node.

[0013] According to another aspect of the embodiments of the present application, a computer program product is also provided, which comprises a computer program, wherein the computer program is executed by a processor to implement the above-mentioned decision support method based on large model technology.

[0014] According to another aspect of the embodiments of the present application, an electronic device is also provided, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-mentioned decision support method based on large model technology through the computer program.

[0015] In the embodiments of the present application, by acquiring target business process data in the execution of the target business process and a plurality of target business nodes in the business process, a target business graph reflecting the structure of the business process and the relationship between the nodes is constructed; based on the target business graph, the similarity between each business node is determined, and a target similarity matrix is constructed, which can help identify repeated or similar links in the process and provide a basis for deleting redundant nodes; using the target business graph, a target importance score set of each business node is determined from multiple dimensions (such as execution frequency, efficiency, cost, etc.), and an automation feasibility score set is determined from different angles (such as input data standardization, processing logic regularization, abnormal risk, etc.); by pre-training a target large model, the target business process data, the target similarity matrix, the target importance score set and the automation feasibility score set are deeply analyzed, and based on the understanding and prediction ability of the business process, the large model can generate an optimization processing strategy for each business node, such as maintaining the state, deleting redundant nodes or configuring an automated node. These strategies are data-driven and can maximize process efficiency and reduce costs. The analysis of the target large model provides data-driven optimization recommendations to help decision-makers make more scientific and accurate process improvement decisions. By deleting redundant nodes and configuring automated nodes, manual operations are reduced, further shortening the execution time of the business process and improving overall processing efficiency, thereby solving the technical problem of low efficiency caused by cumbersome business processing nodes when processing related businesses. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of the present application and help to explain the present application together with the specification. The illustrative embodiments of the present application and their description serve to explain the present application. In the drawings:

[0017] Figure 1 is a flow diagram of an optional decision support method based on large model technology according to an embodiment of the present application;

[0018] Figure 2 is a structural diagram of an optional decision support system based on large model technology according to an embodiment of the present application;

[0019] Figure 3 is a structural diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order for those skilled in the art to better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0021] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0022] The information collected in the embodiments of the present application is information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards of relevant countries and regions, necessary security measures are taken, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.

[0023] Embodiment 1

[0024] According to the embodiments of the present application, a decision support method based on large model technology is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that herein.

[0025] Figure 1 is a flow diagram of a decision support method based on large model technology according to the embodiments of the present application, as Figure 1 shown, the method comprises the following steps:

[0026] Step S102, acquiring target business process data in the execution target business process and a plurality of target business nodes corresponding to the target business, constructing a target business graph according to the plurality of target business nodes and the target business process data;

[0027] Step S104, determining target similarities between each target business node according to the target business graph, and constructing a target similarity matrix according to each target similarity;

[0028] Step S106, determining a set of target importance scores corresponding to each target business node from multiple dimensions according to the target business graph, and determining a set of target automation feasibility scores corresponding to each target business node from multiple dimensions;

[0029] Step S108, using the pre-trained target large model to analyze the target business process data, the target similarity matrix, the multiple sets of target importance scores, and the multiple sets of target automation feasibility scores, to obtain an optimization processing strategy for each target business node, wherein the optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration;

[0030] Step S110, executing the corresponding optimization processing strategy for each target business node.

[0031] The steps of the decision support method based on large model technology will be described in detail in combination with the specific implementation process.

[0032] First, the target business process data in the execution of the target business process and the multiple target business nodes corresponding to the target business are obtained, and a target business graph is constructed according to the multiple target business nodes and the target business process data. This process can be performed in the following steps:

[0033] The target business process data in the execution of the target business process is obtained, wherein the target business process data includes at least one of the following: operation log, task record, processing timestamp, resource allocation information; the multiple target business nodes corresponding to the target business are determined, and the multidimensional target attribute information of each target business node and the task dependency relationship between each target business node are determined according to the target business process data, wherein the target attribute information includes at least one of the following: node name, execution times, single execution time, single input data, single output data, preset node level, execution exception times, and single resource consumption; and a target business graph corresponding to the target business is constructed according to each target business node and the corresponding target attribute information and task dependency relationship.

[0034] The obtained business process data can be business process data of executing the same business multiple times.

[0035] For example, in a bank processing system, the operation log records the operation records of customer managers, approvers, etc. in the business processing process, including operation time, operation type, operation result, etc. The task record tracks the processing of each business node, records the start and end time of the task, the task state (such as to be done, completed, abnormal), the task executor, etc. The processing timestamp accurately records the time point when each business node starts processing and completes processing, which helps to calculate the processing time of the node. The resource allocation information involves the resources consumed by each node during execution, including manpower, system resources, financial resources, etc., which is an important basis for evaluating the cost and efficiency of the node.

[0036] The target business node refers to a step or link that can be individually identified and analyzed in the target business process. For example, in a credit business, the business nodes may include "customer information collection", "credit evaluation", "loan approval", "contract signing", "loan lending", etc. Determining these nodes and collecting relevant attribute information is the key to building a business graph. Attribute information may include but is not limited to: the name of each node, such as "customer credit evaluation"; the total number of times each node is triggered, reflecting the activity level of the node; the execution time of each execution of the node, used to evaluate efficiency and optimization points; the data types and formats required before and after the execution of the node, and the output results generated; and the preset level according to the importance and risk level of the node, such as levels 1 to 5.

[0037] For a single business, depending on the design of the business process and the needs of the specific business scenario, in some standardized and automated business processes, a business node of the business may be designed to be executed only once to ensure the continuity and efficiency of the business process and avoid repeated operations and resource waste. For example, in a loan approval business process, the "credit evaluation" node may be set to be executed only once after the loan application is submitted to generate a credit report for the customer. This step is critical and unique in the process, ensuring the accuracy and timeliness of credit evaluation. However, in complex business processes or certain specific conditions, a business node may need to be executed multiple times, which usually occurs in scenarios that require loop processing or adjustment of processing logic based on business status. For example, in post-loan management, the "repayment monitoring" node may need to be repeatedly executed periodically to check the repayment of the loan customer and ensure the normal recovery of the loan. In customer consultation or complaint handling, the "customer service response" node may need to be executed multiple times until the customer's problem is completely resolved.

[0038] By analyzing the operation log and task record, the dependency relationship between nodes in the business process is identified, for example, in the credit business, "customer information collection" may need to be completed before "loan approval", and "loan approval" depends on the result of "credit evaluation", the identification of such dependency relationship is crucial for building the correct business process graph, which ensures the accuracy and practicality of the graph.

[0039] After obtaining the target business graph, the target similarity between each target business node is determined according to the target business graph, and a target similarity matrix is constructed according to the target similarity of each target business node. This process can be carried out in the following steps:

[0040] For each two target business nodes, the target sub-similarity between the target attribute information of the two target business nodes in each dimension is calculated respectively; the target sub-similarity in multiple dimensions is weighted and summed according to the preset weight coefficient to obtain the target similarity between the two target business nodes; the target similarity matrix is constructed according to the target similarity between each target business node.

[0041] The most suitable similarity calculation method can be selected according to the nature of the target attribute information and the business requirements, for example, for discrete attributes such as node name, Jaccard similarity can be used, and the setting of weight coefficient is a key step, which determines the bias of similarity calculation, the weight coefficient can be set based on expert experience, or can be automatically determined by machine learning methods such as principal component analysis or feature selection algorithm.

[0042] After obtaining the target business graph, the target importance score set corresponding to each target business node is determined from multiple dimensions according to the target business graph, and the target automation feasibility score set corresponding to each target business node is determined from multiple dimensions, the process can be carried out in the following steps:

[0043] For each target business node, the execution times, single execution time, preset node level, and single resource consumption of the target business node are obtained from the target business graph;

[0044] The execution frequency score of the target business node is determined according to the execution times;

[0045] For example, business node A is customer information audit, and business node B is loan amount calculation, in a single loan approval process, node A may be triggered multiple times, because in a complex process such as loan approval, the information submitted by the customer may need to be reviewed multiple times to ensure the accuracy and completeness of the information, for example, the initial review may need to check the basic information, and then resubmit the review of specific details, or the customer's information needs to be reviewed in different stages of the loan (such as application, loan, post-loan management), on the contrary, node B may be triggered only once in the process, to determine the loan amount of the customer.

[0046] For calculating the execution frequency score, the node with the most execution times among all nodes can be found, and its execution times are recorded as a reference value, then the execution times of the current node are compared with the reference value to calculate a proportion value, and finally the proportion value is converted into a percentage score, the more the execution times of a node, the higher its execution frequency score.

[0047] The average execution time of the target business node is determined according to the single execution time, and the efficiency score of the target business node is determined according to the average execution time;

[0048] For example, the average execution time is calculated by using multiple single execution times, the node with the shortest average execution time among all nodes is found, and its execution time is recorded as a reference value, then the reference value is compared with the average execution time of the current node to calculate a proportion value, and finally the proportion value is converted into a percentage score, the shorter the single execution time of a node, the higher its efficiency score.

[0049] The average resource consumption of the target business node is determined according to the single resource consumption, and the cost score of the target business node is determined according to the average resource consumption;

[0050] For example, the average resource consumption (such as memory, network bandwidth, etc.) of each execution of the node is counted, the node with the least average resource consumption among all nodes is found, and its average resource consumption is recorded as a reference value, then the reference value is compared with the single resource consumption of the current node to calculate a proportion value, and finally the proportion value is converted into a percentage score, the less the average resource consumption of a node, the higher its cost score.

[0051] The importance score of the target business node is determined according to the preset node level;

[0052] For example, when designing a business process, each node is pre-evaluated and assigned a level, which is based on the importance and complexity of the node in the process, a node directly related to fund transactions may be assigned a higher level, while some purely document processing nodes may have a lower level, according to the level of the node, a score range can be set, for example, high-level nodes can be set to 80 to 100 points, medium-level nodes to 50 to 80 points, and low-level nodes to 0 to 50 points, the specific score can be adjusted within the set range according to the level of the node.

[0053] The in-degree and out-degree of the target business node are determined according to the target business graph, and the in-degree score and out-degree score of the target business node are determined according to the in-degree and out-degree respectively;

[0054] The in-degree represents how many nodes point to the current node, i.e., the number of times the current node serves as a precondition of a subsequent node, and the in-degree score reflects the "dependence degree" of the node, i.e., how many working results of other nodes will enter the node. The out-degree represents how many other nodes the current node points to, i.e., how many working results of the current node serve as a precondition of other nodes, and the out-degree score reflects the "influence degree" of the node, i.e., the degree of influence of the node on subsequent processes. The scores of in-degree and out-degree can be determined based on the in-degree and out-degree numbers of the node in the graph. For example, the higher the in-degree and out-degree numbers, the higher the score; otherwise, the lower the score.

[0055] For the in-degree score, the following calculation method can be referred to: find the node with the highest in-degree among all nodes, and record its in-degree value as a reference value, compare the in-degree of the current node with the reference value, calculate a proportion value, and finally convert the proportion value into a percentage score. The higher the in-degree of the node, the higher its in-degree score. This means that the node is relied on by more other nodes and plays a more important role in the business process. For the out-degree score, the following calculation method can be referred to: find the node with the highest out-degree among all nodes, and record its out-degree value as a reference value, compare the out-degree of the current node with the reference value, calculate a proportion value, and finally convert the proportion value into a percentage score. The higher the out-degree of the node, the higher its out-degree score, which indicates that the node has stronger connectivity with other nodes and plays a more complex role in the business process.

[0056] Determine the centrality index of the target business node in the target business graph, wherein the centrality index includes at least one of the following: degree centrality index, closeness centrality index, betweenness centrality index; combine the execution frequency score, efficiency score, cost score, importance score, in-degree score, out-degree score, and centrality index to obtain a target importance score set of the target business node.

[0057] For example, the degree centrality index represents the direct connectivity degree of the node, i.e., the number of other nodes directly associated with the node. A node with high degree centrality has more direct connections in the graph, indicating that it plays a core role in the process. The closeness centrality index reflects the "center degree" of the node in the graph, i.e., the average shortest path length of the node from other nodes in the entire graph. A node with high closeness centrality means that it is in a "central position" in the graph and can quickly and effectively interact with other nodes in the graph. The betweenness centrality index measures the frequency of the node as an intermediate node of the shortest path between other nodes in the graph. A node with high betweenness centrality plays a "bridge" role in the graph and has an important influence on information flow.

[0058] For the degree centrality index, the following calculation method can be referred to: in the business graph, the number of direct connections between the current node and other nodes (i.e. the degree) is counted, the node with the highest degree among all nodes is found, and the degree of the node is recorded as a reference value. The degree of the current node is compared with the reference value to calculate a proportion value, and finally the proportion value is converted into a percentage score. The higher the degree of a node, the higher its degree centrality score, indicating that the node has more direct connections with other nodes in the business process.

[0059] For the closeness centrality index, the following calculation method can be referred to: in the business graph, the shortest path length from the current node to all other nodes is calculated, and the average of these path lengths is found. The node with the smallest average shortest path length among all nodes is found, and the average path length of the node is recorded as a reference value. The reference value is compared with the average shortest path length of the current node to calculate a proportion value, and finally the proportion value is converted into a percentage score. The shorter the average shortest path length of a node, the higher its closeness centrality score, indicating that the node is easier to reach other nodes in the business process.

[0060] For the betweenness centrality index, the following calculation method can be referred to: in the business graph, the number of times the current node acts as an "intermediary" in all shortest paths between other nodes is counted. The node with the highest betweenness centrality value among all nodes is found, and the betweenness centrality value of the node is recorded as a reference value. The betweenness centrality value of the current node is compared with the reference value to calculate a proportion value, and finally the proportion value is converted into a percentage score. The higher the betweenness centrality value of a node, the higher its betweenness centrality score, indicating that the node plays a more important role as a bridge in the business process.

[0061] As an optional implementation, the target automation feasibility score set corresponding to each target business node can be determined from multiple dimensions by taking the following steps: for each target business node, determining the single-input data, single-output data, and execution exception number of the target business node according to the target business graph; determining the input standardization score of the target business node according to the single-input data; determining the processing logic rule of the target business node when executing the target business according to the single-input data and the single-output data, and determining the rule score of the target business node according to the processing logic rule; determining the abnormal risk score of the target business node according to the execution exception number; and combining the input standardization score, the rule score, and the abnormal risk score to obtain the target automation feasibility score set of the target business node.

[0062] In the business process graph, each node has its specific input and output data format and type, single input data refers to the total data set received by the node when processing a specific business, and single output data is the output result set after the node processing, which may include customer information, transaction records, approval status, risk assessment results, etc. The consistency of input data format and content is crucial for automation. If the input data of a node can follow a unified structure and template in most cases, the node has a higher input standardization score, which means it is easier to be automated. For the standardization score, the following calculation method can be referred to: define the standard template or rules of input data (for example, whether the data fields are complete, whether the data types match, etc.), check the matching degree of each input data with the standard template, and give a score according to the matching degree, for example: if it is completely matched, the score is 100, if it is partially matched, the score is calculated according to the matching proportion (such as 80% matching degree, the score is 80), if it is not matched at all, the score is 0.

[0063] The processing logic rules of the node determine whether it can be accurately converted into an automated processing strategy. The higher the rule score, the more feasible the automation implementation. The rule score is determined as follows: rule clarity: analyze the processing logic of the node, check whether it is based on a predefined rule set, and whether the rule description is detailed and clear, evaluate the complexity of the node processing logic, including the number of rules, the dependency between them, and the handling of exceptional cases, investigate whether the node processing logic remains stable over a long period of time, frequent changes in rules will reduce the feasibility of automation, the calculation of rule score can be based on the comprehensive consideration of the above three dimensions, the rule clarity, rule complexity and rule stability are quantified as scores respectively, and then weighted average is performed.

[0064] The number of exceptions reflects the probability of problems that may occur after the node is automated. The more exceptions, the lower the exception risk score. The exception risk score can be calculated as follows: count the total number of executions and the number of exceptions of the node, compare the number of exceptions of the current node with the total number of executions, calculate a proportion value, and finally convert this proportion value into a percentage score.

[0065] After obtaining the target similarity matrix, multiple target importance score sets and multiple target automation feasibility score sets, the pre-trained target large model is used to analyze the target business process data, the target similarity matrix, the multiple target importance score sets and the multiple target automation feasibility score sets, to obtain an optimized processing strategy for each target business node, wherein the optimized processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration.

[0066] As an optional implementation, the training process of the target large model can take the following steps: obtaining multiple sets of business process data corresponding to multiple businesses, and determining multiple business nodes corresponding to each business to construct a business graph of each business; for each business, determining a similarity matrix corresponding to the multiple business nodes of the business, a set of importance scores of each business node, and a set of automation feasibility scores according to the business graph of the business; taking the business process data, the similarity matrix, the multiple sets of importance scores, and the multiple sets of automation feasibility scores corresponding to the business as a training sample, and taking the optimization processing strategy for the business as the corresponding sample label; constructing an initial large model, and iteratively training the initial large model using the multiple training samples and sample labels, and taking the trained large model as the target large model.

[0067] After obtaining the optimization processing strategy for each business node, the corresponding optimization processing strategy is executed for each target business node.

[0068] As an optional implementation, in the case of node automation configuration, the optimization processing strategy further includes an automation script generated by the target large model for the target business node to be subjected to node automation configuration, wherein the automation script includes code instructions and flow control logic required to implement the function of the target business node; executing the corresponding optimization processing strategy for the target business node includes adding the automation script to the target business node.

[0069] For example, the target large model analyzes the target similarity matrix, the target set of importance scores, and the target set of automation feasibility scores according to its understanding of the business process, and outputs an optimization processing strategy. For automation configuration, the model generates an automation script containing code instructions and flow control logic to implement the function of the node, while considering other nodes in the business process and potential exception handling mechanisms. The generation of the automation script depends on the specific demand analysis of the target business node. The large model automatically generates corresponding code instructions according to the input data, processing logic, and output data of the node. For example, if the node needs to process data standardization and format conversion, the script will contain code for data cleaning and conversion; if the node involves decision analysis, the script will include rule-based decision logic or invocation of a machine learning model.

[0070] According to the strategy output of the target large model, a detailed implementation plan can be determined, including a timeline, resource allocation, and responsible personnel. The automation script is deployed in the production environment, the automation node is activated, the running state and performance indicators of the node are monitored, the node deletion operation is performed according to the redesign of the business process for the identified redundant nodes, the process is simplified, and after removing the redundant nodes and activating the automation nodes, the entire business process is re-evaluated and adjusted to ensure the coherence and efficiency of the process.

[0071] Based on these indicators, the bank can evaluate the effect of the optimization strategy and make adjustments as needed, for example, regularly collecting performance data of the business process, comparing with the baseline data before optimization, collecting customer feedback on the satisfaction of the improved process, understanding the impact of automated processing on customer experience, regularly reviewing the business process, identifying new optimization opportunities or correcting the implemented strategy, adjusting the automation script according to the monitoring results and customer feedback, optimizing the business process, and ensuring the continuous effectiveness and adaptability of the strategy.

[0072] In the embodiments of the present application, by acquiring target business process data in the execution of the target business process and a plurality of target business nodes in the business process, a target business graph reflecting the structure of the business process and the relationship between the nodes is constructed; based on the target business graph, the similarity between each business node is determined, and a target similarity matrix is constructed, which can help identify repeated or similar links in the process and provide a basis for deleting redundant nodes; using the target business graph, a target importance score set of each business node is determined from multiple dimensions (such as execution frequency, efficiency, cost, etc.), and an automation feasibility score set is determined from different angles (such as input data standardization, processing logic regularization, abnormal risk, etc.); by pre-training a target large model, the target business process data, the target similarity matrix, the target importance score set and the automation feasibility score set are deeply analyzed, and based on the understanding and prediction ability of the business process, the large model can generate optimization processing strategies for each business node, such as maintaining the state, deleting redundant nodes or configuring automation nodes. These strategies are data-driven and can maximize process efficiency and reduce costs. The analysis of the target large model provides data-driven optimization recommendations to help decision-makers make more scientific and accurate process improvement decisions. By deleting redundant nodes and configuring automation nodes, manual operations are reduced, further shortening the execution time of the business process and improving overall processing efficiency, thereby solving the technical problem of low efficiency caused by tedious business processing nodes when processing related businesses.

[0073] Embodiment 2

[0074] According to the embodiments of the present application, a large model technology-based decision support system for implementing the large model technology-based decision support method in Embodiment 1 is also provided, as shown in Figure 2 The large model technology-based decision support system at least includes an acquisition module 21, a construction module 22, a determination module 23, a strategy generation module 24 and an execution module 25, wherein:

[0075] The acquisition module 21 is configured to acquire target business process data in a target business process and a plurality of target business nodes corresponding to the target business, and construct a target business graph according to the plurality of target business nodes and the target business process data.

[0076] The construction module 22 is configured to determine a target similarity between each of the target business nodes according to the target business graph, and construct a target similarity matrix according to each of the target similarities.

[0077] The determination module 23 is configured to determine a target importance score set corresponding to each of the target business nodes from multiple dimensions according to the target business graph, and determine a target automation feasibility score set corresponding to each of the target business nodes from multiple dimensions.

[0078] The strategy generation module 24 is configured to analyze the target business process data, the target similarity matrix, a plurality of the target importance score sets, and a plurality of the target automation feasibility score sets by using a pre-trained target large model, to obtain an optimization processing strategy for each of the target business nodes, wherein the optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration.

[0079] The execution module 25 is configured to execute the corresponding optimization processing strategy for each of the target business nodes.

[0080] The functions of each module of the decision support system based on the large model technology will be described in detail in combination with the specific implementation process.

[0081] First, the acquisition module acquires target business process data in a target business process and a plurality of target business nodes corresponding to the target business, and constructs a target business graph according to the plurality of target business nodes and the target business process data. This process can be performed by the following steps:

[0082] The target business process data in the execution target business process is acquired, wherein the target business process data includes at least one of the following: operation log, task record, processing timestamp, resource allocation information; a plurality of target business nodes corresponding to the target business are determined, and multi-dimensional target attribute information of each target business node and a task dependency relationship between each target business node are determined according to the target business process data, wherein the target attribute information includes at least one of the following: node name, execution times, single execution time, single input data, single output data, preset node level, execution exception times, and single resource consumption; and a target business graph corresponding to the target business is constructed according to each of the target business nodes and the corresponding target attribute information and the task dependency relationship.

[0083] After obtaining the target business graph, the constructing module determines target similarities between each target business node according to the target business graph, and constructs a target similarity matrix according to the target similarities. The process can be performed by the following steps:

[0084] For each two target business nodes, a target sub-similarity between target attribute information of the two target business nodes in each dimension is calculated respectively; target sub-similarities in multiple dimensions are weighted and summed according to a preset weight coefficient to obtain a target similarity between the two target business nodes; and a target similarity matrix is constructed according to the target similarities between each target business node.

[0085] After obtaining the target business graph, the determining module determines a target importance score set corresponding to each target business node from multiple dimensions according to the target business graph, and determines a target automation feasibility score set corresponding to each target business node from multiple dimensions. The process can be performed by the following steps:

[0086] For each target business node, the execution times, single execution time, preset node level, and single resource consumption of the target business node are obtained from the target business graph; an execution frequency score of the target business node is determined according to the execution times; an average execution time of the target business node is determined according to the single execution time, and an efficiency score of the target business node is determined according to the average execution time; an average resource consumption of the target business node is determined according to the single resource consumption, and a cost score of the target business node is determined according to the average resource consumption; an importance score of the target business node is determined according to the preset node level; the in-degree and out-degree of the target business node are determined according to the target business graph, and an in-degree score and an out-degree score of the target business node are determined according to the in-degree and the out-degree respectively; a centrality index of the target business node in the target business graph is determined, wherein the centrality index includes at least one of the following: a degree centrality index, a closeness centrality index, and a betweenness centrality index; and the target importance score set of the target business node is obtained by combining the execution frequency score, the efficiency score, the cost score, the importance score, the in-degree score, the out-degree score, and the centrality index.

[0087] As an optional implementation, the following steps can be taken to determine the target automation feasibility score set corresponding to each target business node from multiple dimensions: for each target business node, determining the single-time input data, single-time output data, and execution exception number of the target business node according to the target business graph; determining the input standardization score of the target business node according to the single-time input data; determining the processing logic rule of the target business node when executing the target business according to the single-time input data and the single-time output data, and determining the rule score of the target business node according to the processing logic rule; determining the abnormal risk score of the target business node according to the execution exception number; and combining the input standardization score, the rule score, and the abnormal risk score to obtain the target automation feasibility score set of the target business node.

[0088] After obtaining the target similarity matrix, the plurality of target importance score sets, and the plurality of target automation feasibility score sets, the strategy generation module analyzes the target business process data, the target similarity matrix, the plurality of target importance score sets, and the plurality of target automation feasibility score sets using the pre-trained target large model to obtain an optimization processing strategy for each target business node, wherein the optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration.

[0089] As an optional implementation, the following steps can be taken to train the target large model: obtaining a plurality of sets of business process data corresponding to a plurality of businesses, and determining a plurality of business nodes corresponding to each business to construct a business graph of each business; for each business, determining a similarity matrix corresponding to the plurality of business nodes of the business, an importance score set of each business node, and an automation feasibility score set of each business node according to the business graph of the business; taking the business process data, the similarity matrix, the plurality of importance score sets, and the plurality of automation feasibility score sets corresponding to the business as a training sample, and taking the optimization processing strategy for the business as the corresponding sample label; constructing an initial large model, and iteratively training the initial large model using the plurality of training samples and the sample labels, and taking the trained large model as the target large model.

[0090] After obtaining the optimization processing strategy for each business node, the execution module executes the corresponding optimization processing strategy for each target business node.

[0091] As an optional implementation, in the case where the optimization processing strategy is node automation configuration, the optimization processing strategy further includes an automation script generated by the target large model for the target business node to be configured with node automation, wherein the automation script includes code instructions and flow control logic required to implement the function of the target business node; and executing the corresponding optimization processing strategy for the target business node includes adding the automation script to the target business node.

[0092] It should be noted that the modules in the decision support system based on large model technology in the embodiments of the present application correspond one by one to the implementation steps of the decision support method based on large model technology in Embodiment 1. Since Embodiment 1 has been described in detail, the details not embodied in this embodiment can be referred to Embodiment 1, and will not be described in detail here.

[0093] Embodiment 3

[0094] According to the embodiments of the present application, a computer program product is also provided, which includes a computer program. When the computer program is executed by a processor, the decision support method based on large model technology in Embodiment 1 is implemented.

[0095] According to the embodiments of the present application, a non-volatile storage medium is also provided, which includes a stored computer program. The device in which the non-volatile storage medium is located executes the decision support method based on large model technology in Embodiment 1 by running the computer program.

[0096] According to the embodiments of the present application, a processor is also provided, which is used to run a computer program. When the computer program is run, the decision support method based on large model technology in Embodiment 1 is executed.

[0097] According to the embodiments of the present application, an electronic device is also provided, which includes a memory and a processor. The memory stores a computer program, and the processor is configured to execute the decision support method based on large model technology in Embodiment 1 by the computer program.

[0098] Specifically, when the computer program is run, the following steps are implemented: obtaining target business process data in an execution target business process and a plurality of target business nodes corresponding to the target business, constructing a target business graph according to the plurality of target business nodes and the target business process data; determining a target similarity between each target business node according to the target business graph, and constructing a target similarity matrix according to each target similarity; determining a target importance score set corresponding to each target business node from multiple dimensions according to the target business graph, and determining a target automation feasibility score set corresponding to each target business node from multiple dimensions; analyzing the target business process data, the target similarity matrix, the plurality of target importance score sets and the plurality of target automation feasibility score sets by using a pre-trained target large model to obtain an optimization processing strategy for each target business node, wherein the optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and node automation configuration; and executing the corresponding optimization processing strategy for each target business node.

[0099] As an optional implementation, the electronic device can exist in the form of a mobile terminal, a computer terminal, or a similar computing device. Figure 3 A hardware structure block diagram of an electronic device for implementing a decision support method based on large model technology is shown. As shown in Figure 3 The electronic device 30 can include one or more processors 302 (the processor 302 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 304 for storing data, and a transmission device 306 for communication functions. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports in the BUS bus), a network interface, a power supply and / or a camera. Those skilled in the art can understand that Figure 3 The structure shown is only schematic, which does not limit the structure of the above-mentioned electronic device. For example, the electronic device 30 can include more or fewer components than Figure 3 shown, or have a different configuration than Figure 3 shown.

[0100] It should be noted that the one or more processors 302 and / or other data processing circuits described above can be referred to herein as "data processing circuits" in general. The data processing circuit can be embodied in whole or in part as software, hardware, firmware or any other combination. In addition, the data processing circuit can be a single independent processing module, or all or part of any one of the other elements combined into the electronic device 30. As referred to in the embodiments of the present application, the data processing circuit serves as a processor control (for example, selection of a variable resistance terminal path connected to an interface).

[0101] The memory 304 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the decision support method based on large model technology in the embodiments of the present application. The processor 302 executes various functional applications and data processing by running the software programs and modules stored in the memory 304, that is, implements the vulnerability detection method of the application program described above. The memory 304 can include a high-speed random access memory, and can also include a non-volatile memory such as one or more magnetic storage devices, flash memory, or other non-volatile solid state memory. In some examples, the memory 304 can further include a memory remotely located with respect to the processor 302, which can be connected to the electronic device 30 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0102] The transmission device 306 is configured to receive or send data via a network. The network can include a wireless network provided by a communication provider of the electronic device 30. In one embodiment, the transmission device 306 includes a network interface controller (NIC) that can connect to other network devices through a base station to communicate with the Internet. In one embodiment, the transmission device 306 can be a radio frequency (RF) module that is configured to communicate with the Internet wirelessly.

[0103] The display can be a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the electronic device 30.

[0104] The above-mentioned embodiment numbers are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0105] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0106] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the device embodiment described above is only illustrative, and for example, the division of units can be a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, unit or module, and can be electrical or other forms.

[0107] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0108] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0109] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0110] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

Claims

1. A decision support method based on large model technology, characterized in that, include: Obtain target business process data and multiple target business nodes corresponding to the target business during the execution of the target business, and construct a target business graph based on the multiple target business nodes and the target business process data; The target similarity between each target service node is determined based on the target service graph, and a target similarity matrix is ​​constructed based on each target similarity. Based on the target business map, a target importance score set corresponding to each target business node is determined from multiple dimensions, and a target automation feasibility score set corresponding to each target business node is determined from multiple dimensions; The target business process data, the target similarity matrix, multiple target importance score sets, and multiple target automation feasibility score sets are analyzed using a pre-trained target large model to obtain an optimization processing strategy for each target business node. The optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and automating node configuration. Execute the corresponding optimization processing strategy for each of the target service nodes; The process of determining the target automation feasibility score set for each target business node from multiple dimensions includes: for each target business node, determining the single input data, single output data, and number of execution anomalies of the target business node based on the target business graph; determining the input standardization score of the target business node based on the single input data; determining the processing logic rules of the target business node when executing the target business based on the single input data and single output data, and determining the rule score of the target business node based on the processing logic rules; determining the anomaly risk score of the target business node based on the number of execution anomalies; and combining the input standardization score, rule score, and anomaly risk score to obtain the target automation feasibility score set for the target business node. The training process of the target large model includes: acquiring multiple sets of business process data corresponding to multiple businesses, and determining multiple business nodes corresponding to each business to construct a business graph for each business; for each business, determining the similarity matrix, the importance score set, and the automation feasibility score set corresponding to the multiple business nodes of the business based on the business graph of the business; using the business process data, similarity matrix, multiple importance score sets, and multiple automation feasibility score sets corresponding to the business as a training sample, and using the optimization processing strategy for the business as the corresponding sample label; constructing an initial large model, and iteratively training the initial large model using multiple training samples and sample labels, and using the trained large model as the target large model.

2. The method according to claim 1, characterized in that, Acquire target business process data and multiple target business nodes corresponding to the target business during the execution of the target business, and construct a target business graph based on the multiple target business nodes and the target business process data, including: Obtain target business process data during the execution of the target business, wherein the target business process data includes at least one of the following: operation log, task record, processing timestamp, and resource allocation information; Multiple target business nodes corresponding to the target business are identified, and multi-dimensional target attribute information of each target business node and task dependency relationships between each target business node are determined based on the target business process data. The target attribute information includes at least one of the following: node name, number of executions, single execution time, single input data, single output data, preset node level, number of execution exceptions, and single resource consumption. Based on each target business node and its corresponding target attribute information, as well as the task dependency relationship, the target business map corresponding to the target business is constructed.

3. The method according to claim 2, characterized in that, Determine the target similarity between each target service node based on the target service graph, and construct a target similarity matrix based on each target similarity, including: For every two target business nodes, calculate the target sub-similarity between the target attribute information of the two target business nodes in each dimension; The target similarity between the two target business nodes is obtained by weighting and summing the target sub-similarity under multiple dimensions according to the preset weight coefficients. The target similarity matrix is ​​constructed based on the target similarity between each target business node.

4. The method according to claim 2, characterized in that, Based on the target business graph, a target importance score set corresponding to each target business node is determined from multiple dimensions, including: For each target business node, obtain the number of executions, single execution time, preset node level, and single resource consumption of the target business node from the target business graph; The execution frequency score of the target business node is determined based on the number of executions. The average execution time of the target business node is determined based on the single execution time, and the efficiency score of the target business node is determined based on the average execution time. The average resource consumption of the target business node is determined based on the single resource consumption, and the cost score of the target business node is determined based on the average resource consumption. The importance score of the target business node is determined based on the preset node level; The in-degree and out-degree of the target service node are determined based on the target service graph, and the in-degree score and out-degree score of the target service node are determined based on the in-degree and out-degree respectively. Determine the centrality index of the target service node in the target service graph, wherein the centrality index includes at least one of the following: degree centrality index, proximity centrality index, and betweenness centrality index; The execution frequency score, efficiency score, cost score, importance score, in-degree score, out-degree score, and centrality index are combined to obtain the target importance score set for the target business node.

5. The method according to claim 1, characterized in that, When the optimization processing strategy is node automated configuration, the optimization processing strategy also includes an automated script generated by the target large model for the target business node to be configured automatically, wherein the automated script includes code instructions and flow control logic required to implement the functions of the target business node; Execute corresponding optimization processing strategies on the target business node, including adding the automation script to the target business node.

6. A decision support system based on large model technology, characterized in that, include: The acquisition module is used to acquire target business process data and multiple target business nodes corresponding to the target business during the execution of the target business, and to construct a target business graph based on the multiple target business nodes and the target business process data; A construction module is used to determine the target similarity between each of the target service nodes based on the target service graph, and to construct a target similarity matrix based on each of the target similarities; The determination module is used to determine the target importance score set corresponding to each target business node from multiple dimensions based on the target business graph, and to determine the target automation feasibility score set corresponding to each target business node from multiple dimensions. The process of determining the target automation feasibility score set includes: for each target business node, determining the single input data, single output data, and execution anomaly count of the target business node based on the target business graph; determining the input standardization score of the target business node based on the single input data; determining the processing logic rules of the target business node when executing the target business based on the single input data and single output data, and determining the rule score of the target business node based on the processing logic rules; determining the anomaly risk score of the target business node based on the execution anomaly count; and combining the input standardization score, rule score, and anomaly risk score to obtain the target automation feasibility score set of the target business node. A strategy generation module is used to analyze the target business process data, the target similarity matrix, multiple target importance score sets, and multiple target automation feasibility score sets using a pre-trained target large model to obtain an optimization processing strategy for each target business node. The optimization processing strategy includes at least one of the following: maintaining the original state, deleting redundant nodes, and automating node configuration. The training process of the target large model includes: acquiring multiple sets of business process data corresponding to multiple businesses, determining multiple business nodes corresponding to each business, and constructing a business graph for each business; for each business, determining the similarity matrix, importance score set, and automation feasibility score set corresponding to the multiple business nodes based on the business graph; using the business process data, similarity matrix, multiple importance score sets, and multiple automation feasibility score sets corresponding to the business as a training sample, and using the optimization processing strategy for the business as the corresponding sample label; constructing an initial large model, and iteratively training the initial large model using multiple training samples and sample labels, and using the trained large model as the target large model. The execution module is used to execute the corresponding optimization processing strategy for each of the target business nodes.

7. A computer program product, characterized in that, include: A computer program, wherein when executed by a processor, the computer program implements the decision support method based on large model technology as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the decision support method based on large model technology according to any one of claims 1 to 5 through the computer program.

Images