Index anomaly root cause positioning method and system for complex process

Abstract

The invention provides a root cause positioning method and system for index abnormity of a complex process, and aims at solving the problems of numerous complex business process nodes, complex DAG structures, diversified operation types among the nodes, inconsistent index granularity of the nodes, high root cause analysis difficulty and the like, and sequentially constructing process DAG with different abstract levels from bottom to top. The influence degrees of operations such as conversion, shunting and aggregation of all stages on abnormal indexes are calculated layer by layer from high to low according to abstract levels, the influence degrees can be compared with one another, an operator is helped to rapidly recognize and position root cause paths and individual stage conversion items with the large influence degrees, and conduction calculation of the influence degrees is achieved through a chain rule. And thus, cross-layer stage conversion analysis is realized. Furthermore, based on a large language model and a graph retrieval enhancement generation technology, a root cause diagnosis question-answering system is constructed in combination with an exception record document, reasons of stage transformation exceptions are further analyzed through the root cause diagnosis question-answering system, and the analysis efficiency of stage transformation exceptions is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a method and system for locating the root cause of abnormal indicators in complex processes. Background Art

[0002] Most traditional BI products use visualization to display year-on-year and month-on-month changes in indicators. They lack the ability to diagnose and locate business problems. Business problem analysis relies too much on manual judgment, which is relatively inefficient and prone to misjudgment. In recent years, with the development of big model technology, many BI vendors have launched ChatBI products based on big models, enabling operators to conduct coherent data analysis and root cause location through multiple rounds of dialogue. Limited by the reasoning ability of big models, as well as the lack of scenario-based root cause location algorithms and contribution quantification indicators, many in-depth root cause analyses still need to be judged manually. Therefore, for various complex business processes of enterprises, when business indicators decline or change, how to quickly and accurately locate the root cause of the anomaly has become a problem that many managers need to solve urgently.

[0003] The business process of an enterprise is usually represented by a process DAG (Directed Acyclic Graph), where a node represents a certain stage or task unit in the process, such as page access, telemarketing call, quotation, transaction, etc. Complex business processes often contain many process nodes, complex DAG graph structure, diverse operation types between nodes (such as conversion, aggregation, diversion, etc.), and inconsistent indicator granularity of different nodes, making root cause analysis difficult. In recent years, technologies such as multi-agent, large language model, multimodal, graph neural network, and association rule mining have been successively applied in the diagnosis and positioning of operation and maintenance faults, but these methods do not consider factors such as process stages and conversion rates, and are difficult to apply to the root cause positioning scenario of complex business processes. For example, the patent (CN202110195548.6) completes the root cause positioning by calculating the impact value and impact degree of the initial stage and the subsequent stages of transformation on the final indicator. This method is mainly aimed at business processes with fewer nodes and relatively simple, and it is difficult to directly apply to complex business processes. Summary of the invention

[0004] The present invention provides a root cause location method and system for indicator anomalies in complex processes, so as to solve the defect that the prior art is difficult to apply to the root cause location scenario of complex business processes.

[0005] The present invention provides a method for locating the root cause of abnormal indicators in complex processes, comprising: Construct an initial process DAG based on the current business process and establish an association between each node in the initial process DAG and its indicator; the node of the initial process DAG corresponds to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; Abstracting the aggregation nodes and the diversion nodes in the initial process DAG to obtain the backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; Determine the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; Starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse order, and determining the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; The abnormal performance of the abnormal indicators and the root cause information after manual determination are recorded to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user performs a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage conversion in which an abnormality exists; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage conversion in which the abnormality exists.

[0006] According to a root cause location method for indicator anomalies of complex processes provided by the present invention, the aggregation nodes and the diversion nodes in the initial process DAG are abstractly processed to obtain the backbone process DAG, including: starting from the node with an in-degree of 0 in the initial process DAG, forward traversal, if the currently visited node is an aggregation node, calling the aggregation processing step for the currently visited node and continuing the traversal; if the currently visited node is a diversion node, calling the diversion processing step for the currently visited node and continuing the traversal; after the traversal is completed, the backbone process DAG is obtained; Among them, the aggregation processing step for any aggregation node includes: reverse traversing the unvisited transformation branches of the aggregation node, if any transformation branch contains the aggregation node, then calling the aggregation processing step for the aggregation node in the transformation branch to obtain the current process DAG, building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; if all transformation branches do not contain the aggregation node, then building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from the transformation branch of any node to the node; The diversion processing step for any diversion node includes: forward traversing the branches of the diversion node, if there is no diversion node in the branch of the diversion node, then determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, creating a directed edge from any diversion node to the virtual node corresponding to the aggregation node, and obtaining a new process DAG; otherwise, calling the diversion processing step for the diversion node in the branch of the diversion node to obtain the current process DAG, determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtaining a new process DAG; there is a directed edge from any node to the branch of this node.

[0007] According to a root cause location method for abnormal indicators of complex processes provided by the present invention, the method of determining the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal indicator based on the backbone process DAG includes: Calculate the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the backbone process DAG on the abnormal index; Traverse the backbone process DAG forward. If the currently visited node is a virtual node and there is no shunt node pointing to the node, call the aggregation node drill-down step for the currently visited node and continue traversing. If the currently visited node is a shunt node in the initial process DAG and the node is connected to a virtual node through a directed edge, call the shunt node drill-down step for the currently visited node and its connected virtual node and continue traversing. Among them, the aggregation node drill-down step for any virtual node includes: based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG, construct each conversion branch of the aggregation node; taking the indicator corresponding to the end node of each conversion branch of the aggregation node as the key indicator, calculating the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, based on the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, the influence of the end node of each conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, determine the influence of the conversion process corresponding to the initial node and adjacent nodes of each conversion branch on the abnormal indicator; if each conversion branch contains a virtual node, call the aggregation node drill-down step for the virtual node in the corresponding conversion branch; The shunt node drill-down step for any shunt node and its connected virtual node includes: constructing each branch of the shunt node based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG; taking the indicator corresponding to the end node of each branch of the shunt node as the key indicator, calculating the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, and determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the abnormal indicator based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator; if each branch contains a virtual node, calling the shunt node corresponding to the virtual node in the corresponding branch in the initial process DAG and the shunt node drill-down step of the virtual node.

[0008] According to a root cause location method for indicator anomalies in complex processes provided by the present invention, based on the influence of the transformation process corresponding to the initial node and adjacent nodes in each transformation branch on the key indicator, the influence of the final node of each transformation branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, the influence of the transformation process corresponding to the initial node and adjacent nodes of each transformation branch on the abnormal indicator is determined, including: For the initial node in any conversion branch, calculate the product of the influence of the initial node in the conversion branch on the key indicator of the same conversion branch, the influence of the final node of the conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the initial node of the conversion branch on the abnormal indicator; For the transformation process corresponding to the adjacent nodes in any transformation branch, calculate the product of the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the key indicator of the same transformation branch, the influence of the transformation branch end node corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the abnormal indicator; The influence of the final node of the transformation branch corresponding to any key indicator on the virtual node is the ratio of the difference between the key indicator in the current cycle and the previous cycle to the difference between the indicator associated with the aggregation node corresponding to the virtual node in the initial process DAG in the current cycle and the previous cycle.

[0009] According to a root cause location method for abnormal indicators of complex processes provided by the present invention, based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator, the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the abnormal indicator is determined, including: For the shunt node, calculate the product of the influence of the shunt node on the key indicator in any branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the conversion process corresponding to the shunt node and the virtual node on the abnormal indicator as the influence of the shunt node on the abnormal indicator; For any transformation process among the transformation processes corresponding to the diversion node and the initial nodes of each branch and the transformation processes corresponding to the adjacent nodes of each branch, calculate the product of the influence of the transformation process on the key indicator of the same branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the diversion node and the virtual node on the abnormal indicator as the influence of the transformation process on the abnormal indicator.

[0010] According to a root cause location method for abnormal indicators of complex processes provided by the present invention, the initial process DAG is traversed in reverse from the node corresponding to the abnormal indicator in the initial process DAG, and the root cause location path of the abnormal indicator is determined based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator, including: The initial process DAG is traversed in reverse from the node corresponding to the abnormal index. If the node currently visited is an aggregation node, when the change direction of the index value of the node currently visited is the same as the abnormal index, the influence of the end nodes of each conversion branch of the node currently visited on the node currently visited is sorted from large to small, and the conversion branches with positive influence and greater than the preset threshold are selected in turn by a greedy strategy until the sum of the influence of the end nodes of the selected conversion branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected conversion branches; when the change direction of the index value of the node currently visited is different from the abnormal index, the influence of the end nodes of each conversion branch of the node currently visited on the node currently visited is sorted from large to small according to the absolute value, and the conversion branches with negative influence and absolute value greater than the preset threshold are selected in turn by a greedy strategy until the sum of the absolute value of the influence of the end nodes of the selected conversion branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected conversion branches; the reverse traversal is performed until a node with an in-degree of 0 is reached, and the root cause location path of the abnormal index is obtained.

[0011] According to a root cause location method for abnormal indicators of complex processes provided by the present invention, a root cause location path of the abnormal indicators is determined, and then the following steps are further included: Calculate the root cause discrimination of each root cause location path, and select the root cause location path with the largest root cause discrimination as the most significant root cause location path; The root cause discrimination of any root cause location path i is calculated based on the following method: C(i)=α× +β×

[0012] Where C(i) is the root cause discrimination of the root cause location path i; α and β are custom coefficients, and α+β=1; AND global (v) The influence of the transformation process corresponding to the initial node or any adjacent node in the root cause location path i on the abnormal index; standard(s) is the standard deviation of the influence of the transformation process corresponding to the initial node and each adjacent node in the root cause location path i on the abnormal index, and k is the number of root cause location paths.

[0013] The present invention also provides a root cause location system for abnormal indicators of complex processes, including: An initial DAG construction module is used to construct an initial process DAG based on the current business process and establish an association between each node in the initial process DAG and its indicator; the nodes of the initial process DAG correspond to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; A DAG abstraction module is used to abstract the aggregation nodes and the shunt nodes in the initial process DAG to obtain a backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the shunt node is greater than 1; An influence degree calculation module, used to determine the influence degree of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; A root cause location path output module is used to reversely traverse the initial process DAG from the node corresponding to the abnormal indicator in the initial process DAG, and determine the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; The root cause diagnosis question and answer system module is used to record the abnormal performance of the abnormal indicators and the root cause information after manual determination to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user makes a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage transformation with an abnormality; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage transformation with the abnormality.

[0014] According to a root cause location system for indicator anomalies of complex processes provided by the present invention, the aggregation nodes and the diversion nodes in the initial process DAG are abstractly processed to obtain the backbone process DAG, including: starting from the node with an in-degree of 0 in the initial process DAG, forward traversal, if the currently visited node is an aggregation node, calling the aggregation processing step for the currently visited node and continuing the traversal; if the currently visited node is a diversion node, calling the diversion processing step for the currently visited node and continuing the traversal; after the traversal is completed, the backbone process DAG is obtained; Among them, the aggregation processing step for any aggregation node includes: reverse traversing the unvisited transformation branches of the aggregation node, if any transformation branch contains the aggregation node, then calling the aggregation processing step for the aggregation node in the transformation branch to obtain the current process DAG, building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; if all transformation branches do not contain the aggregation node, then building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from the transformation branch of any node to the node; The diversion processing step for any diversion node includes: forward traversing the branches of the diversion node, if there is no diversion node in the branch of the diversion node, then determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between the aggregation node and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; otherwise, calling the diversion processing step for the diversion node in the branch of the diversion node to obtain the current process DAG, determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between the aggregation node and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from any node to the branch of this node.

[0015] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for locating the root cause of an indicator anomaly for a complex process as described in any one of the above is implemented.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for locating the root cause of an indicator anomaly for a complex process.

[0017] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for locating the root cause of indicator anomalies for complex processes.

[0018] The present invention provides a root cause location method and system for indicator anomalies in complex processes. In view of the problems of a large number of nodes in complex business processes, a complex DAG graph structure, various node types, inconsistent indicator granularity of different nodes, and great difficulty in root cause analysis, a process DAG graph with different abstraction levels is constructed from the bottom up in sequence, and the influence of operations such as transformation, diversion, and aggregation at each stage on the abnormal index is calculated layer by layer from high to low according to the abstraction level. The influences are comparable to each other, which can help operators quickly identify and locate the root cause path and individual stage transformation items with large influences. The conduction calculation of the influence is realized by the chain rule, thereby realizing cross-layer stage transformation analysis. In addition, a root cause diagnosis question-and-answer system can be constructed based on a large model + graph retrieval generation technology and combined with anomaly record documents. The root cause diagnosis question-and-answer system further analyzes the causes of abnormalities caused by stage transformation, effectively utilizes past diagnosis experience, and improves the analysis efficiency of stage transformation abnormalities. Subsequently, according to actual conditions, the abnormal record documents are added or updated, and the update of the knowledge graph is completed, so that the large language model retrieves the updated knowledge graph, and the question-and-answer results of the root cause diagnosis will be more accurate and comprehensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is a flow chart of a method for locating the root cause of abnormal indicators in complex processes provided by the present invention; Figure 2 It is a schematic diagram of the aggregation node and the transformation branch provided by the present invention; Figure 3 It is one of the schematic diagrams of the polymerization process steps provided by the present invention; Figure 4 This is the second schematic diagram of the polymerization process steps provided by the present invention; Figure 5 is a schematic diagram of a shunt node and its branches provided by the present invention; Figure 6 It is one of the schematic diagrams of the diversion processing steps provided by the present invention; Figure 7 This is the second schematic diagram of the diversion processing step provided by the present invention; Figure 8 It is a structural schematic diagram of a root cause location system for abnormal indicators of complex processes provided by the present invention; Figure 9 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Figure 1 : is a flow chart of a method for locating the root cause of abnormal indicators in complex processes provided by the present invention, such as Figure 1 As shown, the method includes: Step 110, constructing an initial process DAG based on the current business process and establishing associations between each node in the initial process DAG and its indicators; the nodes of the initial process DAG correspond to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; Step 120, abstracting the aggregation nodes and the diversion nodes in the initial process DAG to obtain a backbone process DAG; Among them, the in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; Step 130, determining the influence of the transformation processes corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; Step 140, starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse order, and determining the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; Step 150, record the abnormal performance of the abnormal indicator and the root cause information after manual determination to form an abnormal record document; use GraphRAG technology to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user performs a root cause diagnosis query, receive the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system; the context information includes the root cause path and a stage conversion related indicator information where an abnormality exists; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the result to the user, and determines the root cause of the stage conversion where an abnormality exists.

[0023] Here, step 120 includes: starting forward traversal from a node with an in-degree of 0 in the initial process DAG, if the currently visited node is an aggregation node, calling an aggregation processing step for the currently visited node and continuing traversal; if the currently visited node is a diversion node, calling a diversion processing step for the currently visited node and continuing traversal; after the traversal is completed, the backbone process DAG is obtained; Among them, the aggregation processing step for any aggregation node includes: reverse traversing the unvisited transformation branches of the aggregation node, if any transformation branch contains the aggregation node, then calling the aggregation processing step for the aggregation node in the transformation branch to obtain the current process DAG, building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; if all transformation branches do not contain the aggregation node, then building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from the transformation branch of any node to the node; The diversion processing step for any diversion node includes: forward traversing the branches of the diversion node, if there is no diversion node in the branch of the diversion node, then determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, creating a directed edge from any diversion node to the virtual node corresponding to the aggregation node, and obtaining a new process DAG; otherwise, calling the diversion processing step for the diversion node in the branch of the diversion node to obtain the current process DAG, determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtaining a new process DAG; there is a directed edge from any node to the branch of this node.

[0024] Specifically, an initial process DAG is constructed based on the business advancement sequence of the current business process. Among them, the node in the initial process DAG corresponds to a business stage or task unit, and the end node of the initial process DAG is the business stage or task unit corresponding to the abnormal indicator. Subsequently, an association can be established between each node in the initial process DAG and its indicator, so that the subsequent root cause location step can locate the root cause accordingly. It should be noted that the in-degree or out-degree of the node in the initial process DAG can be greater than 1. If the in-degree of the node is greater than 1, it can be called an aggregate node, which represents that multiple different stages or task units are merged into one stage or task unit. Assume that the relevant business stages are A, B, C and D, and the corresponding indicators are m respectively. a ,m b ,m c and m d The values ​​of these indicators in the current period are A(m a ) , A(m b ) , A(m c ) and A(m d ), then there exists A(m a )+A(m b )+A(m c )=A(m d ). The goal of aggregation is to achieve the convergence of several business stages / task units and their indicator values. If the out-degree of a node is greater than 1, it can be called a diversion node, which means that a business stage or task unit is diverted into multiple different stages or task units. Assume that the relevant business stages are A, B, C and D, and the corresponding indicators are m a ,m b ,m c and m d The values ​​of these indicators in the current period are A(m a ) , A(m b ) , A(m c ) and A(m d ), then there exists A(m a )=A(m b )+A(m c )+A(m d The goal of diversion is to decompose the indicator value of a certain business stage and distribute it to different sub-business stages or task units.

[0025] After obtaining the initial process DAG, we can build process DAGs with different abstraction levels from bottom to top until we get the backbone process DAG. On this basis, we first calculate the impact of the backbone process DAG, and drill down on the virtual nodes representing the sub-processes in the backbone process DAG one by one, and assign the relevant impact to different sub-processes. We drill down step by step from high to low according to the abstraction level until all virtual nodes are expanded and analyzed. Finally, we can get the impact of the transformation process corresponding to the initial node and any two adjacent nodes (two nodes with directed edges are called adjacent nodes) in the initial process DAG on the indicator abnormality.

[0026] In view of the problems of large number of nodes in complex business processes, complex DAG graph structure, diverse node types (such as aggregation nodes, diversion nodes), inconsistent indicator granularity of different nodes, and difficulty in root cause analysis, the embodiment of the present invention proposes a multi-level DAG graph construction algorithm, which constructs process DAG graphs with different abstraction levels from bottom to top, and continuously divides and abstracts the DAG graph until the backbone process DAG with the highest abstraction level is obtained. Specifically, it can be forward traversed from the node with an in-degree of 0 in the initial process DAG (if there are multiple nodes with an in-degree of 0, one is randomly selected as the starting point of the traversal). It should be noted that forward traversal refers to traversal in the direction of the directed edge, while reverse traversal refers to traversal in the opposite direction of the directed edge, which will not be repeated later. If the node currently accessed is an aggregation node, the aggregation processing step for the currently accessed node is called and the traversal continues; if the node currently accessed is a diversion node, the diversion processing step for the currently accessed node is called and the traversal continues; after the traversal is completed, the backbone process DAG can be obtained.

[0027] The aggregation processing step for any aggregation node includes: traversing the unvisited transformation branches of the aggregation node in reverse. Here, there is a directed edge from the transformation branch of any node to the node, such as Figure 2 As shown, the transformation branches of the aggregation node D are A, B and C. If any transformation branch also contains an aggregation node, the aggregation processing step for the aggregation node in the transformation branch is recursively called to obtain the current process DAG, and then the node set of the aggregation node is constructed based on the nodes in the aggregation node and its transformation branches in the current process DAG, and a virtual node corresponding to the aggregation node is created to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and a new process DAG is obtained. If all transformation branches do not contain an aggregation node, the node set of the aggregation node is constructed based on the nodes in the aggregation node and its transformation branches in the current process DAG, and a virtual node corresponding to the aggregation node is created to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and a new process DAG is obtained.

[0028] like Figure 3As shown, the currently visited aggregation node is D, where nodes A, B, and C are located in two different transformation branches of the aggregation node, and each transformation branch is traversed in reverse to see if it contains other aggregation operations. Since the transformation branches of the aggregation node D do not contain aggregation nodes, the node set of the aggregation node D is constructed as {A, B, C, D}, and the identified part of the subgraph is replaced by the new virtual node D'. Continue forward traversal from the virtual node D'. When the aggregation node J is accessed, the transformation branch of the aggregation node is traversed in reverse to confirm that node H is an aggregation node. Therefore, the above-mentioned aggregation processing steps will be performed on node H, that is, continue to traverse the transformation branch of node H in reverse and confirm whether there is an aggregation node on the transformation branch. If there is, continue to recursively call the aggregation processing steps for the newly appeared aggregation node. The current situation is that there is no aggregation node on the transformation branch of node H, so the node set of node H can be constructed {G, I, H}, and the identified part of the subgraph is replaced by the new virtual node H'. Back to the aggregation node J, since all the aggregation nodes on the transformation branch of the aggregation node have been replaced by the corresponding virtual nodes and no longer contain the aggregation node, the node set {E, F, D', H', J} of the aggregation node J is constructed, and the identified subgraph is replaced by the new virtual node J', such as Figure 4 shown.

[0029] The flow splitting processing step for any flow splitting node includes: forward traversing the branches of the flow splitting node. Wherein, there is a directed edge from any node to the branch of the node, such as Figure 5 As shown, the branches of node A are B, C and D. If there is no shunt node in the branch of the shunt node, the corresponding aggregation node of the shunt node is determined (generally, shunt and aggregation appear in pairs, and each branch of the shunt node converges at its corresponding aggregation node), and the node set of the aggregation node is constructed based on the aggregation node and the nodes between it and the shunt node, and a virtual node corresponding to the aggregation node is created to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and a directed edge from the shunt node to the virtual node corresponding to the aggregation node is created to obtain a new process DAG. If there are other shunt nodes in the branch of the shunt node, the shunt processing step for the shunt node in the corresponding branch is called to obtain the current process DAG, and then the corresponding aggregation node of the shunt node is determined, and the node set of the aggregation node is constructed based on the aggregation node and the nodes between it and the shunt node, so as to create a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtain a new process DAG.

[0030] like Figure 6As shown, the currently visited branch node is A. Traversing forward along its left branch, it is found that there is a branch node B. Therefore, the branch processing step is performed on the branch node B, that is, the branch of the branch node B is traversed forward to determine whether there are other branch nodes in its branches. If there are, the branch processing step is continued to be recursively called to process the branch nodes on its branches. After confirming that there are no branch nodes in its branches, the aggregation node G corresponding to the branch node B is determined, and the node set {C, D, E, F, G} of the aggregation node is constructed based on the aggregation node G and the nodes between it and the branch node B, and a virtual node G' corresponding to the aggregation node is created to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG. Subsequently, continue to confirm whether there are other branch nodes on other branches of the branch node A. If there are, the branch processing step is recursively called to process the branch nodes on the branches. After executing the above processing on all branches of the diversion node A, there will be no more diversion nodes on the branches of the diversion node A. Therefore, we return to the diversion node A, locate its corresponding aggregation node J, build the node set {B, G', H, I, J} of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and create a virtual node J' corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtain a new process DAG, such as Figure 7 shown.

[0031] After obtaining the backbone process DAG, take the backbone process DAG as the starting point, and gradually drill down to analyze the virtual nodes in the backbone process DAG to determine the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index. In some embodiments, the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the backbone process DAG on the abnormal index can be calculated first. Subsequently, the backbone process DAG is traversed forward again. If the currently accessed node is a virtual node and there is no shunt node pointing to the node, the traversal is continued after calling the aggregation node drill-down step for the currently accessed node. If the currently accessed node is a shunt node in the initial process DAG and the node is connected to a virtual node through a directed edge, the shunt node drill-down step for the currently accessed node and its connected virtual node is called and the traversal is continued until the traversal is completed.

[0032] Among them, the method given in patent CN202110195548.6 can be used to calculate the impact of the transformation process corresponding to the initial node and any two adjacent nodes in the backbone process DAG on the abnormal index. It should be noted that for any subgraph, this method can be used to calculate the impact of the transformation process corresponding to the initial node and any two adjacent nodes of the subgraph on the end node, and will not be repeated later when related calculations are involved.

[0033] For example, calculate each adjacent node P separatelyi P i+1 The conversion rate between this cycle and the previous cycle (1≦i≦k-1), where P i P i+1 Conversion rate in this cycle F (P i ->P i+1 ) and the conversion rate of the previous cycle F '(P i ->P i+1 ) are respectively expressed as follows.

[0034]

[0035]

[0036] Among them, mi is the index of node Pi, A(mi) is the index value of node Pi in this cycle, and A'(mi) is the index value of node Pi in the previous cycle.

[0037] Calculate the impact value and impact rate of the transformation process corresponding to the initial node P1 and the next adjacent nodes on the final node mk. The impact value of the initial node P1 on mk is calculated as follows:

[0038] P1 to m k The influence is calculated as follows:

[0039] The transformation process from P1 to P2 has an effect on m k The impact value is calculated as follows:

[0040] The transformation process from P1 to P2 has an effect on m k The influence is calculated as follows:

[0041] Correspondingly, P i To P i+1 The conversion of m k The influence value and influence degree are expressed as follows:

[0042] Among them, the aggregation node drill-down step for any virtual node includes: based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG, construct each conversion branch of the aggregation node; take the index corresponding to the end node of each conversion branch of the aggregation node as the key index, calculate the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key index of the same conversion branch (the calculation method is the same as above), based on the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key index of the same conversion branch, the influence of the conversion branch end node corresponding to each key index on the virtual node, and the influence of the virtual node on the abnormal index, determine the influence of the conversion process corresponding to the initial node and adjacent nodes of each conversion branch on the abnormal index. Subsequently, confirm whether each conversion branch contains a virtual node. If each conversion branch contains a virtual node, the aggregation node drill-down step is performed for the virtual node in the corresponding conversion branch.

[0043] In some embodiments, for the initial node in any conversion branch, the product of the influence of the initial node in the conversion branch on the key indicator of the same conversion branch, the influence of the conversion branch end node corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator is calculated as the influence of the initial node of the conversion branch on the abnormal indicator; for the conversion process corresponding to the adjacent node in any conversion branch, the product of the influence of the conversion process corresponding to the adjacent node in the conversion branch on the key indicator of the same conversion branch, the influence of the conversion branch end node corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator is calculated as the influence of the conversion process corresponding to the adjacent node in the conversion branch on the abnormal indicator; wherein the influence of the conversion branch end node corresponding to any key indicator on the virtual node is the ratio of the index difference between the key indicator in the current cycle and the previous cycle to the index difference between the index associated with the aggregation node corresponding to the virtual node in the initial process DAG in the current cycle and the previous cycle.

[0044] Assume that the virtual node P' w The two transformation branches are {P h , P h+1 , ..., P h+m} and {P j , P j+1 , ..., P j+n}, neither of the two transformation branches contains a virtual node, P h+m and P j+n Converge into P w Node, that is, virtual node P' w The corresponding aggregation node is P w The indicators corresponding to the end nodes of each conversion branch are taken as key indicators (m h+m , mj+n ), calculate the impact of the initial node of each transformation branch and the transformation process corresponding to each adjacent node on the key indicators of the same branch { AND (P h ), AND (P h →P h+1 ), ..., AND (P h+m-1 →P h+m )},{ AND (P j ), AND (P j →P j+1 ), ..., AND (P j+n-1 →P j+n )}. After the previous analysis and calculation, the virtual node P' can be obtained w The impact on abnormal indicators is AND global (P' w ). Next, calculate P h+m P j+n P w The influence of the indicator change is expressed as formula (1) and formula (2). Finally, the influence of the branch node and transformation on the indicator abnormality is calculated, here P h and P h →P h+1 For example, their influence on the abnormal index is expressed as formula (3) and formula (4) respectively. Multiply the three together and use the chain rule to realize the conduction calculation of the influence, so as to obtain the influence of the transformation process corresponding to the initial node and each adjacent node in the virtual node on the abnormal index.

[0045] AND (P h+m )=(A(m h+m )-A'(m h+m )) / (A(m w )-A'(m w ))(1) AND (P j+n )=(A(m j+n )-A'(m j+n )) / (A(m w )-A'(m w ))(2) AND global (P h ) =AND (P h )× AND (P h+m )× AND global (P' w )(3) AND global (P h →P h+1 ) =AND (P h →P h+1 )× AND (P h+m )× AND global (P' w )(4) Assume that the virtual node P' w The two transformation branches are {P h , P h+1 , ..., P h+m} and {P' j , P j+1 , ..., P j+n}, and the conversion branch contains a virtual node P' j , P' j is not the end node of the transformation branch, P h+m and P j+n Converge into P w Node, that is, virtual node P' w The corresponding aggregation node is P w . P' j The influence on the abnormal index is expressed as formula (5). If the second transformation branch set only contains P' j Node P' j is the end node of the transformation branch, then P' j The impact on the abnormal index can be expressed as formula (6). j Continue drilling down and calculate P' in a similar way. j The influence of the transformation process corresponding to the initial node and adjacent nodes in the transformation branch on the abnormal index.

[0046] AND global (P' j ) =AND (P' j )× AND (P j+n )× AND global (P' w )(5) AND global (P' j ) = ((A(m' j)-A'(m' j )) / (A(m w )-A'(m w )))× AND global (P' w )(6) The shunt node drill-down step for any shunt node and its connected virtual node includes: constructing each branch of the shunt node based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG; taking the indicator corresponding to the end node of each branch of the shunt node as the key indicator, calculating the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, and determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the abnormal indicator based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator; if each branch contains a virtual node, calling the shunt node corresponding to the virtual node in the corresponding branch in the initial process DAG and the shunt node drill-down step of the virtual node.

[0047] In some embodiments, for the diversion node, the influence of the diversion node on the key indicator in any branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the product of the influence of the transformation process corresponding to the diversion node and the virtual node on the abnormal indicator are calculated as the influence of the diversion node on the abnormal indicator; for the transformation process corresponding to the diversion node and the initial node of each branch and any transformation process among the transformation processes corresponding to the adjacent nodes of each branch, the influence of the transformation process on the key indicator of the same branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the product of the influence of the transformation process corresponding to the diversion node and the virtual node on the abnormal indicator are calculated as the influence of the transformation process on the abnormal indicator.

[0048] Assume that the split node is P g , whose two branches are {P h , P h+1 , ..., P h+m} and {P j , P j+1 , ..., P j+n}, neither branch contains a virtual node, P h+m and P j+n Converge into P wNode, its corresponding virtual node is P' w The index corresponding to the end node of each branch is taken as the key index (m h+m , m j+n ), calculate the shunt node P g The impact of each stage transformation (i.e., the transformation process corresponding to the branch node and the initial node of the branch and the transformation process corresponding to the adjacent nodes in the branch) on the key indicators of the same branch { AND (P g ), AND (P g →P h ), AND (P h →P h+1 ), ..., AND (P h+m-1 →P h+m )},{ AND (P g ), AND (P g →P j ), AND (P j →P j+1 ), ..., AND (P j+n-1 →P j+n )}. After the previous calculation and analysis, the diversion node P can be obtained g To virtual node P' w The influence of the corresponding transformation process on the abnormal index is AND global (P g →P' w ). Next, calculate P h+m P j+n P w The influence of the indicator change is the same as the above formula (1) and formula (2). Finally, the influence of the diversion node and each stage transformation on the abnormal indicator is calculated. Here, P g →P h and P h →P h+1 For example, their influence on the abnormal index is expressed as formula (7) and formula (8) respectively. Multiplying the three together, the influence conduction calculation is also realized through the chain rule, so as to obtain the influence of each stage transformation in the diversion node and virtual node on the abnormal index.

[0049] AND global (P g →P h ) =AND (P g →Ph )× AND (P h+m )× AND global (P g →P' w )(7) AND global (P h →P h+1 ) =AND (P h →P h+1 )× AND (P h+m )× AND global (P g →P' w )(8) Assume that the split node is P g , the two branches are {P h , P h+1 , ..., P h+m} and {P j , P' j+1 , ..., P j+n}, and the branch contains a virtual node P' j+1 , P j Also a diversion node, P h+m and P j+n Converge into P w Node, which corresponds to the virtual node P' w .P j →P' j+1 The influence on the abnormal index is expressed as formula (9). Then, for the virtual node P' in the branch j+1 , determine its corresponding diversion node P j , and for the shunt node P j and virtual node P' j+1 Execute the diversion node drill-down step. j There are still virtual nodes in the branch, and the processing method is the same as above.

[0050] AND global (P j →P' j+1 ) =AND (P j →P' j+1 )× AND (P j+n )× AND global (P g →P' w )(9) After the impact is calculated, the initial process DAG is traversed in reverse from the node corresponding to the abnormal indicator in the initial process DAG. Based on the impact of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator, the root cause location path of the abnormal indicator is determined. Among them, the initial process DAG can be traversed in reverse from the node corresponding to the abnormal index. If the node currently visited is an aggregation node, when the change direction of the index value of the currently visited node is the same as the abnormal index, the influence of the end nodes of each transformation branch of the currently visited node on the currently visited node is sorted from large to small, and the transformation branches with positive influence and greater than the preset threshold are selected in turn by a greedy strategy until the sum of the influence of the end nodes of the selected transformation branches on the currently visited node is greater than the preset threshold, and the reverse traversal is continued along the selected transformation branches; when the change direction of the index value of the currently visited node is different from the abnormal index, the influence of the end nodes of each transformation branch of the currently visited node on the currently visited node is sorted from large to small according to the absolute value, and the transformation branches with negative influence and absolute value greater than the preset threshold are selected in turn by a greedy strategy until the sum of the absolute value of the influence of the end nodes of the selected transformation branches on the currently visited node is greater than the preset threshold, and the reverse traversal is continued along the selected transformation branches; the reverse traversal is performed until a node with an in-degree of 0 is reached, thereby obtaining the root cause location path of the abnormal index. It can be seen that there can be multiple root cause location paths for abnormal indicators. In this regard, after obtaining multiple root cause location paths, the root cause discrimination of each root cause location path can be calculated, and the root cause location path with the largest root cause discrimination can be selected as the most significant root cause location path; The root cause discrimination of any root cause location path i is calculated based on the following method: C(i)=α× +β×

[0051] Here, C(i) is the root cause discrimination of the root cause location path i; α and β are custom coefficients, and α+β=1; AND global (v) The influence of the transformation process corresponding to the initial node or any adjacent node in the root cause location path i on the abnormal index; standard(s) is the standard deviation of the influence of the transformation process corresponding to the initial node and each adjacent node in the root cause location path i on the abnormal index, and k is the number of root cause location paths.

[0052] After determining the root cause location path of the abnormal indicator, we can also build a root cause diagnosis question-and-answer system based on the big model and GraphRAG (Graph-based Retrieval Augmented Generation, text generation based on knowledge graph retrieval enhancement) technology, combined with the abnormal record document, and further analyze the reasons for the abnormality of the stage conversion through the root cause diagnosis question-and-answer system. Specifically, the abnormal performance of the abnormal indicator and the root cause information after manual determination can be recorded to form an abnormal record document. Subsequently, GraphRAG technology is used to extract information such as entities, relationships, and covariates from the abnormal record document, generate communities and community summaries, and construct a knowledge graph. When a user makes a root cause diagnosis query, the root cause diagnosis question and context information (including the root cause path and indicator information related to a stage conversion with an abnormality) input by the user to the root cause diagnosis question and answer system are received. The system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the big language model at the back end. The big language model summarizes the input and further feeds back the results to the user to determine the root cause of the stage conversion with an abnormality. Subsequently, according to the actual situation, the abnormal record documents are added or updated to complete the update of the knowledge graph. The big language model retrieves the updated knowledge graph, and the question and answer results of the root cause diagnosis will be more accurate and comprehensive.

[0053] In summary, the method provided by the embodiment of the present invention aims at the problems of a large number of nodes in a complex business process, a complex DAG graph structure, various node types, inconsistent indicator granularity of different nodes, and great difficulty in root cause analysis. The process DAG graphs with different abstraction levels are constructed from the bottom up in sequence, and the influence of operations such as transformation, diversion, and aggregation at each stage on the abnormal indicators is calculated layer by layer from high to low according to the abstraction level. The influences are comparable to each other, which can help operators quickly identify and locate the root cause path and individual stage transformation items with large influences. The conduction calculation of the influence is realized by the chain rule, thereby realizing cross-layer stage transformation analysis. In addition, a root cause diagnosis question and answer system can be constructed based on a large model + graph retrieval generation technology and combined with anomaly record documents. The root cause diagnosis question and answer system further analyzes the causes of abnormalities caused by stage transformation, effectively utilizes past diagnosis experience, and improves the analysis efficiency of stage transformation abnormalities. Subsequently, according to actual conditions, the abnormal record documents are added or updated, and then the knowledge graph is updated, so that the large language model retrieves the updated knowledge graph, and the root cause diagnosis question and answer results will be more accurate and comprehensive.

[0054] The root cause location system for indicator anomalies for complex processes provided by the present invention is described below. The root cause location system for indicator anomalies for complex processes described below and the root cause location method for indicator anomalies for complex processes described above can be referenced to each other.

[0055] Based on any of the above embodiments, Figure 8 : is a schematic diagram of the structure of the root cause location system for abnormal indicators of complex processes provided by the present invention, such as Figure 8 As shown, the system includes: The initial DAG construction module 810 is used to construct an initial process DAG based on the current business process and establish the association between each node in the initial process DAG and its indicator; the node of the initial process DAG corresponds to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; The DAG abstraction module 820 is used to abstract the aggregation nodes and the shunt nodes in the initial process DAG to obtain the backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the shunt node is greater than 1; An influence calculation module 830 is used to determine the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; A root cause location path output module 840 is used to reversely traverse the initial process DAG from the node corresponding to the abnormal indicator in the initial process DAG, and determine the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; The root cause diagnosis question and answer system module 850 is used to record the abnormal performance of the abnormal indicators and the root cause information after manual determination to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user performs a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage transformation in which an abnormality exists; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage transformation in which the abnormality exists.

[0056] The system provided by the embodiment of the present invention aims at the problems of a large number of nodes in a complex business process, a complex DAG graph structure, various node types, inconsistent indicator granularity of different nodes, and great difficulty in root cause analysis. The system constructs process DAG graphs with different abstraction levels from bottom to top, and calculates the influence of operations such as transformation, diversion, and aggregation at each stage on abnormal indicators layer by layer from high to low according to the abstraction level. The influences are comparable to each other, which can help operators quickly identify and locate the root cause path and individual stage transformation items with large influences. The conduction calculation of the influence is realized by the chain rule, thereby realizing cross-layer stage transformation analysis. In addition, a root cause diagnosis question-and-answer system can be constructed based on a large model + graph retrieval generation technology and combined with abnormal record documents. The root cause diagnosis question-and-answer system further analyzes the causes of abnormalities caused by stage transformation, effectively utilizes past diagnosis experience, and improves the analysis efficiency of stage transformation abnormalities. Subsequently, according to actual conditions, the abnormal record documents are added or updated, and the knowledge graph is updated, so that the large language model retrieves the updated knowledge graph, and the root cause diagnosis question-and-answer results will be more accurate and comprehensive.

[0057] Based on any of the above embodiments, the aggregation nodes and the diversion nodes in the initial process DAG are abstracted to obtain the backbone process DAG, including: Starting forward traversal from the node with in-degree 0 in the initial process DAG, if the currently visited node is an aggregation node, then invoking the aggregation processing step for the currently visited node and continuing traversal; if the currently visited node is a diversion node, then invoking the diversion processing step for the currently visited node and continuing traversal; after the traversal is completed, the backbone process DAG is obtained; Among them, the aggregation processing step for any aggregation node includes: reverse traversing the unvisited transformation branches of the aggregation node, if any transformation branch contains the aggregation node, then calling the aggregation processing step for the aggregation node in the transformation branch to obtain the current process DAG, building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; if all transformation branches do not contain the aggregation node, then building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from the transformation branch of any node to the node; The diversion processing step for any diversion node includes: forward traversing the branches of the diversion node, if there is no diversion node in the branch of the diversion node, then determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, creating a directed edge from any diversion node to the virtual node corresponding to the aggregation node, and obtaining a new process DAG; otherwise, calling the diversion processing step for the diversion node in the branch of the diversion node to obtain the current process DAG, determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtaining a new process DAG; there is a directed edge from any node to the branch of this node.

[0058] Based on any of the above embodiments, determining the influence of the transformation processes corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG includes: Calculate the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the backbone process DAG on the abnormal index; Traverse the backbone process DAG forward. If the currently visited node is a virtual node and there is no shunt node pointing to the node, call the aggregation node drill-down step for the currently visited node and continue traversing. If the currently visited node is a shunt node in the initial process DAG and the node is connected to a virtual node through a directed edge, call the shunt node drill-down step for the currently visited node and its connected virtual node and continue traversing. Among them, the aggregation node drill-down step for any virtual node includes: based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG, construct each conversion branch of the aggregation node; taking the indicator corresponding to the end node of each conversion branch of the aggregation node as the key indicator, calculating the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, based on the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, the influence of the end node of each conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, determine the influence of the conversion process corresponding to the initial node and adjacent nodes of each conversion branch on the abnormal indicator; if each conversion branch contains a virtual node, call the aggregation node drill-down step for the virtual node in the corresponding conversion branch; The shunt node drill-down step for any shunt node and its connected virtual node includes: constructing each branch of the shunt node based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG; taking the indicator corresponding to the end node of each branch of the shunt node as the key indicator, calculating the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, and determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the abnormal indicator based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator; if each branch contains a virtual node, calling the shunt node corresponding to the virtual node in the corresponding branch in the initial process DAG and the shunt node drill-down step of the virtual node.

[0059] Based on any of the above embodiments, based on the influence of the conversion process corresponding to the initial node and the adjacent nodes in each conversion branch on the key indicator, the influence of the end node of each conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, determining the influence of the conversion process corresponding to the initial node and the adjacent nodes of each conversion branch on the abnormal indicator, including: For the initial node in any conversion branch, calculate the product of the influence of the initial node in the conversion branch on the key indicator of the same conversion branch, the influence of the final node of the conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the initial node of the conversion branch on the abnormal indicator; For the transformation process corresponding to the adjacent nodes in any transformation branch, calculate the product of the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the key indicator of the same transformation branch, the influence of the transformation branch end node corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the abnormal indicator; The influence of the final node of the transformation branch corresponding to any key indicator on the virtual node is the ratio of the difference between the key indicator in the current cycle and the previous cycle to the difference between the indicator associated with the aggregation node corresponding to the virtual node in the initial process DAG in the current cycle and the previous cycle.

[0060] Based on any of the above embodiments, based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator, determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the abnormal indicator, including: For the shunt node, calculate the product of the influence of the shunt node on the key indicator in any branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the conversion process corresponding to the shunt node and the virtual node on the abnormal indicator as the influence of the shunt node on the abnormal indicator; For any transformation process among the transformation processes corresponding to the diversion node and the initial nodes of each branch and the transformation processes corresponding to the adjacent nodes of each branch, calculate the product of the influence of the transformation process on the key indicator of the same branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the diversion node and the virtual node on the abnormal indicator as the influence of the transformation process on the abnormal indicator.

[0061] Based on any of the above embodiments, starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse order, and determining the root cause location path of the abnormal indicator based on the influence of the initial node and the transformation processes corresponding to any two adjacent nodes of the initial process DAG on the abnormal indicator, including: The initial process DAG is traversed in reverse from the node corresponding to the abnormal index. If the node currently visited is an aggregation node, when the change direction of the index value of the node currently visited is the same as the abnormal index, the influence of the end nodes of each conversion branch of the node currently visited on the node currently visited is sorted from large to small, and the conversion branches with positive influence and greater than the preset threshold are selected in turn by a greedy strategy until the sum of the influence of the end nodes of the selected conversion branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected conversion branches; when the change direction of the index value of the node currently visited is different from the abnormal index, the influence of the end nodes of each conversion branch of the node currently visited on the node currently visited is sorted from large to small according to the absolute value, and the conversion branches with negative influence and absolute value greater than the preset threshold are selected in turn by a greedy strategy until the sum of the absolute value of the influence of the end nodes of the selected conversion branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected conversion branches; the reverse traversal is performed until a node with an in-degree of 0 is reached, and the root cause location path of the abnormal index is obtained.

[0062] Based on any of the above embodiments, after determining the root cause location path of the abnormal indicator, the root cause location path output module is further used to: Calculate the root cause discrimination of each root cause location path, and select the root cause location path with the largest root cause discrimination as the most significant root cause location path; The root cause discrimination of any root cause location path i is calculated based on the following method: C(i)=α× +β×

[0063] Where C(i) is the root cause discrimination of the root cause location path i; α and β are custom coefficients, and α+β=1; AND global (v) The influence of the transformation process corresponding to the initial node or any adjacent node in the root cause location path i on the abnormal index; standard(s) is the standard deviation of the influence of the transformation process corresponding to the initial node and each adjacent node in the root cause location path i on the abnormal index, and k is the number of root cause location paths.

[0064] Figure 9 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 9As shown, the electronic device may include: a processor (processor) 910, a memory (memory) 920, a communication interface (Communications Interface) 930 and a communication bus 940, wherein the processor 910, the memory 920, and the communication interface 930 communicate with each other through the communication bus 940. The processor 910 can call the logic instructions in the memory 920 to execute the root cause location method of the indicator abnormality for complex processes, which includes: constructing an initial process DAG based on the current business process and establishing the association between each node in the initial process DAG and its indicator; the node of the initial process DAG corresponds to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; abstracting the aggregation nodes and the diversion nodes in the initial process DAG to obtain the backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; based on the backbone process DAG, determining the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal indicator; starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse, based on the initial node and any two adjacent nodes corresponding to the initial process DAG The influence of the transformation process on the abnormal indicator is determined to determine the root cause location path of the abnormal indicator; the abnormal performance of the abnormal indicator and the root cause information after manual determination are recorded to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user performs a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and indicator information related to a stage conversion with an abnormality; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage conversion with the abnormality.

[0065] In addition, the logic instructions in the above-mentioned memory 920 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0066] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the root cause location method for indicator anomalies for complex processes provided by the above methods, the method including: constructing an initial process DAG based on the current business process and establishing an association between each node in the initial process DAG and its indicator; the nodes of the initial process DAG correspond to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; abstracting the aggregation nodes and diversion nodes in the initial process DAG to obtain a backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; based on the backbone process DAG, determining the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal indicator; starting from the node corresponding to the abnormal indicator in the initial process DAG, reverse traversal The initial process DAG determines the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; records the abnormal performance of the abnormal indicator and the root cause information after manual determination to form an abnormal record document; uses GraphRAG technology to extract entity and relationship information from the abnormal record document, generates communities and community summaries, and constructs a knowledge graph; when a user performs a root cause diagnosis query, receives the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system; the context information includes the root cause path and indicator information related to a stage conversion with an abnormality; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the result to the user, and determines the root cause generated by the stage conversion with the abnormality.

[0067] On the other hand, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it is implemented to execute the root cause location method of the indicator abnormality for the complex process provided above, the method comprising: constructing an initial process DAG based on the current business process and establishing an association between each node in the initial process DAG and its indicator; the nodes of the initial process DAG correspond to the business stage or task unit where the abnormal indicator is located and the business stage or task unit before it; abstracting the aggregation nodes and the diversion nodes in the initial process DAG to obtain a backbone process DAG; the in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; determining the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal indicator based on the backbone process DAG; starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse, based on the initial process DAG The influence of the transformation process corresponding to the initial node of AG and any two adjacent nodes on the abnormal index is used to determine the root cause location path of the abnormal index; the abnormal performance of the abnormal index and the root cause information after manual determination are recorded to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when the user performs a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage transformation with an abnormality; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage transformation with the abnormality.

[0068] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0069] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating the root cause of abnormal indicators in complex processes, characterized in that: include: Construct an initial process DAG based on the current business process and establish a relationship between each node in the initial process DAG and its indicator; The nodes of the initial process DAG correspond to the business phase or task unit where the abnormal indicator is located and the business phase or task unit before it; Abstracting the aggregation nodes and the diversion nodes in the initial process DAG to obtain the backbone process DAG; The in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; Determine the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; Starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse order, and determining the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; The abnormal performance of the abnormal indicators and the root cause information after manual determination are recorded to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user performs a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage conversion in which an abnormality exists; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage conversion in which the abnormality exists.

2. The root cause location method for abnormal indicators of complex processes according to claim 1 is characterized in that: Abstract the aggregation nodes and diversion nodes in the initial process DAG to obtain the backbone process DAG, including: Starting forward traversal from the node with in-degree 0 in the initial process DAG, if the currently visited node is an aggregation node, then invoking the aggregation processing step for the currently visited node and continuing traversal; if the currently visited node is a diversion node, then invoking the diversion processing step for the currently visited node and continuing traversal; after the traversal is completed, the backbone process DAG is obtained; Among them, the aggregation processing step for any aggregation node includes: reverse traversing the unvisited transformation branches of the aggregation node, if any transformation branch contains the aggregation node, then calling the aggregation processing step for the aggregation node in the transformation branch to obtain the current process DAG, building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; if all transformation branches do not contain the aggregation node, then building a node set of the aggregation node based on the nodes in the aggregation node and its transformation branches in the current process DAG, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, to obtain a new process DAG; there is a directed edge from the transformation branch of any node to the node; The diversion processing step for any diversion node includes: forward traversing the branches of the diversion node, if there is no diversion node in the branch of the diversion node, then determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, creating a directed edge from any diversion node to the virtual node corresponding to the aggregation node, and obtaining a new process DAG; otherwise, calling the diversion processing step for the diversion node in the branch of the diversion node to obtain the current process DAG, determining the aggregation node corresponding to the diversion node, building the node set of the aggregation node based on the aggregation node and the nodes between it and the diversion node, and creating a virtual node corresponding to the aggregation node to replace the subgraph corresponding to the node set of the aggregation node in the current process DAG, and obtaining a new process DAG; there is a directed edge from any node to the branch of this node.

3. The root cause location method for abnormal indicators of complex processes according to claim 2 is characterized in that: The determining, based on the backbone process DAG, the influence of the transformation processes corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index includes: Calculate the influence of the transformation process corresponding to the initial node and any two adjacent nodes in the backbone process DAG on the abnormal index; Traverse the backbone process DAG forward. If the currently visited node is a virtual node and there is no shunt node pointing to the node, call the aggregation node drill-down step for the currently visited node and continue traversing. If the currently visited node is a shunt node in the initial process DAG and the node is connected to a virtual node through a directed edge, call the shunt node drill-down step for the currently visited node and its connected virtual node and continue traversing. Among them, the aggregation node drill-down step for any virtual node includes: based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG, construct each conversion branch of the aggregation node; taking the indicator corresponding to the end node of each conversion branch of the aggregation node as the key indicator, calculating the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, based on the influence of the conversion process corresponding to the initial node and adjacent nodes in each conversion branch on the key indicator, the influence of the end node of each conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, determine the influence of the conversion process corresponding to the initial node and adjacent nodes of each conversion branch on the abnormal indicator; if each conversion branch contains a virtual node, call the aggregation node drill-down step for the virtual node in the corresponding conversion branch; The shunt node drill-down step for any shunt node and its connected virtual node includes: constructing each branch of the shunt node based on the node set of the aggregation node corresponding to the virtual node in the initial process DAG; taking the indicator corresponding to the end node of each branch of the shunt node as the key indicator, calculating the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, and determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the abnormal indicator based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial node of each branch, and the transformation process corresponding to the adjacent node of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator; if each branch contains a virtual node, calling the shunt node corresponding to the virtual node in the corresponding branch in the initial process DAG and the shunt node drill-down step of the virtual node.

4. The root cause location method for abnormal indicators of complex processes according to claim 3 is characterized in that: Based on the influence of the conversion process corresponding to the initial node and the adjacent nodes in each conversion branch on the key indicator, the influence of the final node of each conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, determining the influence of the conversion process corresponding to the initial node and the adjacent nodes of each conversion branch on the abnormal indicator, including: For the initial node in any conversion branch, calculate the product of the influence of the initial node in the conversion branch on the key indicator of the same conversion branch, the influence of the final node of the conversion branch corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the initial node of the conversion branch on the abnormal indicator; For the transformation process corresponding to the adjacent nodes in any transformation branch, calculate the product of the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the key indicator of the same transformation branch, the influence of the transformation branch end node corresponding to the key indicator on the virtual node, and the influence of the virtual node on the abnormal indicator, as the influence of the transformation process corresponding to the adjacent nodes in the transformation branch on the abnormal indicator; The influence of the final node of the transformation branch corresponding to any key indicator on the virtual node is the ratio of the difference between the key indicator in the current cycle and the previous cycle to the difference between the indicator associated with the aggregation node corresponding to the virtual node in the initial process DAG in the current cycle and the previous cycle.

5. The root cause location method for abnormal indicators of complex processes according to claim 4 is characterized in that: Based on the influence of the shunt node, the transformation process corresponding to the shunt node and the initial nodes of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the key indicator, the influence of the end nodes of each branch corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the shunt node and the virtual node on the abnormal indicator, determining the influence of the shunt node, the transformation process corresponding to the shunt node and the initial nodes of each branch, and the transformation process corresponding to the adjacent nodes of each branch on the abnormal indicator, including: For the shunt node, calculate the product of the influence of the shunt node on the key indicator in any branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the conversion process corresponding to the shunt node and the virtual node on the abnormal indicator as the influence of the shunt node on the abnormal indicator; For any transformation process among the transformation processes corresponding to the diversion node and the initial nodes of each branch and the transformation processes corresponding to the adjacent nodes of each branch, calculate the product of the influence of the transformation process on the key indicator of the same branch, the influence of the branch end node corresponding to the key indicator on the virtual node, and the influence of the transformation process corresponding to the diversion node and the virtual node on the abnormal indicator as the influence of the transformation process on the abnormal indicator.

6. The root cause location method for abnormal indicators of complex processes according to claim 1 is characterized in that: Starting from the node corresponding to the abnormal indicator in the initial process DAG, traversing the initial process DAG in reverse order, and determining the root cause location path of the abnormal indicator based on the influence of the initial node and the transformation processes corresponding to any two adjacent nodes of the initial process DAG on the abnormal indicator, including: The initial process DAG is traversed in reverse from the node corresponding to the abnormal index. If the node currently visited is an aggregation node, when the change direction of the index value of the node currently visited is the same as the abnormal index, the influence of the end nodes of each transformation branch of the node currently visited on the node currently visited is sorted from large to small, and the transformation branches with positive influence and greater than the preset threshold are selected in turn by a greedy strategy, until the sum of the influence of the end nodes of the selected transformation branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected transformation branches; when the change direction of the index value of the node currently visited is different from the abnormal index, the influence of the end nodes of each transformation branch of the node currently visited on the node currently visited is sorted from large to small according to the absolute value, and the transformation branches with negative influence and absolute value greater than the preset threshold are selected in turn by a greedy strategy, until the sum of the absolute value of the influence of the end nodes of the selected transformation branches on the node currently visited is greater than the preset threshold, and the reverse traversal is continued along the selected transformation branches; the reverse traversal is performed until a node with an in-degree of 0 is reached, and the root cause location path of the abnormal index is obtained.

7. The root cause location method for abnormal indicators of complex processes according to claim 6 is characterized in that: Determine the root cause location path of the abnormal indicator, and then also include: Calculate the root cause discrimination of each root cause location path, and select the root cause location path with the largest root cause discrimination as the most significant root cause location path; The root cause discrimination of any root cause location path i is calculated based on the following method: C(i)=α× +β× Where C(i) is the root cause discrimination of the root cause location path i; α and β are custom coefficients, and α+β=1; IR global (v) The influence of the transformation process corresponding to the initial node or any adjacent node in the root cause location path i on the abnormal index; std(i) is the standard deviation of the influence of the transformation process corresponding to the initial node and each adjacent node in the root cause location path i on the abnormal index, and k is the number of root cause location paths.

8. A root cause location system for abnormal indicators of complex processes, characterized by: include: An initial DAG construction module is used to construct an initial process DAG based on the current business process and establish an association between each node in the initial process DAG and its indicator; The nodes of the initial process DAG correspond to the business phase or task unit where the abnormal indicator is located and the business phase or task unit before it; A DAG abstraction module is used to abstract the aggregation nodes and the diversion nodes in the initial process DAG to obtain the backbone process DAG; The in-degree of the aggregation node is greater than 1, and the out-degree of the diversion node is greater than 1; An influence degree calculation module, used to determine the influence degree of the transformation process corresponding to the initial node and any two adjacent nodes in the initial process DAG on the abnormal index based on the backbone process DAG; A root cause location path output module is used to reversely traverse the initial process DAG from the node corresponding to the abnormal indicator in the initial process DAG, and determine the root cause location path of the abnormal indicator based on the influence of the transformation process corresponding to the initial node and any two adjacent nodes of the initial process DAG on the abnormal indicator; The root cause diagnosis question and answer system module is used to record the abnormal performance of the abnormal indicators and the root cause information after manual determination to form an abnormal record document; the GraphRAG technology is used to extract entity and relationship information from the abnormal record document, generate communities and community summaries, and construct a knowledge graph; when a user makes a root cause diagnosis query, the root cause diagnosis question and context information input by the user to the root cause diagnosis question and answer system are received; the context information includes the root cause path and the indicator information related to a stage transformation with an abnormality; the system searches the knowledge graph based on the root cause diagnosis question and context information to obtain the community summary with the highest semantic similarity to the query, and inputs the matching community summary in the knowledge graph as a prompt to the large language model at the back end; the large language model summarizes the input, further feeds back the results to the user, and determines the root cause of the stage transformation with the abnormality.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for locating the root cause of indicator anomalies for complex processes as described in any one of claims 1 to 7 is implemented.

Images