Method and system for extracting system dynamics causal relationship diagram from process log data

Abstract

The present invention provides a method and system for extracting a system dynamics causal relationship diagram from process log data, including: reading a process node log table; obtaining a process node relationship table; matching each cause attribute indicator and each result attribute indicator one by one, performing mining analysis, and obtaining a set of attribute indicator causal matching pairs with enhanced causal relationships; calculating correlation coefficients, and drawing a causal relationship diagram. By analyzing the process node log, causal matching pairs of attribute indicators with causal relationships are mined; then a causal relationship identification method is used to remove causal matching pairs of attribute indicators with weak causal relationships, and causal matching pairs of attribute indicators with enhanced causal relationships are obtained, and then the correlation coefficients reflecting the strength of the causal relationship between the attribute indicators are calculated and converted into a causal relationship diagram. Therefore, a causal relationship diagram with a strong causal relationship can be efficiently established based on the process node log to meet the needs of business process analysis and system dynamics analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of data mining, and in particular relates to a method and system for extracting a system dynamics causal relationship diagram from process log data. Background Art

[0002] Process mining is a specialized data mining technique that uses process log data to analyze business processes. System dynamics is a modeling methodology used to understand the behavior and feedback loops of complex systems. By combining system dynamics with process mining, we can further analyze business processes and create causal diagrams with strong cause-and-effect relationships to guide and predict actual business processes.

[0003] In the existing technology, manual methods are mainly used to select causal variables to establish a causal relationship diagram. Since the business process involves a large number of process nodes, each process node has a large number of attribute variables. Therefore, there are problems such as low efficiency in obtaining the causal relationship diagram and too many variables in the causal relationship diagram, which makes it impossible to specifically reflect the causal relationship diagram with strong causal relationships. Summary of the Invention

[0004] In view of the defects of the existing technology, the present invention provides a method and system for extracting system dynamics causal relationship diagrams from process log data, which can effectively solve the above problems.

[0005] The technical solution adopted in the present invention is as follows:

[0006] The present invention provides a method for extracting a system dynamics causal relationship diagram from process log data, comprising the following steps:

[0007] Step 1: Read a process node log table from a process mining log database; the process node log table is used to record the process node name, process node occurrence time, process node attribute index, and attribute index value of each process node attribute index;

[0008] Step 2: Mining and analyzing the process node log table to obtain a process node relationship table; each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumed by the current process node, the attribute indicator set of the current process node, the name of the next-level directly associated process node, and the attribute indicator set of the next-level directly associated process node; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the next-level directly associated process node is called a result attribute indicator set, including multiple result attribute indicators arranged in chronological order;

[0009] Step 3: Analyze each process node relationship in the process node relationship table, and achieve one-to-one matching between each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, thereby obtaining a set of attribute indicator causal matching pairs corresponding to each process node relationship; the set of attribute indicator causal matching pairs corresponding to each process node relationship constitutes an attribute indicator causal matching pair set;

[0010] Step 4: Mining and analyzing the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtaining the causal delay time of each attribute indicator causal matching pair, i.e., the delay time of the result attribute indicator relative to the cause attribute indicator;

[0011] Step 5: Calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 based on the causal delay time of each attribute indicator causal matching pair and the process node log table;

[0012] Step 6: Based on the causal matching relationship of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationships obtained in Step 4, a causal relationship graph is drawn to show the causal relationship of each attribute indicator. Each vertex in the causal relationship graph represents an attribute indicator in each attribute indicator causal matching pair. The edges between two adjacent vertices with direct causal relationships in the causal relationship graph are assigned corresponding correlation coefficients.

[0013] Step 7: Based on the cause-effect relationship diagram, conduct cause-effect relationship analysis and system dynamics analysis on the business process.

[0014] Preferably, step 3 is specifically:

[0015] Step 3.1, assume that there are N process node relationships in the process node relationship table;

[0016] Step 3.2, for the i-th process node relationship, i = 1, 2, ..., N, its average time is t i , the cause attribute indicator set is expressed as: A i =(a i1 ,a i2 ,…,a im ), the resulting attribute index set is expressed as: A j =(a j1 ,a j2 ,…,a jn ), where m represents the number of cause attribute indicators in the cause attribute indicator set, and n represents the number of result attribute indicators in the result attribute indicator set;

[0017] The cause attribute index set Ai =(a i1 ,a i2 ,…,a im ) in each cause attribute index, and the result attribute index set A j =(a j1 ,a j2 ,…,a jn ) are matched with each result attribute indicator in pairs to obtain m*n pairs of attribute indicator causal matching pairs corresponding to the relationship of the i-th process node, thereby forming a subset of attribute indicator causal matching pairs;

[0018] Step 3.3: After the relationship traversal of N process nodes is completed, all attribute indicator causal matching pair subsets are obtained to form an attribute indicator causal matching pair set.

[0019] Preferably, step 4 is specifically as follows:

[0020] Step 4.1, suppose that the attribute index causal matching pair set has a total of M attribute index causal matching pairs, which are expressed as: C1=(a1,b1), C2=(a2,b2),…,C M =(a M ,b M );

[0021] Pre-establish an initially empty attribute indicator causal matching pair record table;

[0022] Traverse and analyze each attribute indicator causal matching pair in turn, and for the current attribute indicator causal matching pair C k =(a k ,b k ), where k = 1, 2, M, and the following judgment and analysis of the first and second removal rules are performed:

[0023] Removal Rule 1: Self-referential removal rule:

[0024] Judgment cause attribute index a k Is it equal to the result attribute index b k If it is equal, delete the attribute indicator causal matching pair C k =(a k ,b k ), not recorded in the attribute indicator causal matching pair record table; if not equal, the second removal rule is judged:

[0025] Second removal rule: Unique matching rule for attribute indicators belonging to the same process node relationship:

[0026] Further obtain the current attribute index causal matching pair C k =(a k ,bk ) in the process node relationship, and judge whether the attribute indicator causal matching record table has been recorded in the same process node relationship, and is about the cause attribute indicator a k Or result attribute index b k If there is a causal matching pair of attribute indicators, use the current attribute indicator causal matching pair C k =(a k ,b k ), update the attribute index causal matching record table about the cause attribute index a k Or result attribute index b k causal matching pairs of attribute indicators are obtained, thereby ensuring that for the same process node relationship, there are no cases where multiple cause attribute indicators correspond to the same result attribute indicator, and there is no case where the same cause attribute indicator corresponds to multiple result attribute indicators. Based on the last interaction principle, only the attribute indicator causal matching pair that appears last is retained as the attribute indicator causal matching pair with the strongest causal relationship;

[0027] Step 4.2: After executing the first and second removal rules for all attribute indicator causal matching pairs in the attribute indicator causal matching pair set, record each attribute indicator causal matching pair C in the attribute indicator causal matching pair record table. k =(a k ,b k ), increase the causal delay time t k , where the causal delay time t k The average time consumed by the process node relationship where the causal matching of this attribute indicator is located; then execute step 4.3;

[0028] Step 4.3: Analyze and determine the merge rules:

[0029] Traverse each attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k ), obtain the process node relationship where it is located, and determine whether the attribute indicator causal matching pair record table has recorded the same attribute indicator causal matching pair C belonging to different process node relationships k =(a k ,b k ), if so, execute the merge rule and update the attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k ) causal delay time t k , is the sum of the average time taken to execute the merged relationships of each process node.

[0030] Preferably, step 5 is specifically as follows:

[0031] Step 5.1: Assume that the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 includes g attribute indicator causal matching pairs, and each attribute indicator causal matching pair is C f =(a f ,b f ), the causal delay time is t f ; where f = 1, 2, …, g;

[0032] Step 5.2: Read w cause attribute indicators a within the design history time period f The historical attribute index values ​​are: a f1 ,a f2 ,…,a fw ;

[0033] The cause attribute index a f The sampling time offset t of each historical attribute indicator value f , read the result attribute index b of the corresponding historical time in sequence f The historical attribute index values ​​are: b f1 ,b f2 ,…,b fw ;

[0034] Step 5.3, for a f1 ,a f2 ,…,a fw and b f1 ,b f2 ,…,b fw Perform correlation calculation to obtain the correlation coefficient r.

[0035] Preferably, step 5.3 is specifically as follows:

[0036] Step 5.3.1, calculate the cause attribute index a f1 ,a f2 ,…,a fw median(a f );Calculate the attribute index b of the result f1 ,b f2 ,…,b fw The median(b f );

[0037] Step 5.3.2, calculate the correlation coefficient numerator as follows:

[0038] Correlation coefficient numerator = (a f1 -median(a f ))*(b f1 -median(bf ))+(a f2 -median(a f ))*(b f2 -median(b f ))+…+(a fw -median(a f ))*(b fw -median(b f ))

[0039] Step 5.3.3, calculate the denominator of the correlation coefficient as follows:

[0040] Correlation coefficient denominator = {(a f1 -median(a f )) 2 +(a f2 -median(a f )) 2 +…+(a fw -median(a f )) 2} 1 / 2 *{(b f1 -median(b f )) 2 +(b f2 -median(b f )) 2 +…+(b fw -median(b f )) 2} 1 / 2 ;

[0041] In step 5.3.4, the correlation coefficient r is obtained by the formula: correlation coefficient r = correlation coefficient numerator / correlation coefficient denominator.

[0042] The present invention also provides a system for extracting a system dynamics causal relationship diagram from process log data, comprising:

[0043] A process node log table reading module is used to read a process node log table from a process mining log database; the process node log table is used to record the process node name, process node occurrence time, process node attribute index and attribute index value of each process node attribute index of each process node;

[0044] A process node relationship table acquisition module is used to mine and analyze the process node log table to obtain a process node relationship table; each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumed by the current process node, the attribute indicator set of the current process node, the name of the directly associated process node at the next level, and the attribute indicator set of the directly associated process node at the next level; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the directly associated process node at the next level is called a result attribute indicator set, including multiple result attribute indicators arranged in chronological order;

[0045] An attribute indicator causal matching pair set acquisition module is used to analyze each process node relationship in the process node relationship table, achieve one-to-one matching of each cause attribute indicator in the cause attribute indicator set with each result attribute indicator in the result attribute indicator set in each process node relationship, and obtain an attribute indicator causal matching pair subset corresponding to each process node relationship; the attribute indicator causal matching pair subsets corresponding to each process node relationship constitute an attribute indicator causal matching pair set;

[0046] A causal relationship mining and analysis module is used to mine and analyze the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtain the causal delay time of each attribute indicator causal matching pair, that is, the delay time of the result attribute indicator relative to the cause attribute indicator;

[0047] A correlation coefficient calculation module is used to calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship based on the causal delay time of each attribute indicator causal matching pair and the process node log table;

[0048] A causal relationship graph drawing module is used to draw a causal relationship graph showing the causal relationship of each attribute indicator causal matching pair based on the causal matching relationship of each attribute indicator causal matching pair in the causal relationship enhanced attribute indicator causal matching pair set. Each vertex in the causal relationship graph represents an attribute indicator in each attribute indicator causal matching pair. The edges between two adjacent vertices with direct causal relationship in the causal relationship graph are assigned corresponding correlation coefficients.

[0049] The causal relationship analysis and system dynamics analysis modules are used to perform causal relationship analysis and system dynamics analysis on business processes based on the causal relationship diagram.

[0050] The method and system for extracting system dynamics causal relationship diagrams from process log data provided by the present invention have the following advantages:

[0051] The present invention provides a method and system for extracting a system dynamics causal relationship diagram from process log data. By analyzing process node logs, the system mines causal matching pairs of attribute indicators with causal relationships. A causal relationship identification method is then used to remove causal matching pairs with weaker causal relationships, resulting in causal matching pairs with stronger causal relationships. Correlation coefficients reflecting the strength of causal relationships between attribute indicators are then calculated and converted into a causal relationship diagram. This method effectively creates a causal relationship diagram with strong causal relationships based on process node logs, meeting the needs of business process analysis and system dynamics analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The overall flow chart of the method for extracting system dynamics causal relationship diagram from process log data provided by the present invention;

[0053] Figure 2 A process node relationship diagram generated for Example 1 of the present invention;

[0054] Figure 3 A cause-and-effect relationship diagram generated for Example 1 of the present invention;

[0055] Figure 4 A process node relationship diagram generated for Example 2 of the present invention;

[0056] Figure 5 This is a causal relationship diagram generated in Example 2 of the present invention. DETAILED DESCRIPTION

[0057] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] The present invention provides a method and system for extracting a system dynamics causal relationship diagram from process log data. By analyzing process node logs, the system mines causal matching pairs of attribute indicators with causal relationships. A causal relationship identification method is then used to remove causal matching pairs with weaker causal relationships, resulting in causal matching pairs with stronger causal relationships. Correlation coefficients reflecting the strength of causal relationships between attribute indicators are then calculated and converted into a causal relationship diagram. This method effectively creates a causal relationship diagram with strong causal relationships based on process node logs, meeting the needs of business process analysis and system dynamics analysis.

[0059] Specifically, the present invention provides a method for extracting a system dynamics causal relationship diagram from process log data, referring to Figure 1 , including the following steps:

[0060] Step 1: Read a process node log table from a process mining log database; the process node log table is used to record the process node name, process node occurrence time, process node attribute index, and attribute index value of each process node attribute index;

[0061] Step 2: Mining and analyzing the process node log table to obtain a process node relationship table; each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumed by the current process node, the attribute indicator set of the current process node, the name of the next-level directly associated process node, and the attribute indicator set of the next-level directly associated process node; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the next-level directly associated process node is called a result attribute indicator set, including multiple result attribute indicators arranged in chronological order;

[0062] Step 3: Analyze each process node relationship in the process node relationship table, and achieve one-to-one matching between each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, thereby obtaining a set of attribute indicator causal matching pairs corresponding to each process node relationship; the set of attribute indicator causal matching pairs corresponding to each process node relationship constitutes an attribute indicator causal matching pair set;

[0063] Step 3 is as follows:

[0064] Step 3.1, assume that there are N process node relationships in the process node relationship table;

[0065] Step 3.2, for the i-th process node relationship, i = 1, 2, ..., N, its average time is t i , the cause attribute indicator set is expressed as: A i =(a i1 ,a i2 ,…,a im ), the resulting attribute index set is expressed as: A j =(a j1 ,a j2 ,…,a jn ), where m represents the number of cause attribute indicators in the cause attribute indicator set, and n represents the number of result attribute indicators in the result attribute indicator set;

[0066] The cause attribute index set A i =(a i1 ,a i2 ,…,a im) in each cause attribute index, and the result attribute index set A j =(a j1 ,a j2 ,…,a jn ) are matched with each result attribute indicator in pairs to obtain m*n pairs of attribute indicator causal matching pairs corresponding to the relationship of the i-th process node, thereby forming a subset of attribute indicator causal matching pairs;

[0067] Step 3.3: After the relationship traversal of N process nodes is completed, all attribute indicator causal matching pair subsets are obtained to form an attribute indicator causal matching pair set.

[0068] Step 4: Mining and analyzing the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtaining the causal delay time of each attribute indicator causal matching pair, i.e., the delay time of the result attribute indicator relative to the cause attribute indicator;

[0069] Step 4 is as follows:

[0070] Step 4.1, suppose that the attribute index causal matching pair set has a total of M attribute index causal matching pairs, which are expressed as: C1=(a1,b1), C2=(a2,b2),…,C M =(a M ,b M );

[0071] Pre-establish an initially empty attribute indicator causal matching pair record table;

[0072] Traverse and analyze each attribute indicator causal matching pair in turn, and for the current attribute indicator causal matching pair C k =(a k ,b k ), where k = 1, 2, M, and the following judgment and analysis of the first and second removal rules are performed:

[0073] Removal Rule 1: Self-referential removal rule:

[0074] Judgment cause attribute index a k Is it equal to the result attribute index b k If it is equal, delete the attribute indicator causal matching pair C k =(a k ,b k ), not recorded in the attribute indicator causal matching pair record table; if not equal, the second removal rule is judged:

[0075] Second removal rule: Unique matching rule for attribute indicators belonging to the same process node relationship:

[0076] Further obtain the current attribute index causal matching pair C k =(a k ,b k ) in the process node relationship, and judge whether the attribute indicator causal matching record table has been recorded in the same process node relationship, and is about the cause attribute indicator a k Or result attribute index b k If there is a causal matching pair of attribute indicators, use the current attribute indicator causal matching pair C k =(a k ,b k ), update the attribute index causal matching record table about the cause attribute index a k Or result attribute index b k causal matching pairs of attribute indicators are obtained, thereby ensuring that for the same process node relationship, there are no cases where multiple cause attribute indicators correspond to the same result attribute indicator, and there is no case where the same cause attribute indicator corresponds to multiple result attribute indicators. Based on the last interaction principle, only the attribute indicator causal matching pair that appears last is retained as the attribute indicator causal matching pair with the strongest causal relationship;

[0077] Step 4.2: After executing the first and second removal rules for all attribute indicator causal matching pairs in the attribute indicator causal matching pair set, record each attribute indicator causal matching pair C in the attribute indicator causal matching pair record table. k =(a k ,b k ), increase the causal delay time t k , where the causal delay time t k The average time consumed by the process node relationship where the causal matching of this attribute indicator is located; then execute step 4.3;

[0078] Step 4.3: Analyze and determine the merge rules:

[0079] Traverse each attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k ), obtain the process node relationship where it is located, and determine whether the attribute indicator causal matching pair record table has recorded the same attribute indicator causal matching pair C belonging to different process node relationships k =(a k ,b k ), if so, execute the merge rule and update the attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k ) causal delay time t k, is the sum of the average time taken to execute the merged relationships of each process node.

[0080] Step 5: Calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 based on the causal delay time of each attribute indicator causal matching pair and the process node log table;

[0081] Step 5 is as follows:

[0082] Step 5.1: Assume that the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 includes g attribute indicator causal matching pairs, and each attribute indicator causal matching pair is C f =(a f ,b f ), the causal delay time is t f ; where f = 1, 2, …, g;

[0083] Step 5.2: Read w cause attribute indicators a within the design history time period f The historical attribute index values ​​are: a f1 ,a f2 ,…,a fw ;

[0084] The cause attribute index a f The sampling time offset t of each historical attribute indicator value f , read the result attribute index b of the corresponding historical time in sequence f The historical attribute index values ​​are: b f1 ,b f2 ,…,b fw ;

[0085] Step 5.3, for a f1 ,a f2 ,…,a fw and b f1 ,b f2 ,…,b fw Perform correlation calculation to obtain the correlation coefficient r.

[0086] Step 5.3 is as follows:

[0087] Step 5.3.1, calculate the cause attribute index a f1 ,a f2 ,…,a fw median(a f );Calculate the attribute index b of the result f1 ,b f2 ,…,b fw The median(b f );

[0088] Step 5.3.2, calculate the correlation coefficient numerator as follows:

[0089] Correlation coefficient numerator = (a f1 -median(a f ))*(b f1 -median(b f ))+(a f2 -median(a f ))*(b f2 -median(b f ))+…+(a fw -median(a f ))*(b fw -median(b f ))

[0090] Step 5.3.3, calculate the denominator of the correlation coefficient as follows:

[0091] Correlation coefficient denominator = {(a f1 -median(a f )) 2 +(a f2 -median(a f )) 2 +…+(a fw -median(a f )) 2} 1 / 2 *{(b f1 -median(b f )) 2 +(b f2 -median(b f )) 2 +…+(b fw -median(b f )) 2} 1 / 2 ;

[0092] In step 5.3.4, the correlation coefficient r is obtained by the formula: correlation coefficient r = correlation coefficient numerator / correlation coefficient denominator.

[0093] Step 6: Based on the causal matching relationship of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationships obtained in Step 4, a causal relationship graph is drawn to show the causal relationship of each attribute indicator. Each vertex in the causal relationship graph represents an attribute indicator in each attribute indicator causal matching pair. The edges between two adjacent vertices with direct causal relationships in the causal relationship graph are assigned corresponding correlation coefficients.

[0094] Step 7: Based on the cause-effect relationship diagram, conduct cause-effect relationship analysis and system dynamics analysis on the business process.

[0095] The present invention also provides a system for extracting a system dynamics causal relationship diagram from process log data, comprising:

[0096] A process node log table reading module is used to read a process node log table from a process mining log database; the process node log table is used to record the process node name, process node occurrence time, process node attribute index and attribute index value of each process node attribute index of each process node;

[0097] A process node relationship table acquisition module is used to mine and analyze the process node log table to obtain a process node relationship table; each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumed by the current process node, the attribute indicator set of the current process node, the name of the directly associated process node at the next level, and the attribute indicator set of the directly associated process node at the next level; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the directly associated process node at the next level is called a result attribute indicator set, including multiple result attribute indicators arranged in chronological order;

[0098] An attribute indicator causal matching pair set acquisition module is used to analyze each process node relationship in the process node relationship table, achieve one-to-one matching of each cause attribute indicator in the cause attribute indicator set with each result attribute indicator in the result attribute indicator set in each process node relationship, and obtain an attribute indicator causal matching pair subset corresponding to each process node relationship; the attribute indicator causal matching pair subsets corresponding to each process node relationship constitute an attribute indicator causal matching pair set;

[0099] A causal relationship mining and analysis module is used to mine and analyze the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtain the causal delay time of each attribute indicator causal matching pair, that is, the delay time of the result attribute indicator relative to the cause attribute indicator;

[0100] A correlation coefficient calculation module is used to calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship based on the causal delay time of each attribute indicator causal matching pair and the process node log table;

[0101] A causal relationship graph drawing module is used to draw a causal relationship graph showing the causal relationship of each attribute indicator causal matching pair based on the causal matching relationship of each attribute indicator causal matching pair in the causal relationship enhanced attribute indicator causal matching pair set. Each vertex in the causal relationship graph represents an attribute indicator in each attribute indicator causal matching pair. The edges between two adjacent vertices with direct causal relationship in the causal relationship graph are assigned corresponding correlation coefficients.

[0102] The causal relationship analysis and system dynamics analysis modules are used to perform causal relationship analysis and system dynamics analysis on business processes based on the causal relationship diagram.

[0103] Two embodiments are described below:

[0104] Example 1: Analysis of production and sales process:

[0105] Step 1: Read the process node log table shown in Table 1 from the process mining log database;

[0106] Table 1: Process node log table

[0107]

[0108]

[0109] Therefore, the process node log table records the process node name, occurrence time, process node attribute indicator, and attribute indicator value of each process node in the entire production and sales process. For example, for the process node "Inspecting Raw Material Quality", its occurrence time is recorded as 2022-01-03 9:00:01, its attribute indicator is the raw material defect rate, and its attribute indicator value is 80%.

[0110] Step 2: Mining and analyzing the process node log table shown in Table 1 to obtain the process node relationship table shown in Table 2;

[0111] Table 2: Process node relationship table

[0112]

[0113] Therefore, each process node relationship in the process node relationship table is used to record the process node information of each process node, including: the name of the current process node, the average time consumed by the current process node, the attribute indicator set of the current process node, the name of the process node directly associated with the next level, and the attribute indicator set of the process node directly associated with the next level; wherein, for each process node relationship, the attribute indicator set of the current process node is called the cause attribute indicator set, which includes multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the process node directly associated with the next level is called the result attribute indicator set, which includes multiple result attribute indicators arranged in chronological order;

[0114] Specifically, by analyzing the occurrence time of each process node belonging to the same case ID in Table 1, we can obtain the previous and next level association relationship of each process node and the average time consumption of each process node, thereby obtaining Table 2.

[0115] After obtaining Table 2, we can Figure 2 The process node relationship diagram is represented by .

[0116] Step 3: Analyze each process node relationship in the process node relationship table, and achieve one-to-one matching between each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, thereby obtaining a set of attribute indicator causal matching pairs corresponding to each process node relationship; the set of attribute indicator causal matching pairs corresponding to each process node relationship constitutes an attribute indicator causal matching pair set;

[0117] Step 4: Mining and analyzing the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtaining the causal delay time of each attribute indicator causal matching pair, i.e., the delay time of the result attribute indicator relative to the cause attribute indicator;

[0118] The process of step 3 and step 4 is shown in Table 3:

[0119] Table 3:

[0120]

[0121] The final set of attribute indicator causal matching pairs with enhanced causal relationships and the causal delay time of each attribute indicator causal matching pair can be expressed in Table 4:

[0122] Table 4:

[0123] Cause attribute index Result attribute indicators Causal delay time Raw material defect rate Production abnormal events 3 days Production abnormal events Worker fatigue level 0.5 days Worker fatigue level Production abnormal events 1 day Production abnormal events Finished product inventory adequacy rate 7 days Finished product inventory adequacy rate On-time delivery rate 14 days On-time delivery rate On-time payment rate 5 days On-time payment rate Cash flow adequacy ratio 10 days Cash flow adequacy ratio Supplier quality 4 days Supplier quality Raw material defect rate 6 months

[0124] Among them, in order to realize the closed-loop causal relationship, a causal matching pair between the last attribute indicator and the first attribute indicator is added, and the causal delay time is set to a preset value of 6 months.

[0125] Step 5: Based on the causal delay time of each attribute indicator causal matching pair and the process node log table shown in Table 1, calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in Step 4;

[0126] For example, for the causal matching pair of production abnormality events and finished product inventory adequacy rate, the causal delay time is 7 days. The following method is used to calculate the correlation coefficient:

[0127] 1) Obtain the number of abnormal production events in January 2022, that is, between January 1 and January 31, which is 50; delay 7 days and obtain the finished product inventory adequacy rate between January 8 and February 7, 2022, which is 90%;

[0128] The number of abnormal production events in February 2022, that is, between February 1 and February 28, is obtained as 70 times; delayed by 7 days, the finished product inventory adequacy rate between February 8 and March 7, 2022, is obtained as 80%;

[0129] And so on

[0130] The number of production abnormal events in March 2022 is 90, the delay is 7 days, and the finished product inventory adequacy rate is 76%;

[0131] Obtained the number of production abnormal events in April 2022: 120, delays of 7 days, and finished product inventory adequacy rate of 72%;

[0132] Obtain the number of production abnormal events in May 2022: 90, delays: 7 days, and finished product inventory adequacy rate: 82%;

[0133] The number of production abnormal events in June 2022 was obtained as 39, with a delay of 7 days and a finished product inventory adequacy rate of 90%.

[0134] The production abnormal event index value is obtained as follows: a=[a f1 ,a f2 ,…,a f6 ]=[50,70,90,120,90,39]; the finished product inventory adequacy ratio index value is: b=[b f1 ,b f2 ,…,b f6 ]=[90,80,76,72,82,90];

[0135] 2) Median of production abnormal events median(a) = (70 + 90) / 2 = 80

[0136] Finished goods inventory adequacy ratio median(b) = (80 + 82) / 2 = 81

[0137] 3) Correlation coefficient numerator = -1029; where:

[0138] For January: (a f1 -median(a))(b f1 -median(b))=(50-80)(90-81)=-270

[0139] For February: (a f2 -median(a))(b f2 -median(b))=(70-80)(80-81)=10

[0140] For March: (a f3 -median(a))(b f3 -median(b))=(90-80)(76-81)=-50

[0141] For April: (a f4 -median(a))(b f4 -median(b))=(120-80)(72-81)=-360

[0142] For May: (a f5 -median(a))(b f5 -median(b))=(90-80)(82-81)=10

[0143] For June: (a f6 -median(a))(b f6 -median(b))=(39-80)(90-81)=-369

[0144] Therefore, the correlation coefficient numerator = -270 + 10-50-360 + 10-369 = -1029

[0145] 4) Calculate the correlation coefficient denominator = 1099.82;

[0146] in:

[0147] {(a f1 -median(a f )) 2 +(a f2 -median(a f )) 2 +…+(a f6-median(a f )) 2} 1 / 2 =66.94;

[0148] {(b f1 -median(b f )) 2 +(b f2 -median(b f )) 2 +…+(b fw -median(b f )) 2} 1 / 2 =16.43

[0149] Therefore, the correlation coefficient denominator = 66.94*16.43 = 1099.82

[0150] 5) Correlation coefficient r = correlation coefficient numerator / correlation coefficient denominator = -1029 / 1099.82 = -0.94

[0151] This indicates a strong negative correlation between the number of production anomalies and the finished goods inventory adequacy ratio, with a correlation coefficient of approximately -0.94. Therefore, generally speaking, an increase in production anomalies will lead to a decrease in the finished goods inventory adequacy ratio, and conversely, a decrease in production anomalies will lead to an increase in the finished goods inventory adequacy ratio.

[0152] Therefore, when calculating the correlation coefficient between a pair of cause attribute indicators and result attribute indicators, if the correlation coefficient is positive, it means that the influence of the cause attribute indicator on the result attribute indicator is positively correlated; if the correlation coefficient is negative, it means that the influence of the cause attribute indicator on the result attribute indicator is negatively correlated.

[0153] Step 6: Based on the causal relationship of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4, a causal relationship graph is drawn to show the causal relationship of each attribute indicator. Each vertex in the causal relationship graph represents the attribute indicator in each attribute indicator causal matching pair. The edge between two adjacent vertices with direct causal relationship in the causal relationship graph is assigned a corresponding correlation coefficient. In this embodiment, according to Table 4, Figure 3 Cause and effect diagram shown.

[0154] Step 7: Based on the cause-effect relationship diagram, conduct cause-effect relationship analysis and system dynamics analysis on the business process.

[0155] Example 2: Process Analysis of Carbon Emissions from Group Office Printing

[0156] Step 1: Read the process node log table shown in Table 5 from the process mining log database;

[0157] Table 5: Process node log table

[0158]

[0159]

[0160] Step 2: Mining and analyzing the process node log table shown in Table 5 to obtain the process node relationship table shown in Table 6; Figure 4 The process node relationship diagram is represented by .

[0161] Table 6: Process node relationship table

[0162]

[0163] Step 3: Analyze each process node relationship in the process node relationship table, and achieve one-to-one matching between each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, thereby obtaining a set of attribute indicator causal matching pairs corresponding to each process node relationship; the set of attribute indicator causal matching pairs corresponding to each process node relationship constitutes an attribute indicator causal matching pair set;

[0164] Step 4: Mining and analyzing the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and simultaneously obtaining the causal delay time of each attribute indicator causal matching pair, i.e., the delay time of the result attribute indicator relative to the cause attribute indicator;

[0165] The process of step 3 and step 4 is shown in Table 7:

[0166] Table 7:

[0167]

[0168]

[0169] The final set of attribute index causal matching pairs with enhanced causal relationships and the causal delay time of each attribute index causal matching pair can be expressed in Table 8:

[0170] Table 8:

[0171] Cause attribute index Result attribute indicators Causal delay time Annual target sales Number of customers reached through offline activities 5 days Number of customers reached through offline activities carbon emissions 3 days Number of customers reached through offline activities Purchase quantity of paper products 3 days Purchase quantity of paper products carbon emissions 10 days carbon emissions Annual target sales 1 year

[0172] In order to realize the closed-loop causal relationship, a causal matching pair between the last attribute indicator and the first attribute indicator is added, and the causal delay time is set to a pre-value of 1 year.

[0173] Step 5: Based on the causal delay time of each attribute indicator causal matching pair and the process node log table shown in Table 1, calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in Step 4;

[0174] For example, for the causal matching pair of the attribute indicator of the number of customers reached by offline activities and carbon emissions, the causal delay time is 3 days. The following method is used to calculate the correlation coefficient:

[0175] 1) Get the historical sequence:

[0176] In 2018, offline activities reached 12,000 customers, with a delay of 3 days and carbon emissions of 192,000g.

[0177] In 2019, offline activities reached 15,000 customers, with a delay of 3 days and carbon emissions of 795,000g

[0178] In 2020, offline activities reached 50 customers, were delayed by 3 days, and had a carbon emission of 1000g.

[0179] In 2021, offline activities reached 100 customers, with a delay of 3 days and carbon emissions of 5200g

[0180] In 2022, offline activities will reach 5,000 customers, with a delay of 3 days and carbon emissions of 250,000g

[0181] In 2023, offline activities will reach 20,000 customers, with a delay of 3 days and carbon emissions of 2,000,000g

[0182] The index value of the number of customers reached by offline activities is: a=[a f1 ,a f2 ,…,a f6 ]=[12000,15000,50,100,5000,20000]; Carbon emission index value is: b=[b f1 ,b f2 ,…,b f6 ]=[192000,795000,1000,5200,250000,2000000];

[0183] 2) The calculated correlation coefficient is r=0.79.

[0184] This indicates a strong positive correlation between the number of customers reached through offline activities and carbon emissions, with a correlation coefficient of approximately 0.79. Therefore, generally speaking, an increase in the number of customers reached through offline activities leads to an increase in carbon emissions, while a decrease in the number of customers reached through offline activities leads to a decrease in carbon emissions.

[0185] Step 6, get Figure 5 Cause and effect diagram shown.

[0186] The present invention provides a method and system for extracting a system dynamics causal relationship diagram from process log data. By analyzing process node logs, the system mines causal matching pairs of attribute indicators with causal relationships. A causal relationship identification method is then used to remove causal matching pairs with weaker causal relationships, resulting in causal matching pairs with stronger causal relationships. Correlation coefficients reflecting the strength of causal relationships between attribute indicators are then calculated and converted into a causal relationship diagram. This method effectively creates a causal relationship diagram with strong causal relationships based on process node logs, meeting the needs of business process analysis and system dynamics analysis.

[0187] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for extracting a system dynamics causal relationship diagram from process log data, characterized in that: The following steps are involved: Step 1: Read the process node log table from the process mining log database; The process node log table is used to record the process node name, process node occurrence time, process node attribute index and attribute index value of each process node attribute index of each process node; Step 2, mining and analyzing the process node log table to obtain a process node relationship table; Each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumption of the current process node, the attribute indicator set of the current process node, the name of the process node directly associated with the next level, and the attribute indicator set of the process node directly associated with the next level; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the process node directly associated with the next level is a result attribute indicator set, including multiple result attribute indicators arranged in chronological order; Step 3, analyzing each process node relationship in the process node relationship table, achieving one-to-one matching of each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, and obtaining a causal matching pair subset of attribute indicators corresponding to each process node relationship; the causal matching pair subsets of attribute indicators corresponding to each process node relationship constitute a causal matching pair set of attribute indicators; Step 4, mining and analyzing the entire set of attribute indicator causal matching pairs to obtain a set of attribute indicator causal matching pairs with enhanced causal relationships, and obtaining the causal delay time of each attribute indicator causal matching pair, that is, the delay time of the result attribute indicator relative to the cause attribute indicator; Step 5, calculating the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 according to the causal delay time of each attribute indicator causal matching pair and the process node log table; Step 6, according to the causal matching relationship of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4, a causal relationship graph showing the causal relationship of each attribute indicator is drawn, wherein each vertex in the causal relationship graph represents an attribute indicator in each attribute indicator causal matching pair; the edge between two adjacent vertices with direct causal relationship in the causal relationship graph is assigned a corresponding correlation coefficient; Step 7: Based on the cause-effect diagram, conduct cause-effect analysis and system dynamics analysis on the business process.

2. The method for extracting a system dynamics causal relationship diagram from process log data according to claim 1, characterized in that: Step 3 is as follows: Step 3.1, suppose there are a total of N Relationship between process nodes; Step 3.2, for i The relationship between process nodes, i =1,2,…, N , the average time taken is t i , the cause attribute indicator set is expressed as: A i =(a i1 ,a i2 ,…,a im ), the resulting attribute indicator set is expressed as: A j =(a j1 ,a j2 ,…,a jn ),in, m The number of cause attribute indicators representing the cause attribute indicator set, n The number of result attribute indicators representing the result attribute indicator set; Set the cause attribute indicator A i =(a i1 ,a i2 ,…,a im ) and the result attribute indicator set A j =(a j1 ,a j2 ,…,a jn ) are matched with each result attribute index in pairwise order to obtain i The process node relationship corresponds to m * n Causally matching the attribute indicators to form a subset of causal matching pairs of attribute indicators; Step 3.3, N After the traversal of the relationship between the process nodes is completed, all the attribute indicator causal matching pair subsets obtained constitute the attribute indicator causal matching pair set.

3. The method for extracting a system dynamics causal relationship diagram from process log data according to claim 2, characterized in that: Step 4 is as follows: Step 4.1, suppose that the attribute index causal matching pair set has a total of M attribute index causal matching pairs, which are expressed as: C1=(a1,b1), C2=(a2,b2),…, C M =(a M ,b M ); Pre-establish an initially empty attribute indicator causal matching pair record table; Traverse and analyze each attribute index causal matching pair in turn, and for the current attribute index causal matching pair C k =(a k ,b k ), where k=1,2,M, and the following judgment and analysis of the first and second removal rules are performed: Removal rule 1: Self-referential removal rule: Judgment cause attribute index a k Is it equal to the result attribute index b k If they are equal, delete the causal matching pair C of this attribute indicator. k =(a k ,b k ), not recorded in the attribute indicator causal matching pair record table; if not equal, the second removal rule is judged: The second removal rule: the unique matching rule of attribute indicators belonging to the same process node relationship: Further obtain the current attribute index causal matching pair C k =(a k ,b k ) and determines whether the attribute indicator causal matching pair record table has recorded the same process node relationship and is about the cause attribute indicator a k or result attribute index b k If there is a causal matching pair of attribute indicators, use the current attribute indicator causal matching pair C k =(a k ,b k ), update the attribute indicator causal matching pair record table about the cause attribute indicator a k or result attribute index b k causal matching pairs of attribute indicators are obtained, thereby ensuring that for the same process node relationship, there are no multiple cause attribute indicators corresponding to the same result attribute indicator, and there is no situation where the same cause attribute indicator corresponds to multiple result attribute indicators. Based on the last interaction principle, only the last attribute indicator causal matching pair is retained as the attribute indicator causal matching pair with the strongest causal relationship; Step 4.2, after executing the first removal rule and the second removal rule for all attribute indicator causal matching pairs in the attribute indicator causal matching pair set, record each attribute indicator causal matching pair C in the attribute indicator causal matching pair record table. k =(a k ,b k ), increasing the causal delay time t k , where the causal delay time t k The average time consumption of the process node relationship where the causal matching of this attribute indicator is located; then execute step 4.3; Step 4.3, perform analysis and judgment of the merge rules: Traverse each attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k ), obtain the process node relationship where it is located, and determine whether the attribute indicator causal matching pair record table has recorded the same attribute indicator causal matching pair C belonging to different process node relationships k =(a k ,b k ), if there is, then execute the merge rule and update the attribute indicator causal matching pair C recorded in the attribute indicator causal matching pair record table k =(a k ,b k )’s causal delay time t k , is the sum of the average time consumed by each process node relationship to execute the merge.

4. The method for extracting a system dynamics causal relationship diagram from process log data according to claim 1, characterized in that: Step 5 is as follows: Step 5.1: Assume that the attribute indicator causal matching pair set with enhanced causal relationship obtained in step 4 includes g attribute indicator causal matching pairs, and each attribute indicator causal matching pair is C f =(a f ,b f ), the causal delay time is t f ; where f=1,2,…,g; Step 5.2: Read w cause attribute indicators a within the design history time period f The historical attribute index values ​​are: a f1 , a f2 ,…, a fw ; The cause attribute index a f The sampling time offset t of each historical attribute indicator value f , read the result attribute index b of the corresponding historical time in sequence f The historical attribute index values ​​are: b f1 ,b f2 ,…, b fw ; Step 5.3, for a f1 , a f2 ,…, a fw and b f1 ,b f2 ,…, b fw Perform correlation calculation and obtain the correlation coefficient r.

5. The method for extracting a system dynamics causal relationship diagram from process log data according to claim 4, characterized in that: Step 5.3 is as follows: Step 5.3.1, calculate the cause attribute index a f1 , a f2 ,…, a fw The median(a f );Calculate the attribute index b of the result f1 ,b f2 ,…, b fw The median(b f ); Step 5.3.2, calculate the correlation coefficient numerator, the method is: Sokan series numerator = (a f1 -median(a f ))*(b f1 -median(b f ))+(a f2 -median(a f ))*(b f2 - median(b f ))+…+(a fw - median(a f ))*(b fw - median(b f )) Step 5.3.3, calculate the denominator of the correlation coefficient, the method is: Sokan series denominator={(a f1 -median(a f ) ) 2 +(a f2 -median(a f )) 2 +…+(a fw - median(a f ) ) 2 } 1 / 2 *{(b f1 - median(b f ) ) 2 +(b f2 - median(b f ) ) 2 +…+(b fw - median(b f ) ) 2 } 1 / 2 ; Step 5.3.4, the correlation coefficient r is obtained by the formula: correlation coefficient r = correlation coefficient numerator / correlation coefficient denominator.

6. A system for implementing the method for extracting a system dynamics causal relationship diagram from process log data as claimed in any one of claims 1 to 5, characterized in that: include: The process node log table reading module is used to read the process node log table from the process mining log database; The process node log table is used to record the process node name, process node occurrence time, process node attribute index and attribute index value of each process node attribute index of each process node; A process node relationship table acquisition module is used to mine and analyze the process node log table to obtain a process node relationship table; Each process node relationship in the process node relationship table is used to record the process node relationship information of each process node, including: the name of the current process node, the average time consumption of the current process node, the attribute indicator set of the current process node, the name of the process node directly associated with the next level, and the attribute indicator set of the process node directly associated with the next level; wherein, for each process node relationship, the attribute indicator set of the current process node is called a cause attribute indicator set, including multiple cause attribute indicators arranged in chronological order; the attribute indicator set of the process node directly associated with the next level is a result attribute indicator set, including multiple result attribute indicators arranged in chronological order; The attribute indicator causal matching pair set acquisition module is used to analyze each process node relationship in the process node relationship table, realize one-to-one matching of each cause attribute indicator in the cause attribute indicator set and each result attribute indicator in the result attribute indicator set in each process node relationship, and obtain the attribute indicator causal matching pair subset corresponding to each process node relationship; the attribute indicator causal matching pair subsets corresponding to each process node relationship constitute the attribute indicator causal matching pair set; A causal relationship mining and analysis module is used to mine and analyze the attribute indicator causal matching pair set as a whole, obtain the attribute indicator causal matching pair set with enhanced causal relationship, and obtain the causal delay time of each attribute indicator causal matching pair, that is, the delay time of the result attribute indicator relative to the cause attribute indicator; A correlation coefficient calculation module, used to calculate the correlation coefficient of each attribute indicator causal matching pair in the attribute indicator causal matching pair set with enhanced causal relationship according to the causal delay time of each attribute indicator causal matching pair and the process node log table; A causal relationship graph drawing module is used to draw a causal relationship graph showing the causal relationship of each attribute indicator causal matching pair according to the causal matching relationship of each attribute indicator causal matching pair in the attribute indicator causal matching pair set enhanced by causal relationship, wherein each vertex in the causal relationship graph represents the attribute indicator in each attribute indicator causal matching pair; the edge between two adjacent vertices with direct causal relationship in the causal relationship graph is assigned a corresponding correlation coefficient; The causal relationship analysis and system dynamics analysis modules are used to perform causal relationship analysis and system dynamics analysis on business processes based on the causal relationship diagram.

Images