Task-Type Based Process Mining Consistency Checking Method, Medium and System

Abstract

The present invention claims protection for a method, system, and medium for process mining consistency checking based on task types, belonging to the field of data mining, and a method for evaluating the structure of a process mining model. The present invention is a consistency checking method that compares the mining model with the original model and combines task types. First, traverse the original model and the mining model to obtain the total number of tasks, visible tasks, repeated tasks, and invisible tasks in each model, then calculate the proportions of repeated tasks and invisible tasks in the model, and finally substitute them into the structural evaluation formula for calculation. The invention can effectively perform consistency checking on the mining model, thereby reasonably evaluating the structure of the model to achieve the purpose of evaluating the effectiveness of the mining algorithm.

Description

Technical Field

[0001] The present invention belongs to the field of data mining, and relates to a method for evaluating the structure of a process mining model, which involves a method for comparing a mining model with an original model and combining the consistency check of task types. Background Art

[0002] Nowadays, enterprise business processes have become increasingly complex. For enterprises, it has become increasingly important to obtain a high-quality business process model. The business data information recorded in traditional information systems can be extracted and processed to be converted into event logs that can be input into process mining methods. The purpose of Process Mining is to extract valuable information from the event logs generated by business process systems, so as to discover, monitor, and improve actual business processes. Process mining mainly has three major application fields: process discovery, consistency check, and process enhancement.

[0003] Process discovery refers to extracting a process model from the event logs generated by an information system. There are many ways to represent a process model, such as Petri nets, transition systems, workflow nets, BPMN, YAWL, etc. The graphical representation of Petri nets is intuitive and simple, and can be converted into other representation methods. Most of the existing process mining technologies use process models represented by Petri nets. Therefore, the present invention uses Petri nets to represent process models.

[0004] Gartner's report points out that the importance of consistency check is increasing. The purpose of consistency check is to evaluate the quality of a mining model and discover the deviation between the mining model and the original model, and the quality of the mining model reflects the quality of the mining algorithm.

[0005] It is also difficult to evaluate the quality of a process model related to logs. Process discovery algorithms take event logs as input and aim to output a process model that meets specific attributes, which is usually referred to as the four quality dimensions of process mining. The four dimensions of consistency check are adaptability, precision, generalization, and understandability. Among them, understandability mainly evaluates the simplicity and structure of the model. Over the years, several consistency check methods have been developed for each quality dimension. Currently, the more commonly used structural evaluation methods are a′ s and the S method. The a′ s method regards repeated tasks and invisible tasks as redundant tasks, and believes that the fewer repeated tasks and invisible tasks in the mining model, the better. However, it does not compare the mining model with the original model. The S method compares the weighted average arc connectivity of the original model and the mining model, but it does not consider the task types in the model, so it cannot effectively evaluate the structure of process models containing repeated tasks.

[0006] Publication number of the application is CN106599230A, a method and system for evaluating a distributed data mining model. The method includes: S1, generating a model evaluation module by using a specific evaluation index set; S2, based on an evaluation data set, using the model evaluation module to evaluate at least one data mining model in a distributed strategy. The present invention realizes various algorithm evaluation indexes based on the Spark distributed memory computing technology, forms an evaluation index library, and uses a process-based method to evaluate one or more algorithms. The model evaluation node, as a node of the process-based evaluation, can be reused multiple times in one evaluation, and thus can also realize repeated model evaluation of algorithms and synchronous analysis of multiple models. The evaluation results are displayed through an evaluation index table, a gain graph, and a lift graph to help select the best data mining model that meets various classification data mining conditions. The invention mainly evaluates data mining models and cannot evaluate workflow mining algorithms. However, the present invention evaluates the mining model obtained by workflow mining, so as to achieve the purpose of evaluating workflow mining algorithms. Summary of the Invention

[0007] The present invention aims to solve the above problems of the prior art. A method, medium, and system for checking the consistency of process mining based on task types are proposed. The technical solution of the present invention is as follows:

[0008] A method for checking the consistency of process mining based on task types includes the following steps: obtaining an original model and a mining model, where the original model is a process model determined during software design, and the mining model is a process model extracted from the information system of a business process; traversing the original model and the mining model to respectively obtain the total number of tasks, the number of visible tasks, the number of repeated tasks, and the number of invisible tasks in the original model and the mining model; then calculating the proportions of repeated tasks and invisible tasks in the original model and the mining model, and finally substituting them into a structural evaluation formula for calculation.

[0009] Further, the traversing the original model and the mining model to respectively obtain the total number of tasks, the number of visible tasks, the number of repeated tasks, and the number of invisible tasks in the original model and the mining model specifically includes:

[0010] Obtaining the original model PN and the mining model PN′ in the form of a Petri net;

[0011] Traversing each task t in the original model PN; obtaining all task sets T and visible task sets T in the original model V ; if the number of tasks t, λ(t)>1, then the task is a repeated task and is added to T D ; T I is the set of invisible tasks in the original model, and T is obtained by T-T V ; T I;

[0012] Traverse each task t in the mining model PN'; obtain all task sets T' and visible task sets T' in the original model V ; If the number of tasks λ(t) > 1, then this task is a duplicate task and add it to T' D ; T' I is the set of invisible tasks in the mining model, and obtain T' through T' - T' V '; |T| represents the total number of tasks in the original model; |T'| represents the total number of tasks in the mining model; |T I | represents the number of duplicate tasks in the original model; |T' D '| represents the number of duplicate tasks in the mining model; |T D | represents the number of invisible tasks in the original model; |T' I '| represents the number of invisible tasks in the mining model. I '| represents the number of invisible tasks in the mining model.

[0013] Furthermore, calculating the proportions of duplicate tasks and invisible tasks in the original model and the mining model specifically includes the following steps:

[0014] Calculate the proportion of duplicate tasks in the total tasks Calculate the proportion of invisible tasks in the total tasks Duplicate task D ∈ T D , T D There are k duplicate tasks in total. By comparing the original model and the mining model and combining the task types for model structural evaluation.

[0015] Furthermore, comparing the original model and the mining model and combining the task types for model structural evaluation specifically includes:

[0016] For each duplicate D in the duplicate task set T D , compare the number of times |D i | that appears in the original model and the number of times |D' i '| that appears in the mining model; 1 ≤ i ≤ k, and the calculation formula is as follows: i '| that appears in the mining model; 1 ≤ i ≤ k, and the calculation formula is as follows:

[0017]

[0018] The obtained S * is a value between 0 and 1. The closer S * is to 1, the more similar the mining model is to the original model in structure, and the better the structural performance of the mining model; the closer S * is to 0, the greater the difference between the mining model and the original model in structure, and the worse the structural performance of the mining model.

[0019] A computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, it implements the method for checking the consistency of process mining based on task type as described in any one of claims 1-4.

[0020] A consistency checking system based on the method described in any one of the above, comprising: an acquisition module for acquiring an original model and a mined model, where the original model is a process model determined during software design, and the mined model is a process model extracted from the information system of the business process;

[0021] A traversal module for traversing the original model and the mined model to respectively obtain the total number of tasks, the number of visible tasks, the number of repeated tasks, and the number of invisible tasks in the original model and the mined model;

[0022] A calculation module for calculating the proportions of repeated tasks and invisible tasks in the original model and the mined model, and finally substituting them into the structural evaluation formula for calculation.

[0023] The advantages and beneficial effects of the present invention are as follows:

[0024] The present invention proposes a method and system for checking the consistency by comparing the original model and the mined model. The original model is regarded as the best model, and factors such as different task types are considered. The mined model that is more consistent with the original model has a better structure. The formula in step 3 is proposed, which combines the size of the model and the proportion of the complex structure, and can more reasonably and accurately evaluate the structure of the mined model. Description of the Drawings

[0025] Figure 1 is a simplified flowchart of the usage scenario of the preferred embodiment provided by the present invention;

[0026] Figure 2 is the specific process of the structure evaluation algorithm of the present invention;

[0027] Figure 3 is the original model of the Fig6p31 dataset;

[0028] Figure 4 is the model mined from the Fig6p31 dataset using the non-repeated task mining algorithm (HM);

[0029] Figure 5 is the model mined from Fig6p31 using the repeated task mining algorithm (Fodina).

[0030] Figure 6 is the result of using the present invention to respectively Figure 4 for the model and Figure 5 for the model in terms of structural evaluation. DETAILED DESCRIPTION

[0031] The following will describe the technical solutions in the embodiments of the present invention in detail with reference to the accompanying drawings. The described embodiments are only a part of the embodiments of the present invention.

[0032] The technical solution of the present invention to solve the above technical problems is:

[0033] The present invention proposes a process mining consistency checking method that can evaluate repeated tasks, such as Figure 1 , specifically including the following methods:

[0034] First, traverse the original model and the two mining models to obtain the total number of tasks, visible tasks, repeated tasks, and invisible tasks in each model. Then calculate the proportion of repeated tasks and invisible tasks in the model, and finally bring them into the structural evaluation formula for calculation.

[0035] The specific steps include:

[0036] 1) Obtain the original model PN and the mining model PN′ in Petri grid format;

[0037] 2) Traverse each task t in the original model PN;

[0038] 3) Get all the task sets T and visible task sets T in the original model V ;

[0039] 4) If the number of tasks t λ(t)>1, then the task is a repeated task and it is added to T D middle;

[0040] 5)T I is the set of invisible tasks in the original model, through TT V Get T I ;

[0041] 6) Traverse each task t in the mining model PN′;

[0042] 7) Get all the task sets T′ and visible task sets T′ in the original model V ';

[0043] 8) If the number of tasks t λ(t)>1, then the task is a repeated task and it is added to T′ D 'middle;

[0044] 9)T′ I ′ is the set of unseen tasks in the mining model, through T′-T′ V 'Get T' I ';

[0045] 10) Let |T| denote the total number of tasks in the original model; |T′| denote the total number of tasks in the mined model; |T D | denote the number of duplicate tasks in the original model; |T′ D | denote the number of duplicate tasks in the mined model; |T I | denote the number of invisible tasks in the original model; |T′ I ′| denote the number of invisible tasks in the mined model

[0046] 11) Substitute the obtained |T|, |T D |, |T I | into the formula for calculation, where μ1 represents the proportion of duplicate tasks in the total tasks in the original model, and μ2 represents the proportion of invisible tasks in the total tasks in the original model.

[0047] 12) Compare the number of occurrences |D D | of each duplicate D D (a total of k duplicate tasks) in the duplicate task sets T i and T i ′ in the original model and the number of occurrences |D′ i | in the mined model; the main calculation formula is as follows:

[0048]

[0049] The obtained S * is a value between 0 and 1. The closer S * is to 1, the more similar the mined model is to the original model in structure, and the better the structure of the mined model; the closer S * is to 0, the greater the difference between the mined model and the original model in structure, and the worse the structure of the mined model.

[0050] Figure 1 As shown, an information system is constructed according to the original process model designed in the software design stage, and then a mined model is mined from the real event logs generated and recorded by the information system, and the mined model is checked for consistency with the original model.

[0051] The present invention combines different task types and compares the mined model with the original model for model structure evaluation. The specific algorithm process is as Figure 2 . First, traverse each task in the original model and the mined model, and respectively record the total number of tasks, the number of visible tasks, the number of duplicate tasks, and the number of invisible tasks in the original model and the mined model. Substitute into the formula for calculation.

[0052] Figure 3 is the original model of the Fig6p31 dataset, Figure 4The model mined by the non - repetitive task mining algorithm (HM) for the Fig6p31 dataset. Figure 5 The model mined by the repetitive task mining algorithm (Fodina) for Fig6p31. The Fodina algorithm is better than the HM algorithm in mining models containing repetitive tasks, so these two algorithms are used for comparison. There are 3 repetitive tasks in the original model. Figure 4 The model did not mine any repetitive tasks. Figure 5 The model in [[ ]] mined 2 repetitive tasks. It can be found that Figure 5 The model in [[ ]] is more in line with the original model. Figure 5 The model in [[ ]] has better structure.

[0053] Figure 6 For the structural evaluation results of the models of [[ ]] and [[ ]] respectively using the present invention. First, traverse the original model and the two mined models to obtain the total number of tasks, visible tasks, repetitive tasks, and invisible tasks in the models as shown in [[ ]]. Figure 4 and Figure 5 ; Then calculate the proportions of repetitive tasks and invisible tasks in the model through [[ ]], which are Figure 6 ; Substitute all data into [[ ]] for calculation. The structurality of the [[ ]] model is approximately 0.65, 0;

[0054]

[0055] for calculation. Figure 4 The structurality of the [[ ]] model is approximately 0.79. The results show that Figure 5 the structure of the model in [[ ]] is better than that of the model in [[ ]], which is in line with the actual situation. Figure 5 in [[ ]] Figure 4 in [[ ]].

[0056] The systems, devices, modules or units illustrated in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0057] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0058] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0059] The above embodiments should be understood as merely illustrating the present invention and not as limiting the scope of protection of the present invention. After reading the contents of the present invention, technicians may make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A process mining consistency checking method based on task types, characterized in that, Including the following steps: Obtain the original model and the mining model. The original model is the process model determined during software design, and the mining model is the process model extracted from the information system of the business process. Traverse the original model and the mining model to obtain the total number of tasks, visible tasks, repeated tasks, and invisible tasks in the original model and the mining model respectively. Then calculate the proportions of repeated tasks and invisible tasks in the original model and the mining model, and finally substitute them into the structural evaluation formula for calculation; The calculation of the proportions of repeated tasks and invisible tasks in the original model and the mining model specifically includes the following steps: Calculate the proportion of repetitive tasks in the total tasks Calculate the proportion of invisible tasks in the total tasks Repetitive task D ∈ T D , T D There are k repetitive tasks in total in T. By comparing the original model and the mining model and combining the task types for model structural evaluation; Compare the original model and the mining model, and conduct model structural evaluation in combination with task types, specifically including: For each duplicate D D in the set T of duplicate tasks i , compare the number of occurrences |D i | in the original model with the number of occurrences |D′ i | in the mined model; 1 ≤ i ≤ k, and the calculation formula is as follows: The obtained S * is a value between 0 and 1. The closer S * is to 1, the more similar the mining model is to the original model in terms of structure, and the better the structure of the mining model; S * The closer it is to 0, the greater the difference between the mining model and the original model in terms of structure, and the worse the structure of the mining model.

2. The method for process mining consistency checking based on task type according to claim 1, wherein The traversal of the original model and the mining model to obtain the total number of tasks, visible tasks, repeated tasks, and invisible tasks in the original model and the mining model respectively specifically includes: Obtain the original model PN in Petri net format and the mining model PN'; Traverse each task t in the original model PN; obtain all task sets T and visible task sets T in the original model V ; if the number of tasks λ(t) > 1, then the task is a duplicate task and add it to T D ; T I is the set of invisible tasks in the original model, through T - T V obtain T I ; Traverse each task t in the mining model PN′; obtain all task sets T′ and visible task sets T′ in the original model V ; If the number of tasks t λ(t)>1, then the task is a repeated task and it is added to T′ D T′ I To mine the unseen task set in the model, we use T′-T′ V Get T′ I ; |T| represents the number of all tasks in the original model; |T′| represents the number of all tasks in the mining model; |T D | represents the number of repeated tasks in the original model; |T′ D | represents the number of repeated tasks in the mining model; |T I | represents the number of unseen tasks in the original model; |T′ I |Indicates the number of invisible tasks in the mining model.

3. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, it implements the method for checking the consistency of process mining based on task types as described in any one of claims 1-2.

4. A consistency checking system based on the method according to any one of claims 1-2, characterized in that, Including: An acquisition module for obtaining the original model and the mining model. The original model is the process model determined during software design, and the mining model is the process model extracted from the information system of the business process; A traversal module for traversing the original model and the mining model to obtain the total number of tasks, visible tasks, repeated tasks, and invisible tasks in the original model and the mining model respectively; A calculation module for calculating the proportions of repeated tasks and invisible tasks in the original model and the mining model, and finally substituting them into the structural evaluation formula for calculation.

Images