Step management and application method based on process mining

Abstract

The invention discloses a step management and application method based on process mining. Comprising the following steps: S1, collecting data; s2, establishing an index; s3, distributing a filter; s4, selecting a filter; s5, configuring screening conditions; s6, storage of the screening stack; s7, creating a data view; s8, adding a filter; s9, obtaining a data view; s10, preserving the formula; s11, exiting is carried out; s12, calling the object; s13, reading all configuration information; compared with traditional data screening processing, the business log type sequence data is subjected to multi-dimensional and multi-step screening processing, the target requirement is met, the screening processing steps are managed, each newly-created processing step can be flexibly sorted, started or closed, and the purpose of more flexibly facing flow mining and exploration is achieved; and meanwhile, storage and calling of a processing step sequence are supported, continuous and optimal practice is provided for processing work, and time and reuse value experience are greatly saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of process mining, and more specifically, particularly relates to a method for managing and applying process mining steps. Background Art

[0002] Process mining is a data analysis technology that aims to extract process models from enterprise systems (such as enterprise resource planning systems ERP, customer relationship management systems CRM, workflow management systems, etc.) in order to better understand, monitor and optimize business processes. It combines data mining and process modeling techniques, using event logs recorded in enterprise systems to automatically discover, monitor and improve actual business processes.

[0003] As digital transformation of enterprises advances, more and more enterprises rely on automated systems in business management, generating a large amount of data. This data usually contains the execution records of business processes, and can therefore provide valuable clues for process optimization.

[0004] However, there are some problems with the existing technology: traditionally, Excel is often used to filter and penetrate data. After the user imports the original data, the filtering function is enabled in Excel. Then click on the fields to be filtered in the header area and perform the necessary filtering operations. However, there are some problems with the existing technology: the more fields need to be filtered at the same time, the more likely the user will forget the historical filtering process; when there are many filtering fields and many steps, if a certain filtering condition or status in the previous step needs to be adjusted, it will consume the user's constant recall and possible trial operations; it cannot support the flexible adjustment of the order of filtering steps; the filtering step process is a one-time activity, and the next time the same type of data is opened and the same processing is required, each step must be manually executed from the beginning; and for the same project data, after exiting the task halfway and re-entering after a while, the previous operation steps may have been forgotten, and it takes time and operation to recall before proceeding; the filtering step process cannot be replicated and shared as experience and best practices in human activity groups. Therefore, we propose a method based on process mining step management and application. Summary of the invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a method based on process mining step management and application. By managing the steps of data screening operations, the data processing business process is made flexible, controllable and reusable, thereby improving overall business efficiency.

[0006] To achieve the above object, the present invention provides the following technical solution: a method based on process mining step management and application, comprising the following steps:

[0007] S1. Collect data: Create a project and import the business log data that needs to be filtered. The data is stored in the database;

[0008] S2. Create index: Create index for the data field through the database according to its characteristics;

[0009] S3. Assign filters: classify the data into different filters according to its business dimension. The time dimension corresponds to the time interval filter, the business step corresponds to the process filter, the time dimension and the sequence number dimension correspond to the performance filter, the business step and the sequence number correspond to the business tracking filter, and the resource dimension and the business dimension correspond to the object attribute filter.

[0010] S4. Select filter: select a filter through the user interface;

[0011] S5. Configure filtering conditions: The corresponding dimension data is fed back in the filter for the user to observe. The user configures the filtering conditions by clicking on the interface. After the user confirms and submits, the corresponding filtering parameters are obtained;

[0012] S6, storage of the filter stack: temporarily storing the corresponding filter category and filter condition in the filter stack object;

[0013] S7, create data view: generate corresponding SQL according to the filter and filter conditions in the filter stack, filter the database at the same time, and create a data view;

[0014] S8, add filter: if the user needs to add a filter, repeat to S4. If the user needs to adjust the previously configured filtering conditions or steps, adjust the filter step sequence and filtering conditions in the filtering stack through the user interface. After the adjustment is completed, repeat to S6.

[0015] S9, obtaining data view: the user obtains the required data view result;

[0016] S10, recipe saving: the user saves the screening steps and corresponding filters and screening conditions in the screening stack as a recipe, saves it as a screening stack object in the corresponding project, and saves it in the data table corresponding to the project in JSON data format;

[0017] S11, exit: the user exits the project;

[0018] S12, calling the object: the user enters the project again and selects the above-mentioned saved screening stack recipe object through the load button;

[0019] S13, read all configuration information: successfully read all configuration information of the filter stack from the filter stack data table, and then the system executes S7.

[0020] Specifically, the business log data in S1 is preprocessed before being stored in the database; the preprocessing is used to clean and convert the data;

[0021] Cleaning is used to remove noise, detect anomalies, detect missing values, and fill missing values; conversion is used to convert the format of data.

[0022] Specifically, the noise removal is performed for each data x i , use windows of different scales to perform median filtering, and then perform weighted fusion on the filtering results of different scales;

[0023]

[0024] Among them, y i represents the output data after filtering; S represents the number of scales, that is, the number of different scales used when performing the median filtering operation, and each scale corresponds to a filtering window of different sizes; w s Represents the weight of scale S, which is used to perform weighted averaging of filtering results of different scales; represents the result of median filtering the data at scale S. Specifically, It is all the data within the window range at scale S, that is, the neighborhood data at this scale; window s Represents the neighborhood window corresponding to the scale S. At different scales, the size of the neighborhood window is different.

[0025] Specifically, the missing value detection is performed by isna() and isnull(); isna() and isnull() are used to return a Boolean DataFrame or Series with the same structure as the original data, where True represents missing values ​​and False represents non-missing values;

[0026] And notna() and notnull() are the return methods of isna() and isnull() respectively, returning True for non-null values ​​and False for null values;

[0027] And notna() and notnull() are combined with sum() method to count the missing values ​​of each column or row and count the number of missing values;

[0028] And combined with Any() and all() methods to help detect whether a column contains any missing values ​​or all are missing values.

[0029] Specifically, the calculation of the anomaly detection is as follows:

[0030] The Z-score of the data is obtained by calculating the difference between the data and the mean and dividing it by the standard deviation;

[0031]

[0032] Among them, x is the value of the data, μ is the mean of the data, σ is the standard deviation of the data, and points with a value of Z greater than 3 or less than -3 are considered outliers and are removed;

[0033] The missing value filling is used to fill the vacancies of outlier removal and the vacant positions of missing value detection by the mean, and the mean is calculated as follows:

[0034]

[0035] Among them, x i ' is the value of non-missing data, and n is the number of non-missing data.

[0036] Specifically, the format conversion scales the data to the range of [0, 1] by normalization processing, and the calculation of the normalization processing is as follows:

[0037]

[0038] Among them, x i is the input data, x min is the minimum value in the data, x max is the maximum value in the data, scaling the data to the range of [0, 1].

[0039] Specifically, the characteristics of the data field include event ID, resource, timestamp, activity name, and associated case ID;

[0040] The extraction calculation of the features of the data field is as follows:

[0041]

[0042] Among them, TFIDF(x i ,d) reflects the importance of a feature in the data field. The more frequently it appears, the higher the TF value, x i represents the calculated feature, and d represents the feature x i The number of times it occurs in the data field, It is used to measure the prevalence of a feature in the entire corpus. The higher the value, the less common the feature is in the corpus. N is the total number of texts in the corpus. DF(x i ) is represented as containing feature x i The number of texts;

[0043] If a feature appears frequently in the data field to be detected and calculated, but appears less frequently in other texts, TF(x i ,The value of d) will be very high, indicating the importance of the feature to the data field that needs to be detected and calculated;

[0044] On the contrary, if the feature frequency is low or the feature appears in many texts, then TF(x i The value of ,d) will be low, indicating that the feature is not important to the data field that needs to be detected and calculated.

[0045] Specifically, the SQL in S7 queries data through filter categories and screening conditions, and generates a data view based on the query results;

[0046] The steps for generating the data view are as follows:

[0047] Select the view type: Use Tableau to select the corresponding view and generate the view;

[0048] Processing data: SQL queries data through filter categories and screening conditions, and stores the queried data in the view;

[0049] Publish and Share: Publish the generated view in the user interface.

[0050] Specifically, the recipe in S10 includes a screening step, a corresponding filter and a screening condition;

[0051] Filter step: Filter each filter operation in the stack;

[0052] Filters: The filters used in each screening step;

[0053] Filter conditions: specific conditions under the filter;

[0054] The recipe saves the screening steps, corresponding filters and screening conditions in JSON data format, and stores the data in the recipe through the data view.

[0055] Specifically, the screening steps, corresponding filters and screening conditions in the JSON data format are associated by multiple keys;

[0056] The key is generated by a hash function, which is calculated as follows:

[0057] The formula for outputting the hash value is expressed as:

[0058] H=f(H prev ,M),

[0059] Where: H is the current hash value, H previs the initial hash value, M is the currently processed screening step, the corresponding filter and screening condition, and f is the hash function.

[0060] Technical effects and advantages of the present invention:

[0061] Compared with traditional data screening and processing, the present invention performs multi-dimensional and multi-step screening and processing on business log-type sequence data to meet target requirements, manage the steps of screening and processing, and can flexibly sort, enable or disable each newly created processing step, which is more flexible for the purpose of process mining and exploration. At the same time, it also supports the saving and retrieval of processing step sequences, providing continuity and best practices for processing work, greatly saving time and reusing valuable experience;

[0062] When performing process mining, data is queried by assigning filters, and the filter conditions are set to obtain the corresponding filter parameters, and the filter category and filter conditions are temporarily stored in the filter stack object. The filter stack is saved as a recipe and stored in JSON format, which is convenient for loading and applying these filter recipes according to the project ID in the future. It is flexible and can easily expand different filter conditions and step types, and display the query data results in a data view for easy visualization. The filter stack data table is easy to read later, so that all the configuration information of the filter stack can be used for query next time, improving the efficiency of the query.

[0063] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the step flow provided by the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0066] like Figure 1 As shown, a method for process mining step management and application provided by an embodiment of the present invention includes the following steps:

[0067] S1. Collect data: Create a project and import the business log data that needs to be filtered. The data is stored in the database;

[0068] S2. Create index: Create index for the data field through the database according to its characteristics;

[0069] S3. Assign filters: classify the data into different filters according to its business dimension. The time dimension corresponds to the time interval filter, the business step corresponds to the process filter, the time dimension and the sequence number dimension correspond to the performance filter, the business step and the sequence number correspond to the business tracking filter, and the resource dimension and the business dimension correspond to the object attribute filter.

[0070] S4. Select filter: select a filter through the user interface;

[0071] S5. Configure filtering conditions: The corresponding dimension data is fed back in the filter for the user to observe. The user configures the filtering conditions by clicking on the interface. After the user confirms and submits, the corresponding filtering parameters are obtained;

[0072] S6, storage of the filter stack: temporarily storing the corresponding filter category and filter condition in the filter stack object;

[0073] S7, create data view: generate corresponding SQL according to the filter and filter conditions in the filter stack, filter the database at the same time, and create a data view;

[0074] S8, add filter: if the user needs to add a filter, repeat to S4. If the user needs to adjust the previously configured filtering conditions or steps, adjust the filter step sequence and filtering conditions in the filtering stack through the user interface. After the adjustment is completed, repeat to S6.

[0075] S9, obtaining data view: the user obtains the required data view result;

[0076] S10, recipe saving: the user saves the screening steps and corresponding filters and screening conditions in the screening stack as a recipe, saves it as a screening stack object in the corresponding project, and saves it in the data table corresponding to the project in JSON data format;

[0077] S11, exit: the user exits the project;

[0078] S12, calling the object: the user enters the project again and selects the above-mentioned saved screening stack recipe object through the load button;

[0079] S13, read all configuration information: successfully read all configuration information of the filter stack from the filter stack data table, and then the system executes S7.

[0080] In this embodiment, preferably, the business log data in S1 is preprocessed before being stored in the database; the preprocessing is used to clean and convert the data;

[0081] Cleaning is used to remove noise, detect anomalies, detect missing values, and fill missing values; conversion is used to convert the format of data;

[0082] It should be noted that preprocessing is used to clean and convert data, which facilitates the removal of noise in the data, detects outliers and missing values ​​in the data, and fills in missing values ​​to ensure data integrity. The format is converted through conversion to maintain the uniformity of the data format.

[0083] In this embodiment, preferably, the noise removal is performed for each data x i , use windows of different scales to perform median filtering, and then perform weighted fusion on the filtering results of different scales;

[0084]

[0085] Among them, y i represents the output data after filtering; S represents the number of scales, that is, the number of different scales used when performing the median filtering operation, and each scale corresponds to a filtering window of different sizes; w s Represents the weight of scale S, which is used to perform weighted averaging of filtering results of different scales; represents the result of median filtering the data at scale S. Specifically, It is all the data within the window range at scale S, that is, the neighborhood data at this scale; window s Represents the neighborhood window corresponding to the scale S. At different scales, the size of the neighborhood window is different;

[0086] It should be noted that the data is filtered through median filtering, and the filtering results are weighted fused to filter out the noise in the data, prevent the noise in the data from causing errors in subsequent data processing, and improve the accuracy of the data.

[0087] In this embodiment, preferably, the missing value detection is performed by isna() and isnull(); isna() and isnull() are used to return a Boolean DataFrame or Series with the same structure as the original data, where True represents a missing value and False represents a non-missing value;

[0088] And notna() and notnull() are the return methods of isna() and isnull() respectively, returning True for non-null values ​​and False for null values;

[0089] And notna() and notnull() are combined with sum() method to count the missing values ​​of each column or row and count the number of missing values;

[0090] And combined with Any() and all() methods to help detect whether a column contains any missing values ​​or all are missing values;

[0091] It should be noted that the missing values ​​in the data are detected by means of missing value detection, and all detected missing values ​​are counted to facilitate subsequent data filling.

[0092] In this embodiment, preferably, the calculation of the anomaly detection is as follows:

[0093] The Z-score of the data is obtained by calculating the difference between the data and the mean and dividing it by the standard deviation;

[0094]

[0095] Among them, x is the value of the data, μ is the mean of the data, σ is the standard deviation of the data, and points with a value of Z greater than 3 or less than -3 are considered outliers and are removed;

[0096] The missing value filling is used to fill the vacancies of outlier removal and the vacant positions of missing value detection by the mean, and the mean is calculated as follows:

[0097]

[0098] Among them, x i ' is the value of non-missing data, n is the number of non-missing data;

[0099] It should be noted that the Z-score is used to detect outliers in the data, and the outliers are eliminated. The eliminated outliers and missing values ​​are filled with the mean to improve the integrity of the data and maintain the accuracy of the data.

[0100] In this embodiment, preferably, the format conversion scales the data to a range of [0, 1] by normalization processing, and the calculation of the normalization processing is as follows:

[0101]

[0102] Among them, x i is the input data, x min is the minimum value in the data, x max is the maximum value in the data, scaling the data to the range of [0, 1];

[0103] It should be noted that the data is scaled to the range of [0, 1] through normalization processing, which facilitates subsequent calculation processing of the data and reduces the calculation pressure and complexity of the data.

[0104] In this embodiment, preferably, the characteristics of the data field include event ID, resource, timestamp, activity name and associated case ID;

[0105] The extraction calculation of the features of the data field is as follows:

[0106]

[0107] Among them, TFIDF(x i ,d) reflects the importance of a feature in the data field. The more frequently it appears, the higher the TF value. ,xi represents the calculated feature, and d represents the number of times feature xi appears in the data field. It is used to measure the prevalence of a feature in the entire corpus. The higher the value, the less common the feature is in the corpus. N is the total number of texts in the corpus. DF(x i ) is represented as containing feature x i The number of texts;

[0108] If a feature appears frequently in the data field to be detected and calculated, but appears less frequently in other texts, TF(x i ,The value of d) will be very high, indicating the importance of the feature to the data field that needs to be detected and calculated;

[0109] On the contrary, if the feature frequency is low or the feature appears in many texts, then TF(x i ,The value of ,d) will be low, indicating that the feature is not important to the data field that needs to be detected and calculated;

[0110] It should be noted that, through feature extraction of data fields, feature extraction of event ID, resource, timestamp, activity name and associated case ID is achieved, so as to effectively obtain feature information in the data.

[0111] In this embodiment, preferably, the SQL in S7 queries data through filter categories and screening conditions, and generates a data view based on the query results;

[0112] The steps for generating the data view are as follows:

[0113] Select the view type: Use Tableau to select the corresponding view and generate the view;

[0114] Processing data: SQL queries data through filter categories and screening conditions, and stores the queried data in the view;

[0115] Publishing and sharing: Displaying the generated view in the user interface;

[0116] It should be noted that by establishing a view to fill in the queried data and displaying the queried data through the view, it is easy to display the queried data and display the screening steps, corresponding filters and screening conditions.

[0117] In this embodiment, preferably, the recipe in S10 includes a screening step, a corresponding filter and a screening condition;

[0118] Filter step: Filter each filter operation in the stack;

[0119] Filters: The filters used in each screening step;

[0120] Filter conditions: specific conditions under the filter;

[0121] The recipe saves the screening steps, corresponding filters and screening conditions in JSON data format, and stores the data in the recipe through the data view;

[0122] It should be noted that the screening steps, corresponding filters and screening conditions are associated through the JSON data format, so that when performing a query, the entire screening steps, corresponding filters and screening conditions can be called out.

[0123] In this embodiment, preferably, the screening steps, corresponding filters and screening conditions in the JSON data format are associated by multiple keys;

[0124] The key is generated by a hash function, which is calculated as follows:

[0125] The formula for outputting the hash value is expressed as:

[0126] H=f(H prev ,M),

[0127] Where: H is the current hash value, H prev is the initial hash value, M is the currently processed screening step, the corresponding filter and screening condition, and f is the hash function;

[0128] It should be noted that the association value calculation of the screening step, the corresponding filter and the screening condition is implemented by the hash function, so that the screening step, the corresponding filter and the screening condition can be associated with each other.

[0129] The specific operation steps of the present invention are:

[0130] Step 1: Collect data: Create a project and import the business log data that needs to be filtered. The data is stored in the database;

[0131] Step 2: Create an index: Create an index for the data field through the database based on its characteristics;

[0132] Step 3: Assign filters: Classify the data into different filters according to its business dimension. The time dimension corresponds to the time interval filter, the business step corresponds to the process filter, the time dimension and the sequence number dimension correspond to the performance filter, the business step and the sequence number correspond to the business tracking filter, and the resource dimension and the business dimension correspond to the object attribute filter.

[0133] Step 4. Select filter: Select filter through the user interface;

[0134] Step 5: Configure filtering conditions: The corresponding dimension data is fed back in the filter for users to observe. Users configure filtering conditions by clicking on the interface. After users confirm and submit, they obtain the corresponding filtering parameters.

[0135] Step 6: Storage of filter stack: temporarily storing the corresponding filter category and filter condition in the filter stack object;

[0136] Step 7: Create data view: Generate corresponding SQL according to the filters and filter conditions in the filter stack, filter the database at the same time, and create data view;

[0137] Step 8. Add filters: If the user needs to add filters, repeat step 4. If the user needs to adjust the previously configured filtering conditions or steps, adjust the filter step order and filtering conditions in the filtering stack through the user interface. After the adjustment is completed, repeat step 6.

[0138] Step 9: Obtaining data view: The user obtains the required data view result;

[0139] Step 10: Recipe saving: The user saves the filtering steps and corresponding filters and filtering conditions in the filtering stack as a recipe, saves it to the corresponding project as a filtering stack object, and saves it in the data table corresponding to the project in JSON data format;

[0140] Step 11. Exit: The user exits the project;

[0141] Step 12: Calling the object: The user enters the project again and selects the above-mentioned saved filter stack recipe object through the load button;

[0142] Step 13, read all configuration information: successfully read all configuration information of the filter stack from the filter stack data table, and then the system executes step 7.

[0143] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method based on process mining step management and application, characterized in that: The steps include: S1. Collect data: Create a project and import the business log data that needs to be filtered. The data is stored in the database; S2. Create index: Create index for the data field through the database according to its characteristics; S3. Assign filters: classify the data into different filters according to its business dimension. The time dimension corresponds to the time interval filter, the business step corresponds to the process filter, the time dimension and the sequence number dimension correspond to the performance filter, the business step and the sequence number correspond to the business tracking filter, and the resource dimension and the business dimension correspond to the object attribute filter. S4. Select filter: select a filter through the user interface; S5. Configure filtering conditions: The corresponding dimension data is fed back in the filter for the user to observe. The user configures the filtering conditions by clicking on the interface. After the user confirms and submits, the corresponding filtering parameters are obtained; S6, storage of the filter stack: temporarily storing the corresponding filter category and filter condition in the filter stack object; S7, create data view: generate corresponding SQL according to the filter and filter conditions in the filter stack, filter the database at the same time, and create a data view; S8, add filter: if the user needs to add a filter, repeat to S4. If the user needs to adjust the previously configured filtering conditions or steps, adjust the filter step sequence and filtering conditions in the filtering stack through the user interface. After the adjustment is completed, repeat to S6. S9, obtaining data view: the user obtains the required data view result; S10, recipe saving: the user saves the screening steps and corresponding filters and screening conditions in the screening stack as a recipe, saves it as a screening stack object in the corresponding project, and saves it in the data table corresponding to the project in JSON data format; S11, exit: the user exits the project; S12, calling the object: the user enters the project again and selects the above-mentioned saved screening stack recipe object through the load button; S13, read all configuration information: successfully read all configuration information of the filter stack from the filter stack data table, and then the system executes S7.

2. The method for process mining step management and application according to claim 1, characterized in that: The business log data in S1 is preprocessed before being stored in the database; the preprocessing is used to clean and convert the data; Cleaning is used to remove noise, detect anomalies, detect missing values, and fill missing values; conversion is used to convert the format of data.

3. The method for process mining step management and application according to claim 2, characterized in that: The noise removal is for each data x i , use windows of different scales to perform median filtering, and then perform weighted fusion on the filtering results of different scales; Among them, y i represents the output data after filtering; S represents the number of scales, that is, the number of different scales used when performing the median filtering operation, and each scale corresponds to a filtering window of different sizes; w s Represents the weight of the scale S, which is used to perform weighted averaging of filtering results of different scales; represents the result of median filtering the data at scale S. Specifically, It is all the data within the window range at scale S, that is, the neighborhood data at this scale; window s Represents the neighborhood window corresponding to the scale S. At different scales, the size of the neighborhood window is different.

4. The method for process mining step management and application according to claim 2, characterized in that: The missing value detection is performed by isna() and isnull(); isna() and isnull() are used to return a Boolean DataFrame or Series with the same structure as the original data, where True represents missing values ​​and False represents non-missing values; And notna() and notnull() are the return methods of isna() and isnull() respectively, returning True for non-null values ​​and False for null values; And notna() and notnull() are combined with sum() method to count the missing values ​​of each column or row and count the number of missing values; And combined with Any() and all() methods to help detect whether a column contains any missing values ​​or all are missing values.

5. The method for process mining step management and application according to claim 2, characterized in that: The calculation of the anomaly detection is as follows: The Z-score of the data is obtained by calculating the difference between the data and the mean and dividing it by the standard deviation; Among them, x is the value of the data, μ is the mean of the data, σ is the standard deviation of the data, and points with a Z value greater than 3 or less than -3 are considered outliers and are removed; The missing value filling is used to fill the vacancies of outlier removal and the vacant positions of missing value detection by the mean, and the mean is calculated as follows: Among them, x i ' is the value of non-missing data, and n is the number of non-missing data.

6. The method for process mining step management and application according to claim 1, characterized in that: The format conversion scales the data to the range of [0, 1] by normalization, and the calculation of the normalization is as follows: Among them, x i is the input data, x min is the minimum value in the data, x max is the maximum value in the data, scaling the data to the range of [0, 1].

7. The method for process mining step management and application according to claim 1, characterized in that: Characteristics of the data fields include event ID, resource, timestamp, activity name, and associated case ID; The extraction calculation of the features of the data field is as follows: Among them, TFIDF(x i ,d) reflects the importance of a feature in the data field. The more frequently it appears, the higher the TF value. ,xi represents the calculated feature, and d represents the number of times feature xi appears in the data field. It is used to measure the prevalence of a feature in the entire corpus. The higher the value, the less common the feature is in the corpus. N is the total number of texts in the corpus. DF(x i ) is represented as containing feature x i The number of texts; If a feature appears frequently in the data field to be detected and calculated, but appears less frequently in other texts, TF(x i ,The value of d) will be very high, indicating the importance of the feature to the data field that needs to be detected and calculated; On the contrary, if the feature frequency is low or the feature appears in many texts, then TF(x i The value of ,d) will be low, indicating that the feature is not important to the data field that needs to be detected and calculated.

8. The method for process mining step management and application according to claim 1, characterized in that: The SQL in S7 queries data through filter categories and screening conditions, and generates a data view based on the query results; The steps for generating the data view are as follows: Select the view type: Use Tableau to select the corresponding view and generate the view; Processing data: SQL queries data through filter categories and screening conditions, and stores the queried data in the view; Publish and Share: Publish the generated view in the user interface.

9. The method for process mining step management and application according to claim 8, characterized in that: The recipe in S10 includes a screening step, a corresponding filter and a screening condition; Filter step: Filter each filter operation in the stack; Filters: The filters used in each screening step; Filter conditions: specific conditions under the filter; The recipe saves the screening steps, corresponding filters and screening conditions in JSON data format, and stores the data in the recipe through the data view.

10. The method for process mining step management and application according to claim 9, characterized in that: The screening steps, corresponding filters and screening conditions in the JSON data format are associated by multiple keys; The key is generated by a hash function, which is calculated as follows: The formula for outputting the hash value is expressed as: H=f(H prev ,M), Where: H is the current hash value, H prev is the initial hash value, M is the currently processed screening step, the corresponding filter and screening condition, and f is the hash function.

Images