An RPA process discovery method and device based on multi-view clustering

Abstract

The present invention provides an RPA process discovery method and device based on multi-view clustering, comprising: a feature calculation module, a trajectory clustering module, and a process discovery module; the feature calculation module is used to receive event logs and the number of feature clusters as input, construct a configuration file containing feature information, and cluster process trajectories with similar feature values ​​into clusters based on the similarity matrix between the feature information; the trajectory clustering module is used to collect and adjust the clusters of process clustering under multiple views, fuse the multi-view clustering results into a complete process trajectory similarity, and obtain multiple groups containing similar process trajectories based on threshold clustering; the process discovery module is used to receive, mine, and transform process information, and transform process nodes and trajectories that can be automatically executed by RPA into scheduling instructions and programs that can be recognized by the corresponding RPA system, thereby completing the required flow chart. The beneficial effects of the present invention are: making the process model simpler, reducing complexity and difficulty, and saving a lot of resources.

Description

Technical Field

[0001] The present invention relates to the fields of process discovery and process automation, and in particular to an RPA process discovery method and device based on multi-view clustering. Background Art

[0002] RPA, or Robotic Process Automation, is primarily designed to reduce the time and financial resources consumed by manual operations. It uses software programs to automatically execute rule-based business processes, reduce human errors and repetitive work, and improve work efficiency.

[0003] Process discovery is one of the important technologies in the field of process mining. It mainly uses serialized event logs as input, discovers structured process trajectories, and outputs business process models, providing more valuable process information for RPA process automation.

[0004] Current systems that identify automatable processes through process discovery primarily rely on fully feeding event logs into the process discovery algorithm. This easily results in highly accurate but complex process models. This is primarily because infrequent process nodes cannot be identified as executable instructions, which increases the time and complexity of model discovery and wastes significant system resources. Scientifically processing the input to process algorithms to efficiently meet the needs of RPA process automation is a challenge worth considering. Summary of the Invention

[0005] To solve the above problems, the present invention provides an RPA process discovery method and device based on multi-view clustering. The RPA process discovery method based on multi-view clustering mainly includes: a feature calculation module, a trajectory clustering module and a process discovery module;

[0006] The feature calculation module is used to receive event logs and the number of feature clusters As input, in a process view Build a configuration file containing feature information , based on the similarity matrix between feature information , process trajectories with similar eigenvalues Clustering , ;

[0007] The trajectory clustering module is used to collect and adjust the clusters of process clustering under multiple views, and analyze different process clusters. The classification gap , get the clustering similarity between process trajectories , based on the adaptive view weight value , fuse the multi-view clustering results into a complete process trajectory similarity, based on the threshold Clustering to obtain multiple groups containing similar process trajectories ,in, , , represents the process trajectory of group k, k =1,2,…, n ;

[0008] The process discovery module is used to receive, mine and transform process information. The process groups in the input process discovery algorithm in turn, output the corresponding process model , M k Represent the process model corresponding to the kth group of process trajectories. Input the process model into the instruction converter, convert the process nodes and trajectories that can be automatically executed by RPA into scheduling instructions and programs that can be recognized by the corresponding RPA system, and then complete the required flowchart.

[0009] Furthermore, the feature calculation module includes a feature extraction module and a feature clustering module;

[0010] For the event log L, the feature extraction module is used to generate different process view sets P = {p act ,p res ,p trans ,p perform}, and encode the feature information in the profile into the corresponding feature vector, calculate the spatial distance between different feature vectors, and output the similarity matrix between the corresponding process features under the process view. ,in p * ∈P, represents any process view; p act The process view representing the event type, p res The process view that represents the process consuming resources, p trans A process view that represents event transitions, p perform A process view that represents the process run;

[0011] In multiple different process views, the feature clustering module is used to aggregate the process trajectories with high similarity in the matrix into the same cluster. In, get The number of clusters.

[0012] Furthermore, in the feature extraction module, for event logs Each process track in , ,in, Represents the process trajectory t k The i-th event node in each event node records the attribute content of the event, which includes the event type activity, event consumption resource resource, event start and end time timestamp, in different process views P={p act ,p res ,p trans ,p perform}, respectively record the event type frequency, activity resource utilization frequency, the frequency of close occurrence of different event types, and the time information statistics when the process trajectory is running. The time information statistics include the trajectory running time, the longest event running time, the shortest event running time, the average running time of the event node and the number of event nodes, and output the configuration file corresponding to the process trajectory , among which 1 k n, n represents the number of complete process traces in the log, Respectively represent the corresponding Configuration files;

[0013] In any process view p * Next, based on different process trajectories Corresponding configuration file , Indicates the first The configuration file corresponding to the process trajectory, k=1,2,…,n, converts the quantified feature information into the feature vector corresponding to the process trajectory , calculate the similarity between any two feature vectors , output the similarity matrix between process trajectories ,in, In the process view p * Downstream process trajectory The degree of similarity between Represents the distance between two different feature vectors, the distance between vectors The calculation formula is as follows:

[0014]

[0015] in, Represents different eigenvectors.

[0016] Furthermore, in the feature clustering module, based on the number of clusters initialized by the user , randomly selected in a process view process trajectories as initial cluster centers , extract the corresponding Indexed row data ,for Find the number of rows with the maximum similarity value for each column , r k Indicates the kth row, merging the process traces of the corresponding row and column position indexes to form a clustering result ;

[0017] Then the iterative process begins, randomly selecting The new process trajectory is used as the new cluster center of this iteration, the clustering results are calculated, and the function Evaluate the average dissimilarity between the clustered process trajectories and the center points. The formula is defined as follows:

[0018]

[0019] Among them, locI() means finding according to the process track id The row and column indices corresponding to the matrix, E j Represents the metric value calculated at the jth iteration, if but , when the algorithm converges or exceeds the maximum number of iterations, take E best The corresponding cluster assignment is the result of feature clustering in this view; based on the clustering results, the corresponding trajectory cluster matrix is ​​calculated and output. ,in, , which means in the process view p * Next process trajectory t i and cluster c j The similarity of the corresponding cluster centers.

[0020] Furthermore, the trajectory clustering module includes a process trajectory clustering module and a multi-view aggregation module;

[0021] Use the process trajectory clustering module to collect multi-view process clustering results , in the process view Calculate the cluster similarity between any two process trajectories , output cluster similarity matrix ,in, Representative process trajectory In the process view The degree of clustering under In the process view Downstream process trajectory The value of the cluster similarity between ;

[0022] Using the multi-view aggregation module, the cluster similarity matrix is ​​aggregated The elements in the Fusion into a process trajectory similarity matrix similar T ;

[0023] Using the trajectory clustering module, the complete event log L is divided into multiple groups containing similar process trajectories. , and ensure that the similarity between trajectories within the group is not lower than the threshold .

[0024] Furthermore, in the multi-view aggregation module, for the process view P={p act ,p res ,p trans ,p perform}, set the weight corresponding to the impact process classification:

[0025] ,

[0026] Multiply the weight value by the cluster similarity matrix of the corresponding view to obtain the complete process trajectory similarity matrix ,in, Represents the process trajectory t i ,t j The similarity of multi-view aggregation between , is calculated as follows:

[0027]

[0028] in, represents the normalization function, and Represent the maximum and minimum values ​​of the i-th row of the matrix, Represents the clustering similarity under different process perspectives, They represent the weight values ​​of the corresponding perspectives respectively.

[0029] Furthermore, in the trajectory clustering module, a process trajectory t1 is randomly selected as the center point e1 of the cluster, and in the trajectory similarity matrix W, the process trajectory t1 with a similarity greater than the threshold value is selected. The column index of t1 is used to aggregate the corresponding process tracks and t1 into the same group g1. A new center point e2 is selected from the remaining process tracks, and the center point that meets the threshold is selected. The process trajectories that can be aggregated under the conditions are formed into a new group g2, and the above process is repeated until all the process trajectories are grouped;

[0030] Start the iteration process, randomly select a new process trajectory as the initial center point of this iteration, calculate the clustering results, and The output of the iteration Use function To evaluate the degree of clustering, the formula is defined as follows:

[0031]

[0032] in, Representatives follow the process trajectory turn up The row and column indices corresponding to the matrix, Representative The process trajectory corresponding to the center point of the group, Representative The number of groups calculated in the iterations, represents the balance coefficient, and ;

[0033] like ,but , when the algorithm converges or exceeds the maximum number of iterations, take E best The corresponding clustering results are used as the classification results of the process trajectory ,in .

[0034] Furthermore, in the process discovery module, for n groups containing multiple similar process trajectories, ,in , converted into corresponding n new process log files , and input them into the inductive process discovery algorithm in turn, build a process tree and transform and output multiple process models ;

[0035] Use the RPA instruction converter to filter the process tracks that can be automated in the process model and output the scripts and running instructions of the RPA executable process.

[0036] Furthermore, the content of the instruction converter includes: mapping a complete process trajectory saved in the process model into a script file, including converting each process node into a ScriptAction tag in the script, converting the operation attributes and operation targets required in the process node into ScriptCommand tags, each ScriptCommand contains the complete interaction information required to guide a basic operation, and arranges the order of tags according to the order in the process trajectory.

[0037] An RPA process discovery device based on multi-view clustering includes: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement an RPA process discovery method based on multi-view clustering.

[0038] The beneficial effects brought about by the technical solution provided by the present invention are:

[0039] The present invention proposes an RPA process discovery method based on multi-view clustering, which can extract and encode process trajectory features of different views from event logs, and use a clustering algorithm to realize trajectory clustering based on characteristic values. Then, the similarity between clustering results is calculated and analyzed, the multi-view analysis results are integrated, and the complete process trajectory information is clustered again. The process trajectories with a similarity greater than a threshold are divided into a group. The process discovery algorithm is used to discover multiple process models in multiple groups, identify executable process nodes and trajectories, and convert them into RPA executable instructions and programs, making the process model simple, reducing complexity and difficulty, and saving a lot of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0041] Figure 1 This is an overall framework diagram of an RPA process discovery method based on multi-view clustering in an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of an event log in an embodiment of the present invention;

[0043] Figure 3 is a schematic diagram of a configuration file corresponding to a process trajectory in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of RPA generating running instructions based on a process script in an embodiment of the present invention;

[0045] Figure 5 It is a schematic diagram of the operation of the hardware device in the embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0047] Embodiments of the present invention provide an RPA process discovery method and device based on multi-view clustering.

[0048] Please refer to Figure 1 , Figure 1 This is an overall framework diagram of an RPA process discovery method based on multi-view clustering in an embodiment of the present invention, which specifically includes: a feature calculation module, a process clustering module, and a process discovery module.

[0049] (1) Feature calculation module is used to receive event logs and the number of feature clusters As input, in a process view Next (where ), build a configuration file containing feature information , based on the similarity between feature information , process trajectories with similar eigenvalues Clustering .

[0050] The main methods for calculating process trajectory characteristics include:

[0051] For the event log L, the feature extraction module is used to generate different process view sets P = {p act ,p res ,p trans ,p perform}, and encode the feature information in the profile into the corresponding feature vector, calculate the spatial distance between different feature vectors, and output the similarity matrix between the corresponding process features under the process view. ,in p * ∈P, represents any process view.

[0052] Using the feature clustering module, the process trajectories with high similarity in the matrix are aggregated into the same cluster In, get Number of clusters.

[0053] The specific steps of feature extraction include:

[0054] refer to Figure 2 , for event logs Each process track in , ,in Representative process trajectory Middle Event nodes, each of which records the attributes of the event (including process track identifier , event type , events consume resources , event start and end time ), the attribute feature information under different perspectives is counted and converted into corresponding configuration files, such as Figure 3 shown.

[0055] In any process view Next, different process trajectories Corresponding configuration file , converted into the feature vector corresponding to the process trajectory . Calculate the similarity between any two feature vectors , output the similarity matrix between process trajectories ,in Represents the process view Downstream process trajectory The calculation formula of the distance between vectors is as follows:

[0056]

[0057] The specific methods for generating configuration files include:

[0058] Event type based view , count the frequencies of different event types in the process trace and generate activity profiles ; Based on process resource view , record the utilization of active resources in the process path and generate resource configuration files ; Event-based transition view , focus on the connection order relationship between different nodes in the process trajectory, regard the adjacent process node pairs as an event transition, and generate a configuration file to store the node association relationship ; View based on process operation , quantify the dynamic data generated by the process trajectory during operation, including trajectory running time, the longest event running time, the shortest event running time, the average event node running time and the number of event nodes, etc., and generate a dynamic configuration file .

[0059] The specific steps of feature clustering include:

[0060] Number of clusters based on user initialization , randomly selected in a process view process trajectories as initial cluster centers , extract the corresponding Indexed row data ,for Find the number of rows with the maximum similarity value for each column , merge the process trajectories of the corresponding row and column position indexes to form a clustering result .

[0061] Then the iterative process begins, randomly selecting The new process trajectory is used as the new cluster center of this iteration, the clustering results are calculated, and the function Evaluate the average dissimilarity between the clustered process trajectories and the center points. The formula is defined as follows:

[0062]

[0063] in, Representatives follow the process trajectory turn up The row and column indices corresponding to the matrix, Representative The metric value calculated by the iteration is but , when the algorithm converges or exceeds the maximum number of iterations, take The corresponding cluster assignment is the result of feature-based clustering under this view, and the corresponding trajectory clustering cluster matrix is ​​constructed. ,in ,like Represents the process view Downstream process trajectory Belong to cluster ,otherwise .

[0064] (2) The trajectory clustering module is used to collect and adjust the clusters of process clustering under multiple views, by analyzing different process clusters Room (where ) classification gap , get the clustering similarity between process trajectories Based on adaptive view weight value , fuse the multi-view clustering results into a complete process trajectory similarity, based on the threshold (in ) Clustering to obtain multiple groups containing similar process trajectories .

[0065] The role of the trajectory clustering module is:

[0066] Collect process clustering results from multiple views , in the process view Calculate the cluster similarity between any two process trajectories , output cluster similarity matrix ,in Represents the process view Downstream process trajectory The value of the cluster similarity between .

[0067] Using the multi-view aggregation module, the cluster similarity matrix is ​​aggregated The elements in the Fusion into a process trajectory similarity matrix . Using the trajectory clustering module, the complete event log Divide into n groups containing multiple similar process trajectories , and ensure that the similarity between trajectories within the group is not lower than the threshold .

[0068] Methods for aggregating multi-view clustering results include:

[0069] For process view , set the weight corresponding to the impact process classification In the default scenario, the fixed weight values ​​are .

[0070] Multiply the weight value by the cluster similarity matrix of the corresponding view to obtain the complete process trajectory similarity matrix ,in Representative process trajectory The similarity of multi-view aggregation between , is calculated as follows:

[0071]

[0072] in, represents the normalization function, and Represents the matrix The maximum and minimum values ​​of the row.

[0073] Methods for clustering process trajectories include:

[0074] Randomly select a process trajectory As the center point of the cluster , in the trajectory similarity matrix , filter out the points whose similarity with the center point e1 is greater than the threshold The column index of the corresponding process trajectory and Aggregate into the same group Continue to select new center points from the remaining process tracks , and filter out the points that meet the threshold Under the condition of Repeat the above process until all process tracks are grouped.

[0075] Start the iteration process, randomly select a new process trajectory as the initial center point of this iteration, and calculate the clustering result. Use function To evaluate the degree of clustering, the formula is defined as follows:

[0076]

[0077] in, Representatives follow the process trajectory turn up The row and column indices corresponding to the matrix, Representative The process trajectory corresponding to the center point of the group, Representative The number of groups calculated in the iterations, represents the balance coefficient, and .

[0078] like but , when the algorithm converges or exceeds the maximum number of iterations, take The corresponding clustering results are used as the classification results of the process trajectory ,in .

[0079] (3) The process discovery module is used to receive, mine and transform process information. The process groups in the input process discovery algorithm in turn, output the corresponding process model The instruction converter receives the process model as input and converts the process nodes and trajectories that can be automatically executed into scheduling instructions and programs that can be recognized by the RPA system.

[0080] The functions of the process discovery module are:

[0081] For multiple groups with similar process trajectories , converted into multiple new process log files , and then input them into the InductiveMiner algorithm in turn. By converting the process log into a follow-up activity diagram, and splitting the process nodes based on the follow-up relationship between the process nodes, the process subtree is constructed and converted and output to obtain multiple process models. , where the ways of splitting the following activity diagram include sequential splitting, exclusive splitting, concurrent splitting, and loop splitting.

[0082] Use the RPA instruction converter to filter the process tracks that can be automated in the process model and output the following Figure 4 The RPA shown can execute the script and run instructions of the process to complete the required flowchart.

[0083] Methods for converting process traces into RPA instructions include:

[0084] Use the RPA command converter to map a complete process path into a script file. This involves converting each process node into a ScriptAction tag in the script and converting the required operations in the process node into ScriptCommand tags. The ScriptCommand tag typically contains common RPA command attributes such as the command type (xis:type), command ID (commandId), command name (commandName), command execution status (commandEnable), and command instance name (InstanceName), as well as other specific parameters required for command execution.

[0085] See Figure 5 , Figure 5 4 is a schematic diagram of the working of the hardware device of an embodiment of the present invention, and the hardware device specifically includes: an RPA process discovery device 401 based on multi-view clustering, a processor 402 and a storage device 403.

[0086] An RPA process discovery device 401 based on multi-view clustering: The RPA process discovery device 401 based on multi-view clustering implements the RPA process discovery method based on multi-view clustering.

[0087] Processor 402: The processor 402 loads and executes instructions and data in the storage device 403 to implement the RPA process discovery method based on multi-view clustering.

[0088] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the RPA process discovery method based on multi-view clustering.

[0089] The beneficial effects of the present invention are:

[0090] The present invention proposes an RPA process discovery method based on multi-view clustering, which can extract and encode process trajectory features of different views from event logs, and use a clustering algorithm to realize trajectory clustering based on characteristic values. Then, the similarity between clustering results is calculated and analyzed, the multi-view analysis results are integrated, and the complete process trajectory information is clustered again. The process trajectories with a similarity greater than a threshold are divided into a group. The process discovery algorithm is used to discover multiple process models in multiple groups, identify executable process nodes and trajectories, and convert them into RPA executable instructions and programs, making the process model simple, reducing complexity and difficulty, and saving a lot of resources.

[0091] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An RPA process discovery method based on multi-view clustering, characterized by: include: Feature calculation module, trajectory clustering module and process discovery module; The feature calculation module is used to receive event logs and the number of feature clusters As input, in a process view Build a configuration file containing feature information , based on the similarity matrix between feature information , process trajectories with similar eigenvalues Clustering , ; The feature calculation module includes a feature extraction module and a feature clustering module; For the event log L, the feature extraction module is used to generate different process view sets P = {p act ,p res ,p trans ,p perform }, and encode the feature information in the profile into the corresponding feature vector, calculate the spatial distance between different feature vectors, and output the similarity matrix between the corresponding process features under the process view. ,in p * ∈P, represents any process view; p act The process view representing the event type, p res The process view that represents the process consuming resources, p trans A process view that represents event transitions, p perform A process view that represents the process run; In multiple different process views, the feature clustering module is used to aggregate the process trajectories with high similarity in the matrix into the same cluster. In, get The number of clusters; The trajectory clustering module is used to collect and adjust the clusters of process clustering under multiple views, and analyze different process clusters. The classification gap , get the clustering similarity between process trajectories , based on the adaptive view weight value , fuse the multi-view clustering results into a complete process trajectory similarity, based on the threshold Clustering to obtain multiple groups containing similar process trajectories ,in, , , represents the process trajectory of the kth group, k=1,2,…,n; The process discovery module is used to receive, mine and transform process information. The process groups in the input process discovery algorithm in turn, output the corresponding process model , M k The process model corresponding to the process trajectory of the kth group is represented. The process model is input into the instruction converter, which converts the process nodes and trajectories that can be automatically executed by RPA into scheduling instructions and programs that can be recognized by the corresponding RPA system, thereby completing the required flowchart.

2. The RPA process discovery method based on multi-view clustering according to claim 1, characterized in that: In the feature extraction module, for event logs Each process track in , ,in, Represents the process trajectory t k The i-th event node in each event node records the attribute content of the event, which includes the event type activity, event consumption resource resource, event start and end time timestamp, in different process views P={p act ,p res ,p trans ,p perform }, respectively record the event type frequency, activity resource utilization frequency, the frequency of close occurrence of different event types, and the time information statistics when the process trajectory is running. The time information statistics include the trajectory running time, the longest event running time, the shortest event running time, the average running time of the event node and the number of event nodes, and output the configuration file corresponding to the process trajectory ,in, , n represents the number of complete process trajectories in the log, Respectively represent the corresponding Configuration files; In any process view p * Next, based on different process trajectories Corresponding configuration file , Indicates the first The configuration file corresponding to the process trajectory, k=1,2,…,n, converts the quantified feature information into the feature vector corresponding to the process trajectory , calculate the similarity between any two feature vectors , output the similarity matrix between process trajectories ,in, In the process view Downstream process trajectory The degree of similarity between Represents the distance between two different feature vectors, the distance between vectors The calculation formula is as follows: in, Represents different eigenvectors.

3. The RPA process discovery method based on multi-view clustering according to claim 2, characterized in that: In the feature clustering module, the number of clusters initialized by the user , randomly selected in a process view process trajectories as initial cluster centers , extract the corresponding Indexed row data ,for Find the number of rows with the maximum similarity value for each column , r k Indicates the kth row, merging the process traces of the corresponding row and column position indexes to form a clustering result ; Then the iterative process begins, randomly selecting The new process trajectory is used as the new cluster center of this iteration, the clustering results are calculated, and the function Evaluate the average dissimilarity between the clustered process trajectories and the center points. The formula is defined as follows: Among them, loc() means finding according to the process track id The row and column indices corresponding to the matrix, E j Represents the metric value calculated at the jth iteration, if but , when the algorithm converges or exceeds the maximum number of iterations, take E best The corresponding cluster assignment is the result of feature clustering in this view; based on the clustering results, the corresponding trajectory cluster matrix is ​​calculated and output. ,in, , which means in the process view Next process trajectory t i and cluster c j The similarity of the corresponding cluster centers.

4. The RPA process discovery method based on multi-view clustering according to claim 3, characterized in that: The trajectory clustering module includes a process trajectory clustering module and a multi-view aggregation module; Use the process trajectory clustering module to collect multi-view process clustering results , in the process view Calculate the cluster similarity between any two process trajectories , output cluster similarity matrix ,in, Representative process trajectory In the process view The degree of clustering under In the process view Downstream process trajectory The value of the cluster similarity between ; Using the multi-view aggregation module, the cluster similarity matrix is ​​aggregated The elements in the Fusion into a process trajectory similarity matrix similar T ; Using the trajectory clustering module, the complete event log L is divided into multiple groups containing similar process trajectories. , and ensure that the similarity between trajectories within the group is not lower than the threshold .

5. The RPA process discovery method based on multi-view clustering according to claim 4, characterized in that: In the multi-view aggregation module, for the process view P={p act ,p res ,p trans ,p perform }, set the weight corresponding to the impact process classification , multiply the weight value by the clustering similarity matrix of the corresponding view to obtain the complete process trajectory similarity matrix ,in, Represents the process trajectory t i ,t j The similarity of multi-view aggregation between , is calculated as follows: in, represents the normalization function, and Represent the maximum and minimum values ​​of the i-th row of the matrix, Represents the clustering similarity under different process perspectives, They represent the weight values ​​of the corresponding perspectives respectively.

6. The RPA process discovery method based on multi-view clustering according to claim 5, characterized in that: In the trajectory clustering module, a process trajectory t1 is randomly selected as the center point e1 of the cluster, and in the trajectory similarity matrix W, the process trajectory t1 with a similarity greater than the threshold value is selected. The column index of t1 is used to aggregate the corresponding process tracks and t1 into the same group g1. A new center point e2 is selected from the remaining process tracks, and the center point that meets the threshold is selected. The process trajectories aggregated under the conditions of , form a new group g2, and repeat the above process until all process trajectories are grouped; Start the iteration process, randomly select a new process trajectory as the initial center point of this iteration, calculate the clustering results, and The output of the iteration Use function To evaluate the degree of clustering, the formula is defined as follows: in, Representatives follow the process trajectory turn up The row and column indices corresponding to the matrix, Representative The process trajectory corresponding to the center point of the group, Representative The number of groups calculated in the iterations, represents the balance coefficient, and ; like ,but , when the algorithm converges or exceeds the maximum number of iterations, take E best The corresponding clustering results are used as the classification results of the process trajectory ,in .

7. The RPA process discovery method based on multi-view clustering according to claim 6, characterized in that: In the process discovery module, for n groups containing multiple similar process trajectories, ,in , converted into corresponding n new process log files , and input them into the inductive process discovery algorithm in turn, build the process tree and transform and output the process model ; Use the RPA instruction converter to filter the process tracks that can be automated in the process model and output the scripts and running instructions of the RPA executable process.

8. The RPA process discovery method based on multi-view clustering according to claim 7, characterized in that: The contents of the instruction converter include: mapping a complete process trajectory saved in the process model into a script file, including converting each process node into a ScriptAction tag in the script, converting the operation attributes and operation targets required in the process node into ScriptCommand tags, each ScriptCommand contains the complete interaction information required to guide a basic operation, and arranges the order of tags according to the order in the process trajectory.

9. An RPA process discovery device based on multi-view clustering, characterized by: include: A processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the RPA process discovery method based on multi-view clustering as described in any one of claims 1 to 8.

Images