WO2024004025A1

WO2024004025A1 - Workflow management system, method therefor, and program therefor

Info

Publication number: WO2024004025A1
Application number: PCT/JP2022/025733
Authority: WO
Inventors: 匡生武田; 伸手塚; 拓下沢; 訓大島; 洋司小澤
Original assignee: 株式会社日立製作所
Priority date: 2022-06-28
Filing date: 2022-06-28
Publication date: 2024-01-04

Abstract

The present invention realizes workflow management which makes it possible to reduce the execution cost of a workflow that takes errors into consideration. A computer 1 that manages a workflow has a cost information collection calculation unit 122 and an improvement proposal unit 123. The cost information collection calculation unit 122 uses advance cost information, which has been collected in advance, to calculate the execution costs for one or more tasks to be executed in a workflow and for the entire workflow. The improvement proposal unit 123 identifies a verification process for verifying success of one or more tasks on the basis of execution history information of the workflow, and, via the verification process, detects, prior to task execution, the possibility of occurrence of a task anomaly in the workflow. The improvement proposal unit 123 totals the execution cost of one or more tasks for which execution can be omitted to calculate a task execution cost reduction amount; calculates an overhead, which is the execution cost of the verification process; and, on the basis of the task execution cost reduction amount and the overhead, calculates an improvement expectation amount, which is the amount of reduction in execution costs occurring during workflow operation.

Description

Workflow management system and method and program thereof

The present invention relates to a workflow management system, its method, and its program, and particularly relates to information processing that supports workflow design improvement.

In recent years, workflow management systems have been attracting attention as a technology that automates business processes and services and improves operational efficiency. In a workflow management system, a developer uses workflow to define processing logic for business processes, services, etc., and automatically creates instances of the workflow and executes the processing logic in response to requests from workers. A workflow is composed of tasks that define processing, arrows that define the order of execution, and the like, and may be defined by a workflow notation standard such as BPMN (Business Process Model and Notation). Tasks include tasks that wait for completion by a worker, tasks that call external services via API, etc.

The workflow management system uses a low-code design method that allows workflow development even by developers who lack skills in software development using program code. On the other hand, it is still difficult for developers to develop a workflow that takes into consideration execution costs such as processing time and fees that occur when executing a workflow. For example, there is a technology disclosed in Patent Document 1 that supports reduction of workflow execution costs. This technology uses schedule information of candidate workers to execute a task and automatically determines the appropriate worker from the perspective of availability and reliability, reducing the task failure rate and ensuring the execution of the entire workflow. It can support cost reduction. Other prior art techniques for reducing execution costs include techniques for selecting API call destinations that minimize execution fees, and techniques for reducing processing time by identifying tasks that can be executed in parallel.

US Patent Publication 2017/0337492A1

The technology disclosed in Patent Document 1 does not sufficiently reduce the execution cost by taking into account workflow abnormalities. Here, an abnormal situation refers to a situation where the purpose of the workflow cannot be achieved due to the occurrence of an exception in a task, etc. In an abnormal system, the more a workflow task is executed, the more unnecessary execution costs that do not contribute to achieving the goal of the workflow increase. The inventor believes that the workflow optimization method disclosed in Patent Document 1 makes it difficult to reduce the execution cost by taking abnormalities in the workflow into consideration. Therefore, we believe that there is a need for some kind of method that can detect the possibility of a task abnormality occurring early and terminate the workflow execution early.

In addition, as a method to detect the possibility of an abnormality occurring in a task early and terminate the workflow execution early, a task that detects the possibility of an abnormality occurring in one or more subsequent tasks (hereinafter referred to as verification processing) is added to the workflow. There are ways to add it. However, since the verification process itself is costly to execute, inappropriately adding a verification process will result in an increase in the execution cost of the entire workflow, making appropriate implementation difficult.

An object of the present invention is to realize workflow management that can reduce the cost of executing a workflow that takes abnormal systems into consideration.

In a preferred example of the workflow management system according to the present invention, a computer that manages workflow has a cost information collection calculation section and an improvement proposal section,
The cost information collection calculation unit calculates the execution cost of one or more tasks to be executed in the workflow and the entire workflow using cost advance information collected in advance,
The improvement proposal department is
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
The present invention is a workflow management system that calculates an expected improvement amount, which is a reduction amount of execution costs generated during operation of a workflow, based on the task execution cost reduction amount and the overhead.

In a preferred example of the workflow management method according to the present invention, a computer that manages the workflow executes a cost information collection calculation step and an improvement proposal step,
The cost information collection calculation step calculates the execution cost of one or more tasks to be executed in the workflow and the entire workflow using cost advance information collected in advance,
The improvement proposal step includes:
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
The present invention is a workflow management method that calculates an expected improvement amount, which is a reduction amount of execution costs generated during operation of a workflow, based on the task execution cost reduction amount and the overhead.

A preferred example of the workflow management program according to the present invention is a program that is executed by a computer that manages workflow, and includes:
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
Based on the task execution cost reduction amount and the overhead, calculate an expected improvement amount that is the reduction amount of the execution cost that occurs during operation of the workflow;
This is a program that executes .

According to the present invention, a workflow can be ended before a task is executed and an execution cost is incurred, and unnecessary execution costs that do not contribute to achieving the goal of the workflow can be reduced. This makes it possible to reduce the cost of executing a workflow that takes abnormal systems into consideration.

FIG. 1 is a diagram showing the configuration of a workflow management system according to an embodiment. FIG. 3 is a diagram illustrating an example of a workflow definition. 3 is a diagram showing an example of workflow definition information 131. FIG. 3 is a diagram illustrating an example of workflow instance information 132. FIG. 3 is a diagram illustrating an example of workflow execution history information 133. FIG. 13 is a diagram showing an example of cost advance information 134. FIG. 3 is a diagram illustrating an example of incurred cost information 135. FIG. 3 is a diagram showing an example of learning model information 136. FIG. 13 is a diagram showing an example of improvement plan information 137. FIG. 3 is a diagram showing an example of an improvement proposal template 138. FIG. 3 is a diagram showing a processing flow of the cost information collection calculation unit 122. FIG. 3 is a diagram showing a processing flow of an improvement plan estimating unit 123. FIG. FIG. 6 is a diagram showing a processing flow of an improvement method identification step 2200. FIG. 6 is a diagram showing a processing flow of an improvement amount evaluation step 2300. FIG. 3 is a diagram showing an example of an improvement proposal screen.

An embodiment of the present invention will be described below with reference to the drawings.
In this embodiment, an example in which a workflow management system is applied to a design support system will be described. The design support system uses information about execution costs such as processing time and fees for workflows (hereinafter referred to as flows), as well as flow execution history information that records changes in the state of flow instances, to automatically perform the verification processing presented. This is a system that estimates using a learning model, evaluates the expected amount of improvement, and presents candidates for verification processing to the developer. A flow instance represents the actual execution of processing logic defined by a flow, and a flow instance is started by a flow start command from a flow worker.

FIG. 1 shows an example of the configuration of a computer used in the workflow management system according to this embodiment. The computer 1 is connected to, for example, a processing unit (CPU) 10 that executes programs, a storage unit 20 such as a semiconductor memory or hard disk that retains programs and data, an input/output interface (I/F) 30, and a network 9. The network interface (I/F) 40 is configured.

A terminal 7 used by a developer/operator and a terminal 8 used by a flow worker are connected to the computer 1 via a network 9, and users such as the developer/operator and flow worker can use these terminals. The computer 1 can be used by using Here, the terminals 7 and 8 are connected to an input section that can be operated by a user, a display section that displays data, a processing section that processes data by executing a program, a memory that stores data and programs, and a network 9. It has a network connection section.

The processing unit 10 of the computer 1 includes a user interface (abbreviated as UI) generation unit 11 and a functional unit 12.

The UI generation unit 11 has a development/operation UI 111, a flow work UI 112, and an improvement proposal UI 113. The UI generation unit 11 generates a user interface (UI) used in the terminal 7 and terminal 8 connected to the computer 1. The user can view the screens displayed on the terminals 7 and 8 and perform input operations using the UI. In one example, individual programs for the development/operation UI 111, the flow work UI 112, and the improvement proposal UI 113 are executed by the processing unit 11 to realize their respective functions. Note that one UI program having the functions of multiple UI programs may realize multiple functions.

The improvement proposal UI 113 is realized by executing a program that uses the improvement proposal information 137 managed by the storage unit 20 to generate a UI that presents a flow design improvement proposal to a flow development operator. The improvement proposal UI 113 can provide a screen that displays improvement proposals (for example, verification processing and estimated amount of improvement), and can support development and operation personnel in improving the design of appropriate flows.

The functional unit 12 includes a flow execution management unit 121, a cost information collection calculation unit 122, and an improvement proposal unit 123. These functional units include, for example, a flow execution management program 121′, a cost information collection processing program 122′, and an improvement proposal program 123′ stored in the storage unit 13, which are executed by the processing unit 10 to perform their respective functions. (The code of each program is not shown).

The flow execution management unit 121 executes and manages flows. Examples of basic functions include, for example, a function to store information regarding flow definitions sent from the developer/operator terminal 7 into the flow definition information 131 managed by the storage unit 20, and a function to store information regarding flow definitions sent from the developer/operator terminal 7 into the flow definition information 131 managed by the storage unit 20; This includes a function of creating a flow instance from flow definition information 131 and storing it in flow instance information 132 based on an execution start command, and a function of recording changes in the state of the executed flow instance in flow execution history information 133.

In order to calculate the execution cost of a flow instance from the flow execution history information 133, the cost information collection calculation unit 122 collects cost advance information 134 required in advance, and calculates the execution cost for each flow instance or each task within the flow instance. is calculated and recorded in the incurred cost information 135. The cost advance information 134 includes, for example, personnel costs incurred in human tasks, API usage costs for calling external APIs, and costs for using computational resources during task execution. Note that the execution cost may be determined using processing fees, processing time, or other indicators.

The improvement proposal unit 123 uses the flow execution history information 133 and the generated cost information 135 to update the learning model information 136 regarding the machine learning model for estimating the improvement plan, and calculates the improvement plan information 137. Note that the learning model information 136 may not be managed and held by the storage unit 20 depending on the machine learning method, for example, when a new model is created each time in batch processing. The improvement proposal information records information regarding the improvement proposal presented to the flow development operator, and is converted into a UI by being substituted into the improvement proposal template 138 by the improvement proposal UI 113.

The storage unit 20 holds flow definition information 131, flow instance information 132, flow execution history information 133, cost advance information 134, incurred cost information 135, learning model information 136, improvement plan information 137, and improvement proposal template 138. This information is configured, for example, in a table format. Note that configuration examples of each information will be described later with reference to FIG. 3 and subsequent figures.

FIG. 2 shows a definition example of a flow executed and managed by the computer 1.
The flow definition G10 includes events that indicate the start and end of flow execution such as Start event G1010 and NormalEnd event G1070, tasks that define specific processes to be executed in the flow such as Task1 task G1030 and Task2 task G1050, and Arrow1 arrow G1020. It includes arrows that define the execution order and branching of events and tasks, such as Arrow2 arrow G1040. For example, in a human work task, the worker and the work content are defined in the task, and in the task that calls an external service, the URL information of the API connection destination and information on variables to be input are defined. Each task is set with a cost type, which is prior information regarding the execution cost included in the cost prior information 134, and this cost type allows the execution cost of each task to be calculated from the flow execution history information 133 of this flow. .

The ErrorEnd1 event G1090 and the ErrorEnd2 event G1110 shown in FIG. 2 are end events corresponding to the case where an abnormality occurs in the Task1 task G1030 or the Task2 task G1050 and the task cannot be executed normally. In this embodiment, for example, by proposing to the developer that a verification process, which is a task that verifies the success of Task2 task G1050, be inserted into Arrow1 arrow G1020, a flow instance that occurs abnormally in Task2 task G1050 can be removed from Task1 task G1030. Reduce the execution cost of a flow instance by causing it to fail before execution.

FIG. 3 shows an example of the table structure of the flow definition information 131.
The flow definition information 131 has information regarding flows designed by a flow developer/operator, and is added and updated by the flow execution management unit 121. The flow definition table T1000 includes columns such as flow definition IDT1100, name T1200, version T1300, and definition content T1400.

In the definition content T1400, a flow definition is recorded in a form that can be read by a computer. For example, it may be recorded using xml. In the flow definition, as described above, a cost type is set for each task definition, which allows the cost information collection calculation unit 122 to calculate the execution cost generated by the execution of this task. Additionally, each task or event definition has a unique ID set within each flow, which is used to indicate the execution position of the flow instance being executed.

FIG. 4 shows an example of the table structure of the flow instance information 132.
The flow instance information 132 includes information regarding the state of a flow instance created by a flow worker's start command from the terminal 8. The flow instance table T2000 includes columns such as a flow instance IDT2100, a flow definition IDT2200, a current task IDT2300, and a flow variable T2400. Each record of the flow instance table T2000 corresponds to each flow instance being executed, and the corresponding record is updated every time the flow instance is progressed by the flow execution management unit 121.

The current task IDT 2300 and flow variable T2400 represent information regarding the current execution position and state of each flow instance. In the flow variable T2400, variables held by a flow instance are recorded, and variables are added and updated in each task. In the flow variable T2400, the variable name and value may be expressed in a format other than a dictionary, or some or all of the information similar to the flow variable may be recorded as an individual table. Further, the flow variable T2400 broadly includes information regarding the state of the flow instance, and may include, for example, the initiator of the flow instance, authentication information when calling an external service, and the like.

FIG. 5 shows an example of the table configuration of the flow execution history information 133.
The flow execution history information 133 has records added by the flow execution management unit 121 at the start and end of a flow instance or task, and has information regarding the state of the flow instance at a specific point in time. The flow execution history table T3000 includes columns such as a flow history IDT3100, a flow instance IDT3200, an event occurrence time T3300, an event content T3400, and a flow variable T3500.

The event occurrence time T3300 and the event content T3400 represent information regarding the time and content of an event such as the start of a flow instance or a task. Note that when an event occurs, such as when a task starts or when a flow occurs, a record is to be recorded, which may be defined by a separately defined policy or table. From the information on the event occurrence time T3300 and the event content T3400, it is possible to calculate, for example, which route in the flow definition a particular flow instance took, how much time each task took, and the like.

The flow variable T3500 is recorded based on the flow variable T2400 of the flow instance at the time of each record creation. Note that, as described in the explanation of the flow variable T2400, the flow variable T3500 broadly includes information regarding the state of the flow instance, and may include, for example, the initiator of the flow instance, authentication information when calling an external service, and the like. Using the information on the flow variables T3500, it is possible to collect historical information such as when each flow variable was added and updated.

FIG. 6 shows an example of the table structure of the cost advance information 134.
The cost advance information 134 defines information necessary in advance to calculate the generated cost information 135 from the flow execution history information 133. The cost type table T4000 is set for each task definition in a flow definition, and has information regarding the calculation method and calculation parameters for costs generated when the task is executed. The cost type table T4000 includes columns such as cost type IDT4100, cost type name T4200, fixed cost T4300, and time-varying cost T4400. Each record of the cost type table T4000 is collected by the cost information collection calculation unit 122 in one example, but may be prepared in advance by the provider of the computer 1, or may be created and customized by the flow development manager. good.

The fixed cost T4300 and the time-varying cost T4400 represent parameters necessary for calculating the task execution cost. The fixed cost T4300 is a cost that occurs with respect to the number of times a task is executed, and includes, for example, an API usage fee for a task that calls an external service that is charged for the number of executions. The time-varying cost T4400 is a cost that occurs in proportion to the execution time of a task calculated from the flow execution history information 133, and includes, for example, the labor cost of a human task, or the cost of a task that calls an external service that is charged for execution time. This applies to API usage fees, etc.

Note that the method of calculating the task execution cost is not limited to the method using the fixed cost T4300 or the time-varying cost T4400. For example, other parameters depending on the task execution time are defined, and the task execution cost obtained from the flow execution history information 133 is Information other than time may be used. Furthermore, the method is not limited to the method of defining calculation parameters in a table, such as the fixed cost T4300 and the time-varying cost T4400, but for example, information such as the storage location of the execution cost managed outside the storage unit 20, such as defining a calculation formula. May include.

FIG. 7 shows an example of the table structure of the incurred cost information 135.
The incurred cost information 135 includes, for example, information on a flow cost table T5000, which is an incurred cost table for each flow instance, and a task cost table T6000, which is an incurred cost table for each task of a flow instance. Note that the generated cost information 135 may include, for example, a table that records information regarding execution costs for each flow instance and other than tasks, and statistical information such as the average and total of execution costs for each flow definition. . Further, if the calculation load for calculating the generated cost information 135 from the flow execution history information 133 is light, the generated cost information 135 does not need to be managed and held by the storage unit 20.

The flow cost table T5000 and the task cost table T6000 have information T5100 and T6100 regarding the execution cost of each flow instance and each task. The cost T5300 and cost T6400 columns representing the execution cost are calculated by the cost information collection calculation unit 122 based on the flow execution history information 133 and the cost advance information 134. Note that the cost T5300 and the cost T6400 are not limited to having only one numerical data, and may include a plurality of values or values in a discrete format, such as a processing fee and a processing time, for example.

FIG. 8 shows an example of the table structure of the learning model information 136.
The learning model table T7000 has information regarding a machine learning model used when estimating the improvement plan information 137 using the flow execution history information 133 and the generated cost information 135. The learning model table T7000 includes a model IDT7100, a flow definition IDT7200, a verification process insertion location T7300, a model type T7400, a learning parameter T7500, a model object T7600, and the like. Note that information similar to the learning model information 136 may not be held as a table depending on the machine learning method, for example, when a new model is created each time in batch processing.

The flow definition IDT 7200 and the verification process insertion location T7300 represent which flow definition and arrow location are used by the machine learning model held in each record to estimate the verification process. As mentioned above, the verification process is a task of verifying the success of one or more tasks. Note that the verification process insertion location T7300 does not need to be limited to an arrow, and may include, for example, before and after processing a task or event.

The model type T7400 represents information regarding the type of machine learning model held in each record. The model type T7400 inputs the flow instance state at the verification process insertion point T7300 extracted from the flow execution history information 133, and the end code of the flow instance, such as the Normal End event G1070 or Error End1 event G1090 in FIG. 2. A machine learning model that can output the estimation result can be used, for example, a NN (Neural Network) model or a decision tree model may be used.

The model object T7600 represents information regarding a model object that can be executed by the improvement proposal unit 123. Note that it is not necessary that the model object is directly recorded in the model object T7600; for example, the model object may be recorded in another table or file system, and only a reference thereof may be retained.

FIG. 9 shows an example of the table structure of the improvement plan information 137.
The improvement plan information 137 includes information regarding an improvement plan for adding the verification process estimated and evaluated by the improvement proposal unit 123. The improvement plan table T8000 includes information regarding the contents of the verification process plan, insertion locations, and cost reduction effects. The improvement proposal table T8000 includes improvement proposal IDT8010, flow definition IDT8020, verification process insertion location T8030, improvement proposal type IDT8040, improvement proposal parameter T8050, false positive rate T8060, task execution cost reduction amount T8070, overhead T8080, expected improvement amount T8090, Contains a valid flag T8100, etc.

The flow definition IDT 8020 and the verification process insertion location T8030 represent information regarding the insertion position of the verification process. The flow definition IDT8020 and the verification process insertion location T8030 include information equivalent to the flow definition IDT7200 and the verification process insertion location T7300 of the learning model table T7000 of the machine learning model used when estimating the verification process for each record. . Note that the records in the learning model table T7000 and the improvement plan table T8000 do not need to have a one-to-one correspondence, and a plurality of improvement processing plans may be generated from one machine learning model.

The improvement proposal type IDT8040 and the improvement proposal parameter T8050 represent information regarding the contents of the verification process. The improvement proposal type IDT 8040 specifies template information regarding the improvement proposal included in the improvement proposal template 138, and the improvement proposal parameter T8050 specifies the contents of individual verification processing and the contents of the improvement proposal to be displayed by the improvement proposal UI 113. The improvement proposal parameter T8050 includes, for example, a verification formula for verification processing by checking flow variables, URL information for verification processing by calling an API, and the like. Note that the information regarding the content of the verification process does not necessarily have to be defined using the improvement proposal template 138, and for example, the definition of a task to perform the verification process may be recorded.

The false positive rate T8060, task execution cost reduction amount T8070, overhead T8080, expected improvement amount T8090, and effectiveness flag T8100 represent information regarding the effectiveness of the verification process. The false positive rate T8060 represents the probability that the verification process will erroneously classify a normal flow instance as an abnormal flow instance. Based on the false positive rate T8060, for example, even if the verification process is highly evaluated for the expected improvement amount T8090, if the false positive rate T8060 is high, it is possible to judge that it will not be adopted because it is considered to have a negative impact on the normal processing of the flow. . The task execution cost reduction amount T8070, overhead T8080, and expected improvement amount T8090 represent information regarding the amount of influence on the flow execution cost due to the verification process. The expected improvement amount T8090 is evaluated by subtracting the overhead T8080, which is the execution cost of the task itself that performs the verification process, and it is possible to evaluate whether it is truly effective in reducing the execution cost of the flow. The validity flag T8100 becomes "True" when the effect of the verification processing, such as the false positive rate T8060 or the expected improvement amount T8090, satisfies certain criteria, indicating that it can be presented by the improvement proposal UI 113. The validity flag T8100 enables control such as not presenting verification processing with a high false positive rate T8060 or verification processing that cannot be expected to reduce execution costs.

FIG. 10 shows an example of the table structure of the improvement proposal template 138.
The improvement proposal template 138 has information regarding the template of the content of the verification process and the presentation method of the verification process, and is used when the improvement proposal UI 113 presents the improvement proposal information 137 to the terminal 7 . The improvement proposal template table T9000 includes an improvement proposal type IDT9100, a name T9200, a verification processing task T9300, a message T9400, a cost type T9500, a cost T9600, and the like. The improvement proposal template table T9000 may be prepared in advance by the provider of the computer 1, or may be created and customized by the flow development manager.

The verification processing task T9300 has information regarding the task definition template of the verification processing to be presented. The verification processing task T9300 is defined in, for example, the xml format in accordance with the notation format of the flow definition information 131, and variables to be assigned to the template are expressed, for example, in curly braces as "{script}". It's good. If it is possible to replace the part defined as a template with a variable, other methods may be used.

Message T9400 represents information regarding a message template to be presented when making an improvement proposal. Message T9400 has information regarding the explanation of the contents of the verification process and the effects of the verification process, which allows the flow development operator to understand the contents and effects of the presented verification process and determine whether or not to insert the verification process. be able to. A variable to be assigned to a template may be expressed, for example, in curly braces as "{var}". Note that a plurality of columns containing information similar to message T9400 may be prepared depending on the presentation method and language.

The cost type T9500 and cost T9600 represent information regarding overhead, which is the execution cost of a task that performs verification processing. The cost type T9500 represents information regarding the cost type of the task that performs the verification process, and specifies the cost type IDT4100. The cost type T9500 can be used to estimate overhead. For example, if a task for which the same cost type is specified exists in a flow for which an improvement proposal is made, the average execution cost of that task can be used as the estimated value of overhead. can. Note that cost type T9500 may be included in the task definition of verification processing task T9300. Cost T9600 is an estimated value of overhead, and can be used, for example, when cost is not estimated by cost type T9500. Note that the value of the cost T9600 may be modified or updated, for example, from the execution history of the task that performs the verification process that was actually inserted based on the presentation.

FIG. 11 shows the processing flow of the cost information collection calculation unit 122.

The cost information collection calculation unit 122 executes the following processing steps.

In step S1100, the process of the cost information collection calculation unit 122 is started, and the process advances to step S1200.

In step S1200, preliminary cost information 134 is collected and recorded. Note that, as described above, the cost advance information 134 may be prepared in advance by the provider of the computer 1, or information created and customized by the flow development manager may be collected. After completing the collection and recording of cost advance information 134, the process advances to step S1300 for calculating incurred cost information.

In the incurred cost information calculation step S1300, the incurred cost information 135 is calculated. The generated cost information calculation step S1300 may be executed asynchronously every time a flow instance termination signal transmitted from the flow execution management unit 121 is received. Note that the execution timing of the generated cost information calculation step S1300 is not limited to the timing at each end of a flow instance, and may be executed, for example, at each task of a flow instance or at each end of a plurality of flow instances. The generated cost information calculation step S1300 ends waiting for a flow instance end signal when the computer 1 or the cost information collection calculation unit 122 ends, and proceeds to step S1400.

In step S1310, the process of the incurred cost information calculation step S1300 is started, and the process proceeds to step S1320.

In step S1320, the execution cost is calculated for each task of the completed flow instance based on the flow definition information 131, flow execution history information 133, and cost advance information 134. As described above, a cost type is set for each task definition in the flow definition information 131, and the execution cost of each task can be calculated by referring to the record of the corresponding cost type table T4000. After completing the calculation of the execution cost of each task, the process advances to step S1330.

In step S1330, the execution cost of the flow instance is calculated by summing the execution costs of each task calculated in step S1320. Note that the method for calculating the execution cost of a flow instance is not limited to summing the execution cost of each task; for example, when there are elements other than tasks that should take execution cost into consideration, or when a certain processing cost is taken into account when creating a flow instance. If necessary, these factors may be taken into consideration in the calculation. After completing the calculation of the execution cost of the flow instance, the process advances to step S1340.

In step S1340, the execution cost of each task calculated in step S1320 and the execution cost of the flow instance calculated in step S1330 are recorded in the flow cost table T5000 and task cost table T6000 of the generated cost information 135. Note that in step S1340, for example, information regarding execution costs for each flow instance or other than tasks, and statistical information such as the average or total of execution costs for each flow definition may be calculated and recorded. Once recording to the incurred cost information 135 is completed, the process advances to step S1350.

In step S1350, the process of step S1300 for calculating the generated cost information regarding the received flow instance end signal is ended. However, as described above, the generated cost information calculation step S1300 continues to wait for a flow instance end signal, and the same process is asynchronously executed from step S1310 every time a flow instance end signal is received.

In step S1400, the process of the cost information collection calculation unit 122 ends.

FIG. 12 shows the processing flow of the improvement plan estimation unit 123.

The improvement plan estimation unit 123 executes the following processing steps.

When the process of the improvement plan estimating unit 123 is started, in step S2100, the flow definition to be presented and estimated and evaluated for the improvement plan is specified using the flow definition IDT 1100. Once the flow definition is specified, the process advances to improvement method identification step S2200.

In the improvement method identification step S2200, a column regarding the contents of the verification process in the improvement proposal table T8000 is calculated and recorded. Through the improvement method identification step S2200, a group of candidates for verification processing can be obtained. (Details of this processing step S2200 will be described later with reference to FIG. 13.) Once the column calculation for the content of the verification process is completed in the improvement method identification step S2200, the process advances to improvement amount evaluation step S2300.

In the improvement amount evaluation step S2300, a column related to the evaluation of the verification process in the improvement plan table T8000 is calculated and recorded. In the improvement amount evaluation step S2300, the effectiveness of the candidate group of verification processes estimated in the improvement method identification step S2200 is evaluated, and information for determining whether each verification process should be presented and the priority of presentation is calculated. be able to. (The details of this processing step S2300 will be described later with reference to FIG. 14.) Once the column calculations related to the evaluation of the verification process are completed, the process advances to step S2400.

In step S2400, the processing of the improvement plan estimation program 123 ends.

FIG. 13 shows the processing flow of the improvement method identification step S2200.

In step S2210, the process of improvement method identification step S2200 is started, and the process proceeds to step S2220.

In step S2220, the repeating process starts for each arrow in the specified flow definition, with the repeating range from step S2220 to step S2270. Note that the target of repeated processing is not limited to arrows; for example, if the before and after processing of a task or event are included in the insertion position candidates for the verification process, these may be the targets of repeated processing. Further, it is not necessary to perform repeated processing on all insertion position candidates for verification processing, and it is also possible to appropriately narrow down the repeating targets, such as performing repeated processing only on the arrows of the route in which normal processing is to be performed. Further, the repetitive processing may be performed in parallel. After specifying the flow definition arrow, the process advances to step S2230.

In step S2230, a data set used for learning the machine learning model is created from the flow execution history information 133. From the flow execution history information 133, among the records of the flow execution history table T3000 of the flow instance of the specified flow definition, a set of information regarding the flow state at the time of the specified arrow and an exit code at the end of the flow instance is extracted, Create a dataset. The exit code at the end of the flow instance represents information regarding the execution result of the flow instance, and may use, for example, the end event name (eg, NormalEnd event G1070 or ErrorEnd1 event G1090 in FIG. 2). Note that the method for creating the data set is not limited to the method shown in step S2230, and a data set in other formats may be created, such as by using the flow definition information 131, for example. In addition, it is not necessary to create a dataset based on all records in the flow execution history table T3000; for example, it is possible to selectively use records that have not been used for learning in the past, such as using the most recent records within a certain period of time. Also good. After creating the data set, the process advances to step S2240.

In step S2240, the learning model information 136 is checked and records that can be used as initial values of the machine learning model are extracted. By reusing previously trained models, model learning can be made more efficient. Note that, as described above, depending on the machine learning method, the learning model information 136 may not need to be used or recorded, for example, when a new model is created each time by batch processing. After confirming the learning model information 136, the process advances to step S2250.

In step S2250, a machine learning model is trained using the data set created in step S2230. Note that the processing device that performs learning is not limited to the processing unit 20, and a processing device other than the processing unit 20 included in the computer 1 or another computer may be used. When the learning of the machine learning model is completed, the learning model information 136 is updated and the process advances to step S2260.

In step S2260, the content of the verification process is specified using the machine learning model learned in step S2250. The method for specifying the content of the verification process differs depending on the type of learning model and how it is used, but for example, in a method using a decision tree, the content of the verification process can be specified using information on the branch conditional expression. Further, the content of the verification process to be specified is not limited to the case where it is specifically formulated and specified; for example, in the case of a deep learning model, the model itself may be the content of the verification process. Once the content of the verification process is specified, it is recorded in the corresponding location of the improvement proposal information 137 (for example, improvement proposal type IDT 8040, improvement proposal parameter T8050, etc.), and the process proceeds to step S2270.

In step S2270, it is determined whether the repeated processing from step S2220 to step S2270 has ended. If the repetitive process is to be continued, the process advances to step S2220, and if it is to be terminated, the process is to proceed to step S2280.

In step S2280, the process of improvement method identification step S2200 is ended. The verification processing candidate group identified in the improvement method identification step S2200 (that is, for presentation) is recorded in the improvement proposal information 137.

FIG. 14 shows the processing flow of the improvement amount evaluation step S2300.

In the improvement amount evaluation step S2300, the execution cost reduction effect of the verification processing candidate group obtained in the improvement method identification step S2200 is evaluated.

In step S2301, the process of improvement amount evaluation step S2300 is started, and the process proceeds to step S2302.

In step S2302, iterative processing is started for the group of verification processing candidates recorded in the improvement proposal information 137. A record to be evaluated is specified from the improvement proposal information table T8000, and the process advances to step S2303.

In step S2303, as a pre-process to evaluate the effectiveness of the verification process, the flow execution history information 133 is used to apply the verification process to previously executed flow instances. From the flow execution history information 133, flow instance information at the time of passing through the insertion position of the verification process is collected, and it is determined whether the verification process is applicable. After that, the process advances to step S2304.

In step S2304, a false positive rate T8060 is calculated, which is the probability that the verification process will erroneously classify a normal flow instance as abnormal. The false positive rate T8060 can be calculated by summing up the number of normally completed flow instances based on the result of applying the verification process to the flow execution history in step S2303, and dividing the number classified as abnormal by the verification process. Note that the false positive rate T8060 is an index for determining whether the verification process does not adversely affect normal processing, and is not limited to the false positive rate, but one or more other indicators may be used. The calculated false positive rate is recorded in false positive rate T8060 of the improvement plan information 137.

In step S2305, it is confirmed whether the false positive rate T8060 does not exceed a certain threshold. In step S2305, verification processing that may have a negative impact on normal processing of the flow can be removed from the presentation candidates. Note that the threshold value may be determined in advance by the computer 1, or may be set from the terminal 7 by the developer/operator. If the false positive rate T8060 does not exceed a certain threshold, the process advances to step S2306, and if the false positive rate T8060 exceeds the certain threshold, the process advances to step S2311.

In step S2306, a task execution cost reduction amount T8070, which is the expected value of the total execution cost of tasks whose execution could be omitted through the verification process, is calculated. Based on the result of applying the verification process to the flow execution history in step S2303, the task execution cost is reduced by summing the execution costs of tasks whose execution could be omitted due to the verification process for all flow instances and dividing by the total number of flow instances. The quantity T8070 can be calculated. The task execution cost reduction amount T8070 is a numerical value representing the expected execution cost reduction value per flow instance when the overhead of verification processing is ignored. The calculated task execution cost reduction amount T8070 is recorded in the improvement proposal information 137, and then the process advances to step S2307.

In step S2307, the overhead T8080 of the verification process is calculated. Information such as cost type T9500 and cost T9600 that the improvement proposal template 138 has is used to calculate the overhead T8080. If a task with the cost type specified as cost type T9500 already exists in the flow definition, the overhead T8080 can be estimated by finding the average execution cost of that task from the generated cost information 135. Furthermore, if there is no task with the cost type specified in cost type T9500, cost T9600 is used as the estimated value of overhead T8080. Note that the method for calculating the overhead T8080 is not limited to the method of step S2307, and information on cost type T9500 and cost T9600 may be used in combination. The overhead T8080 of the verification process is recorded in the improvement proposal information 137, and then the process advances to step S2308.

In step S2308, the expected improvement amount T8090 is calculated by subtracting the overhead T8080 from the task execution cost reduction amount T8070. The expected improvement amount T8090 is a numerical value representing the expected reduction in execution cost per flow instance, taking into account the overhead T8080. By using the expected improvement amount T8090, it is possible to present a verification process that is truly effective in reducing flow execution costs. After the calculated expected improvement amount T8090 is recorded in the improvement plan information 137, the process advances to step S2309.

In step S2309, it is confirmed whether the expected improvement amount T8090 exceeds a certain threshold value. In step S2309, verification processing that is not expected to improve the execution cost of the flow can be removed from the presentation candidates. Note that the threshold value may be determined in advance by the computer 1, or may be set by the developer/operator from the terminal 7. If the expected improvement amount T8090 exceeds a certain threshold, the process advances to step S2310, and if the expected improvement amount T8090 does not exceed the certain threshold, the process advances to step S2311.

In step S2310, by setting the validity flag T8100 to "True", it is recorded that the verification process can be presented. Additionally, the terminal 7 may be notified that the verification process can be presented. This allows the developer/operator to know that suggestions for improvement are possible. After setting the validity flag T8100, the process advances to step S2312.

In step S2311, the validity flag T8100 is set to "False" to record that the verification process cannot be presented. Note that the verification processes that cannot be presented may be deleted from the improvement proposal table T8000, and may be deleted at once before the end of the improvement amount evaluation step S2300, for example, in step S2311. After setting the validity flag T8100, the process advances to step S2312.

In step S2312, it is determined whether the repeated processing from step S2302 to step S2312 has ended. If the repetitive process is to be continued, the process advances to step S2302; if it is to be terminated, the process is to proceed to step S2313.

In step S2312, the process of improvement amount evaluation step S2300 is ended. Through the improvement amount evaluation step S2300, the evaluation of the effectiveness of the verification processing candidate group of the improvement plan information 137 is completed.

FIG. 15 shows an example of an improvement proposal screen provided by the computer 1.
The improvement proposal screen is displayed on the developer/operator's terminal 7 when the processing unit 20 executes the improvement proposal UI 113. The improvement proposal UI 113 can be executed, for example, by the developer/operator periodically or as needed, or upon receiving a notification that the verification process can be presented after the process in step S2310. can.

The improvement proposal screen G2000 includes a flow selection section G2100 for selecting a flow definition for which an improvement is proposed, an improvement presentation section G2200 for displaying the contents and effects of the improvement proposal on the illustrated flow, and selection of an improvement proposal to be displayed from a list. The improvement proposal selection unit G2300 includes an improvement proposal selection unit G2300.

The flow selection unit G2100 presents information from the flow definition information 131 about flows for which improvement proposals can be made. The selected flow definition is colored, for example, gray.

The improvement presentation unit G2200 displays the flow from the flow definition information 131, and displays the contents and effects of the improvement proposal from the improvement proposal information 137. The improvement presentation section G2200 includes a verification processing presentation section G2210 and an abnormal cost presentation section G2220.

The verification process presentation unit G2210 refers to and presents information regarding the content of the verification process and the insertion position of the improvement plan information 137. The verification process presentation unit G2210 includes a message G2211 and a task definition G2212.
Message G2211 is a text that explains the contents and effects of the verification process for the developer. Note that a message obtained by substituting the improvement proposal parameter T8050 of the improvement proposal information 137 into the message T9400 of the improvement proposal template 138 may be displayed.
The task definition G2212 represents a task that performs a verification process, and a result obtained by substituting the improvement proposal parameter T8050 of the improvement proposal information 137 for the verification process task T9300 of the improvement proposal template 138 is displayed. The task definition G2212 may be provided with a mechanism that facilitates the implementation of improvement proposals, such as being able to be inserted into the flow immediately by UI operation, for example.

The abnormal system cost presentation unit G2220 presents abnormal system statistical information that can reduce the execution cost through verification processing. The abnormal system cost presentation unit G2220 can make the developer aware of the necessity of improving the execution cost. The abnormal cost presentation unit G2220 includes an execution cost message G2221 and an execution cost graph G2222. The execution cost message G2221 and the execution cost graph G2222 use the generated cost information 135 to display how much the abnormal system accounts for in the total execution cost of the flow.

The improvement proposal selection unit G2300 extracts improvement proposals for the flow definition specified by the flow selection unit G2100 from the improvement proposal information 137, and displays them in a selectable table format. The selected improvement plan is colored, for example, gray. The improvement proposal selection section G2300 includes improvement proposal number G2310, proposal type G2320, expected improvement amount (%) G2330, and expected improvement amount (yen) G2340.
The proposal type G2320 presents information regarding the name T9200 of the improvement proposal template 138.
The expected improvement amount (%) G2330 and the expected improvement amount (yen) G2340 are presented using the expected improvement amount T8080 of the improvement plan information 137 and the generated cost information 135.

The improvement proposal screen G2000 allows the flow developer/operator to quickly understand the content and effects of the proposed verification process and determine whether or not to insert the verification process. Note that the improvement proposal screen G2000 shown in FIG. 15 is an example for presenting an improvement proposal to the flow developer/operator, and other screen contents may be used as long as the UI is for making improvement proposals through verification processing.

Although one embodiment has been described above, the present invention is not limited to the above-described embodiment, and can be implemented with various modifications without departing from the spirit of the present invention.

For example, in the above embodiment, the UI generation unit 11, flow execution management unit 121, cost information collection calculation unit 122, and improvement proposal unit 123 are realized by one processing unit 10 or one computer 1. A plurality of functional units may be realized by a plurality of processing units or a plurality of computers. For example, a processing unit or a computer that realizes the cost information collection calculation unit 122 and the improvement proposal unit 123 newly proposed in this embodiment may be added to a general flow management computer having a flow execution management unit 121. It can be configured as follows. Further, the computer may be provided as a cloud service.
Further, the plurality of tables stored in the storage unit 20 may also be distributed and stored in the plurality of storage units.

Furthermore, in the above embodiment, the terminals 7 and 8 are connected to the computer 1 via the network 9. According to another example, the terminal 7 or 8 may have a display input device having terminal functions directly connected to the computer 1 without going through the network 9. Even in this case, the computer 1 uses the UI generation unit 11 to connect to the display input device.

According to this embodiment, the workflow can be ended before the task is executed and execution costs are incurred, and unnecessary execution costs that do not contribute to achieving the purpose of the workflow can be reduced. It can be presented to users such as developers of verification processes that are effective in reducing flow execution costs, and can support execution cost reductions that take abnormal systems into account.

1: Computer, 10: Processing unit, 20: Storage unit, 40: Input/output I/F, 50: NWI/F
11: UI generation unit 12: Functional unit, 121: Flow execution management unit, 122: Cost information collection calculation unit,
123: Improvement proposal department,
131: Flow definition information, 132: Flow instance information, 133: Flow execution history information, 134: Prior cost information, 135: Incurred cost information, 136: Learning model information, 137: Improvement plan information, 138: Improvement proposal template 7: Terminal, 8: Terminal, 9: Network

Claims

The computer that manages the workflow has a cost information collection calculation section and an improvement proposal section,
The cost information collection calculation unit calculates the execution cost of one or more tasks to be executed in the workflow and the entire workflow using cost advance information collected in advance,
The improvement proposal department is
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
Based on the task execution cost reduction amount and the overhead, calculate an expected improvement amount that is the reduction amount of the execution cost that occurs during operation of the workflow;
A workflow management system characterized by:
a user interface that provides a display of the verification process and the expected amount of improvement calculated by the improvement proposal unit;
The workflow management system according to claim 1.
The improvement proposal unit specifies information including the content and insertion position of the verification process,
having a user interface that provides display of the specified contents of the verification process and the insertion position;
The workflow management system according to claim 1.
4. The workflow management system according to claim 3, wherein the improvement proposal unit specifies a process for estimating a classification of a workflow execution result from information regarding the flow being executed, and uses the process as content of the verification process.
4. The workflow management system according to claim 3, wherein the improvement proposal unit calculates an index related to the accuracy of the verification process and uses the index to determine whether or not to present the information or to determine the presentation method.
The workflow management system according to claim 1, wherein the improvement proposal unit calculates the expected improvement amount by subtracting overhead from the task execution cost reduction amount.
3. The workflow management system according to claim 2, wherein the user interface displays a task for implementing the contents of the verification process on a terminal screen when presenting an improvement proposal.
The workflow management system according to claim 1, wherein the execution cost of the task is evaluated based on processing time and charges incurred.
The cost prior information includes a fixed cost that occurs for the number of times a task is executed, and a time-varying cost that occurs in proportion to the task execution time, in order to calculate the execution cost of tasks and workflows from flow execution history information. The workflow management system according to claim 1, comprising: .
The improvement proposal information calculated by the improvement proposal unit includes information regarding the insertion point of the verification process, improvement proposal parameters, false positive rate, task execution cost reduction amount, overhead T, and expected improvement amount, and the improvement proposal information is stored. The workflow management system according to claim 1, wherein a table is stored in the storage unit.
The computer that manages the workflow executes the cost information collection calculation step and the improvement proposal step.
The cost information collection calculation step calculates the execution cost of one or more tasks to be executed in the workflow and the entire workflow using cost advance information collected in advance,
The improvement proposal step includes:
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
Based on the task execution cost reduction amount and the overhead, calculate an expected improvement amount that is the reduction amount of the execution cost that occurs during operation of the workflow;
A workflow management method characterized by:
12. The workflow management method according to claim 11, wherein a user interface provides a display of the verification process and the expected amount of improvement calculated in the improvement proposal step.
The improvement proposal step specifies information including the content and insertion position of the verification process,
12. The workflow management method according to claim 11, wherein a user interface provides a display of the specified contents of the verification process and the insertion position.
14. The workflow management method according to claim 13, wherein the improvement proposal step specifies a process for estimating a classification of a workflow execution result from information regarding the flow being executed, and uses the process as content of the verification process.
A program that is executed by a computer that manages workflow,
Identify a verification process that verifies the success of one or more tasks based on workflow execution history information,
The verification process detects the possibility of a task abnormality occurring in the workflow before task execution, and calculates a task execution cost reduction amount by totaling the execution cost of one or more tasks that can be omitted from execution;
Calculate the overhead, which is the execution cost of the verification process,
Based on the task execution cost reduction amount and the overhead, calculate an expected improvement amount that is the reduction amount of the execution cost that occurs during operation of the workflow;
A program characterized by executing.