KR20210096875A - Method for generating simulation model and apparatus thereof - Google Patents

Abstract

Disclosed are a simulation model creation method and a device. According to the present invention, the simulation model creation method includes the steps of: obtaining log data; generating a leading and lagging table based on the log data; generating a process model based on the leading and lagging tables; generating a probabilistic model based on the process model; generating an event sequence based on the probabilistic model; predicting a required work time based on the log data; and generating the simulation model based on the probabilistic model, the event sequence, and the required work time. Accordingly, the more accurate simulation model can be generated.

Description

Method and apparatus for generating a simulation model

The present invention relates to a method and apparatus for generating a simulation model, and more particularly, to a method and apparatus for generating a simulation model based on process mining.

Process mining may be a method for collecting event logs generated by systems and devices to facilitate analysis of process models and patterns. If a sequential process model is derived based on the event log using process mining, the user can analyze the characteristics of the system or device that generated the event log based on the derived process model. Process mining can help users to identify bottlenecks such as main flow of work and delayed tasks by providing a process model to users.

On the other hand, process mining may generate a process model by deriving a precedence and trailing relation table from an event log. However, in the case of the existing technology, only the precedence and trailing relationships between past business events are identified by deriving a business process map, and the sequence of events by the probability of branching from a specific event to the next event required for business process planning and the next event after a random event occurs There may be problems that cannot be predicted until the time of occurrence.

An object of the present invention for solving the above problems is to provide a simulation model generating method and apparatus for generating a transition probability model between tasks and a sequence of possible events and predicting a schedule of a task.

According to an embodiment of the present invention for solving the above problems, a simulation model generation method includes: obtaining log data; generating a preceding and succeeding table based on the log data; based on the preceding and succeeding table generating a process model, generating a probabilistic model based on the process model, generating an event sequence based on the probabilistic model, estimating a task required time based on the log data, and the probabilistic model , generating a simulation model based on the sequence of events and the time required for the task.

Here, the log data may include one or more of a case ID (identifier) of a job, a type of the job, and a required time of the job.

Here, the step of generating the precedence table includes: identifying a process pattern based on the log data; identifying a precedence and subsequent relationship between tasks based on the process pattern; It may include creating a table.

Here, the generating of the probabilistic model includes: obtaining a transition point and a transition frequency based on the preceding and succeeding table; obtaining a transition probability based on the transition point and transition frequency; and based on the transition probability and generating the probabilistic model.

Here, the probabilistic model may be a markov chain.

Here, the generating of the event sequence includes: obtaining an order between tasks based on the probabilistic model; obtaining a probability that tasks are performed in the obtained order; and generating an event sequence based on the order and probability. It may include the step of generating.

Here, the predicting of the task required time may include predicting the task required time by a pre-trained machine learning model.

Here, the machine learning model may be Xgboost.

Here, the predicting of the time required for the job may include calculating an average of the time required for each job included in the log data, calculating a standard deviation based on the average, and based on the average and the standard deviation It may include obtaining a normal distribution function and estimating the required work time based on the normal distribution function.

Here, generating the simulation model may include predicting a schedule time of a task based on the task required time, and generating the simulation model based on the event sequence and the schedule time of the probabilistic model. can

A simulation model generating apparatus according to another embodiment of the present invention may include a processor and a memory in which one or more instructions executed by the processor are stored, and the one or more instructions obtain log data and , generating a preceding and succeeding table based on the log data, generating a process model based on the preceding and succeeding table, generating a probabilistic model based on the process model, generating an event sequence based on the probabilistic model, , predicting a task required time based on the log data, and generating a simulation model based on the probabilistic model, the event sequence, and the task required time.

Here, the log data may include at least one of a case ID (identifier) of a job, a type of the job, and a required time of the job.

Here, when generating the precedence table, the one or more commands determine a process pattern based on the log data, determine a precedence relationship between tasks based on the process pattern, and determine the precedence relationship. It may be performed to create a leading and trailing table based on it.

Here, when generating the probabilistic model, the one or more instructions are configured to: obtain a transition point and a transition frequency based on the preceding and subsequent table, obtain a transition probability based on the transition point and the transition frequency, and It may be performed to generate the probabilistic model based on probabilities.

Here, the probabilistic model may be a markov chain.

Here, when generating the sequence of events, the one or more instructions include

Obtaining an order between tasks based on the probabilistic model, obtaining a probability that tasks are performed in the obtained order, and generating an event sequence based on the order and probability.

Here, when predicting the task required time, the one or more instructions may be performed to predict the task required time by a pre-trained machine learning model.

Here, the machine learning model may be Xgboost.

Here, when estimating the task required time, the one or more commands calculate an average of the required time for each task included in the log data, calculate a standard deviation based on the average, and calculate the average and the standard deviation It may be performed to obtain a normal distribution function based on , and to predict the operation time required based on the normal distribution function.

Here, when generating the simulation model, the one or more instructions predict a schedule time of a task based on the task required time, and generate the simulation model based on the event sequence and the schedule time of the probabilistic model can be performed to

According to the present invention, a more accurate simulation model is generated by generating an event sequence based on a Markov Chain probability model, predicting the schedule of a task, and generating a simulation model based on the event sequence and the predicted schedule. can do.

1 is a block diagram of an apparatus for generating a simulation model according to an embodiment of the present invention.
2 is a block diagram of an apparatus for generating a simulation model according to another embodiment of the present invention.
3 is a conceptual diagram of a process model according to an embodiment of the present invention.
4 is a conceptual diagram of a probabilistic model according to an embodiment of the present invention.
5 to 13 are conceptual diagrams for explaining application of a machine learning model according to an embodiment of the present invention.
14 is a flowchart of a simulation generating method according to an embodiment of the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. and/or includes a combination of a plurality of related listed items or any of a plurality of related listed items.

When a component is referred to as being “connected” or “connected” to another component, it is understood that the other component may be directly connected or connected to the other component, but other components may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that no other element is present in the middle.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. In describing the present invention, in order to facilitate the overall understanding, the same reference numerals are used for the same components in the drawings, and duplicate descriptions of the same components are omitted.

1 is a block diagram of an apparatus for generating a simulation model according to an embodiment of the present invention.

Referring to FIG. 1 , the simulation model generating apparatus 100 may include at least one processor 110 , a memory 120 , and a transceiver 130 connected to a network to perform communication. In addition, the simulation model generating apparatus 100 may further include an input interface device 140 , an output interface device 150 , a storage device 160 , and the like. Each of the components included in the simulation model generating apparatus 100 may be connected by a bus 170 to communicate with each other.

However, each of the components included in the simulation model generating apparatus 100 may be connected through an individual interface or an individual bus centered on the processor 110 instead of the common bus 170 . For example, the processor 110 may be connected to at least one of the memory 120 , the transceiver 130 , the input interface device 140 , the output interface device 150 , and the storage device 160 through a dedicated interface. .

The processor 110 may execute a program command stored in at least one of the memory 120 and the storage device 160 . The processor 110 may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. In addition, the processor 110 may include a machine learning algorithm. The machine learning algorithm may be XGBoost. Each of the memory 120 and the storage device 160 may be configured as at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory 120 may be configured as at least one of a read only memory (ROM) and a random access memory (RAM).

The memory 120 and the storage device 160 may include log data. The log data may include an event log in which performance results of events belonging to the process are recorded while the work process is performed. The event log may include a plurality of details that are information on the execution of individual events belonging to a process.

2 is a block diagram of an apparatus for generating a simulation model according to another embodiment of the present invention.

The simulation model generating apparatus 200 of FIG. 2 may be configured the same as or similar to the simulation model generating apparatus 100 of FIG. 1 .

Referring to FIG. 2 , an apparatus 200 for generating a simulation model according to an embodiment of the present invention includes a log data acquisition unit 210 , a processor model generation unit 220 , a probability model generation unit 230 , and a schedule prediction unit ( 240) may be included.

The log data acquisition unit 210 may acquire log data from a memory (eg, the memory 120 of FIG. 1 ) and a storage device (eg, the storage device 160 of FIG. 1 ). The log data may include an event log. The event log can be represented as shown in Table 1 below.

Referring to Table 1, the event log may include a case ID (CaseID) indicating an object of a process, an activity (Activity) constituting a process, and an operation time required. In addition, although not included in Table 1, the event log may further include information on a task performer who is a subject who performed the task, a work environment, and a task goal.

For example, when the simulation model generating apparatus 200 generates a simulation model for the working time of the slab hot rolling process, the operation may include heating of the slab, rough rolling, finishing rolling, winding step, and the like. The event log may further include the weight of the slab, the carbon content of the steel, the thickness of the steel, the length of the steel, the target temperature of the heating furnace, and the like.

When the simulation model generating device 200 generates a simulation model for the time an outpatient stays in the hospital, the activity means reception, treatment, examination, receipt, etc., and includes the time required for the operation and the schedule can do. The event log may further include the age of the outpatient, whether the outpatient is the first or revisit, the treatment subject, and the type of examination. The log data obtaining unit 210 may transmit the log data to the process model generating unit 220 and the schedule predicting unit 240 .

The process model generator 220 may receive log data from the log data acquirer 210 . The process model generator 220 may acquire a process pattern from log data. The process pattern may be a sequence of operations in each case ID.

For example, the process model generator 220 may acquire a process pattern based on the event log of Table 1 . The process model generator 220 may obtain <A, B, C, E, G> as the process pattern of case ID 1 and <A, B, D, E, G> as the process pattern of case ID 2 can be obtained as , and <A, F, G> can be obtained as a process pattern of case ID 3.

Meanwhile, the process pattern of case ID 1 may be performed 20 times, the process pattern of case ID 2 may be performed 10 times, and the process pattern of case ID 3 may be performed 5 times. The process model generator 220 may generate a process pattern table including a process pattern and the number of times the process pattern is performed. The process pattern table generated by the process model generator 220 based on Table 1 may be shown in Table 2 below.

The process model generator 220 may determine a precedence and subsequent relationship between tasks based on the process pattern. Here, the predecessor relationship may include a predecessor task, a successor task, and the number of transitions from the predecessor task to the successor task.

For example, the process model generator 220 may identify a precedence and subsequent relationship between tasks based on the process pattern table of Table 2 . If a predecessor task does not exist, the successor task can transition to A. If the predecessor is A, the successor can transition to either B or F. If the predecessor is B, the successor can transition to either C or D. If the predecessor is C, the successor can transition to E. If the predecessor is D, the successor can transition to E. If the predecessor is E, the successor can transition to G. If the predecessor is F, the successor can transition to G.

The number of transitions to the successor task A without a predecessor task may be 45 times. The number of transitions from the predecessor task A to the successor task B may be 30 times. The number of transitions from the predecessor task B to the successor task C may be 20. The number of transitions from the predecessor task B to the successor task D may be 10.

The number of transitions from the predecessor task C to the successor task E may be 20 times. The number of transitions from the predecessor task D to the successor task E may be 10. The number of transitions from the predecessor task E to the successor task G may be 30. The number of transitions from the predecessor task A to the successor task F may be 5 times. The number of transitions from the predecessor task F to the successor task G may be 5 times. Also, when the predecessor task is G and there is no successor task, the number of transitions may be 45.

The process model generator 220 may generate a preceding and succeeding table including the preceding task, the succeeding task, and the number of transitions. The preceding and subsequent tables generated by the process model generator 220 based on Table 2 may be as shown in Table 3 below.

The process model generator 220 may generate a process model based on the preceding and following tables. For example, the process model generator 220 may generate a process model based on Table 3 . The process model can be:

3 is a conceptual diagram of a process model according to an embodiment of the present invention.

Referring to FIG. 3 , the process model generator (eg, the process model generator 220 of FIG. 2 ) may generate a process model based on a preceding task and a subsequent task. The process model generator may generate a process model to determine a flow between tasks.

That is, when the process model generator creates a process model based on Table 3 (see FIG. 2 ), if the predecessor task is A, the successor task may be B or F, and if the predecessor task is B, the successor task is C or D can be Also, if the predecessor is C or D, the successor may be E, and if the predecessor is E, the successor may be G. Also, if the predecessor is F, the successor may be G. Accordingly, the process model generator may generate a process model in the form of a Markov chain as shown in FIG. 3 based on this.

Referring back to FIG. 2 , the process model generator 220 may transmit the preceding and succeeding table and the process model to the probabilistic model generator 230 . The probabilistic model generator 230 may receive the preceding and succeeding table and the process model from the process model generator 220 .

The simulation model generator 230 may obtain a transition point and a transition frequency based on the preceding and following tables. The transition point may be a predecessor task having a plurality of successor tasks. The transition frequency may be the number of transitions from a predecessor task to a successor task.

For example, the probabilistic model generator 230 may obtain a transition point and a transition frequency based on the preceding and following tables of Table 3 . The probabilistic model generator 230 may acquire A, which may be a successor task as B or F, as a transition point, and may acquire B, where a successor task may become C or D, as a transition point.

The probabilistic model generator 230 may obtain the frequency of transition from task A to task B as 30 and the frequency of transition from task A to task F as 5, respectively. Also, the probabilistic model generator 230 may obtain the transition frequency from the task B to the task C as 20 and the transition frequency from the task B to the task D as 10, respectively.

The probabilistic model generator 230 may calculate a transition probability based on the transition point and the transition frequency. The probabilistic model generator 230 may generate a probabilistic model based on the transition probability and the process model. The probabilistic model may be a Markov chain. The probabilistic model may be as follows.

4 is a conceptual diagram of a probabilistic model according to an embodiment of the present invention.

Referring to FIG. 4 , the probabilistic model generator (eg, the probabilistic model generator 230 of FIG. 2 ) may calculate a transition probability based on the transition frequency. The transition probability may be a probability of transitioning from a predecessor task to a successor task.

For example, the probabilistic model generator may obtain a transition probability based on a transition frequency obtained based on Table 3 (refer to FIG. 2 ). The probability of transition from task A to task B may be 6/7. The probability of transition from task A to task F may be 1/7. The probability of transition from task B to task C may be 2/3, and the probability of transition from task B to task D may be 1/3. Further, the transition probability from task C to task E, the transition probability from task D to task E, the transition probability from task E to task G, and the transition probability from task F to task G may be one. Accordingly, the probabilistic model generator may generate a probabilistic model that is a Markov chain including a transition probability as shown in FIG. 4 .

Referring back to FIG. 2 , the probabilistic model generator 230 may generate an event sequence based on the probabilistic model. The sequence of events may include an order between tasks and a probability that tasks will be performed in that order.

For example, the probabilistic model generator 230 may generate an event sequence based on the probabilistic model of FIG. 4 . The probabilistic model generator 230 may obtain <A, B, C, E, G>, <A, B, D, E, G>, and <A, F, G> in an order between tasks. The probabilistic model generator 230 may obtain 4/7 as a probability that the operation is performed with <A, B, C, E, G>, and the operation is performed with <A, B, D, E, G> You can get 2/7 with the probability of being done, and 1/7 with the probability that the task will be performed with <A, F, G>.

The probabilistic model generator 230 may transmit the probabilistic model and the event sequence to the schedule predictor 240 . The schedule predictor 240 may receive the event sequence from the probabilistic model generator 230 . Also, the schedule predictor 240 may receive log data from the log data obtainer 210 .

The schedule prediction unit 240 may include a machine learning model. The machine learning model may be XGBoost, an artificial neural network, or the like. When the event attribute independent variable exists in the log data, the schedule predictor 240 may predict the task required time based on an algorithm included in the machine learning model. The event attribute independent variable may be a variable that can affect the operation time required. For example, when the schedule prediction unit 240 predicts the slab working time, the weight of the slab, the steel carbon component, the slab thickness, the slab length, and the target temperature for the slab work may be event attribute independent variables. When the schedule prediction unit 240 predicts the time the outpatient stays in the hospital, whether the outpatient is the first visit or the second visit, the age of the outpatient, the treatment subject, the type of examination, etc. may be event attribute independent variables. In addition, machine learning algorithms perform not only the task duration, but also the waiting time to perform the task and when tasks are performed in sequence (for example, <A, B, C, E, G>, <A, B, D, E, G> and <A, F, G> (if operations are performed in the order)) latency and duration can also be predicted.

Meanwhile, the machine learning model may be trained in advance. The machine learning model may be pre-trained based on a plurality of log data. For example, a machine learning model can learn based on log data that includes the slab's weight, steel carbon composition, slab thickness, slab length, and target temperature and slab working time for slab working. This can be described in detail as follows.

5 to 13 are conceptual diagrams for explaining the application of a machine learning model according to an embodiment of the present invention.

5 to 13 , the machine learning model may input log data including the weight of the slab, the steel carbon component, the slab thickness, the slab length, the target temperature for the slab operation, and the slab operation time. A histogram of the slab working time as shown in FIG. 5 can be generated based on log data through a machine learning algorithm. Referring to Figure 5, the slab working time may be distributed between 200 and 550.

6 to 10 , the machine learning model may compare the log data by separating them based on the slab working time. For example, a machine learning model can split log data into 50 slab working time intervals. A machine learning model can compare log data to determine which features affect slab working time. The features may include the weight of the slab, the steel carbon composition, the slab thickness, the slab length and the target temperature for the slab operation.

The machine learning model is divided into 200-250 sections, 250-300 sections, 300-350 sections, 350-400 sections, and 400-450 sections based on the slab working time. can create The weight of the slab, the iron carbon content, and the slab thickness can have the most influence on the slab working time.

The machine learning model is a graph showing the relationship between the weight of the slab and the slab working time, and a graph showing the relationship between the slab thickness and the slab working time, as shown in FIGS. 11 to 13. A graph showing the relationship between the material of the slab and the slab working time can create The graph can be a Jointplot.

Through this learning process, the machine learning model can predict the slab working time according to the weight of the slab, the steel carbon content, the slab thickness, the slab length, and the target temperature for working the slab.

Referring back to FIG. 2, for example, when the schedule prediction unit 240 predicts the slab working time, the machine learning model includes the weight of the slab, the steel carbon component, the slab thickness, the slab length, and the target temperature for the slab operation. etc. can be input as independent variables, and the slab working time can be predicted based on this.

를 연산할 수 있다. 예를 들어, 작업 A의

는

일 수 있다. 스케줄 예측부(240)는 작업 A의 표준 편차인

를 연산할 수 있다. 스케줄 예측부(240)는

및

를 기초로 다음 수학식 1과 같은 정규 분포 함수를 획득할 수 있다.When the event attribute independent variable does not exist in the log data, the schedule predictor 240 may predict the required time of each task based on the task time of the event log. The schedule prediction unit 240 is an average of the work required time of the same task.

can be calculated. For example, in task A

Is

can be The schedule prediction unit 240 is the standard deviation of task A

can be calculated. The schedule prediction unit 240 is

and

It is possible to obtain a normal distribution function as in Equation 1 based on

The schedule prediction unit 240 may predict the required time for the task A based on the normal distribution function. The schedule predictor may predict the task required time of the remaining tasks B, C, D, E, F, and G by repeating this process.

The schedule predictor 240 may predict the job occurrence time based on log data. When the first task execution time is determined, the schedule prediction unit 240 may predict the number and occurrence time of subsequent tasks. The schedule predictor 240 may predict the task occurrence time through Equation 2 below.

In Equation 2, P _ij may be the transition probability from the predecessor task i to the successor task j, N _ij may be the number of transitions from the predecessor task i to the successor task j, and n may be the number of all tasks present in the log data. there is.

The schedule prediction unit 240 may predict the start time of the subsequent task based on the task occurrence time and the task required time. The schedule predictor 240 may obtain the schedule time of the subsequent job based on Equation 3 below.

In Equation 3, S _j may be the start time of the predecessor task, W _i may be the execution time of the predecessor task i, and T _ij may be the transition time from the predecessor task i to the successor task j. Meanwhile, the first task execution interval may follow the event initiation distribution observed in process mining. S _j may be obtained by Equation 2, W _i may be obtained by a machine learning algorithm or a normal distribution function, and T _ij may be obtained from log data. The schedule predictor 240 may generate a simulation model based on a schedule time, a probabilistic model, and an event sequence.

14 is a flowchart of a method for generating a simulation model according to an embodiment of the present invention.

Referring to FIG. 14 , the simulation model generating apparatus (eg, the simulation model generating apparatus 200 of FIG. 2 ) may obtain log data ( S1410 ). The log data acquisition unit (eg, the log data acquisition unit 210 of FIG. 2 ) of the simulation model generating apparatus includes a memory (eg, the memory 120 of FIG. 1 ) and a storage device (eg, FIG. 1 ) Log data may be obtained from the storage device 160 of the The log data may include an event log (eg, Table 1 of FIG. 2 ). The event log may include a case ID, the type of operation, and the duration of the operation.

The log data acquisition unit may transmit the log data to the process model generation unit (eg, the process model generation unit 220 of FIG. 2 ) and the schedule prediction unit (eg, the schedule prediction unit 240 of FIG. 2 ). .

The simulation model generating apparatus may generate a process model (S1420). The process model generating unit of the simulation model generating apparatus may receive log data from the log data obtaining unit. The process model generator may acquire a process pattern based on log data. The process model generator may generate a process pattern table based on the process pattern (eg, Table 2 of FIG. 2 ).

The process model generating unit may identify a precedence and trailing relationship between tasks based on the process pattern. The process model generator may generate a precedence and succession table (eg, Table 3 of FIG. 2 ) based on the precedence and subsequent relationships. The process model generator may generate a process model (eg, the process model of FIG. 3 ) based on the preceding and succeeding tables. The process model generator may transmit the preceding and subsequent tables and the process model to the probabilistic model generator (eg, the probabilistic model generator 230 of FIG. 2 ).

The simulation model generating apparatus may generate a probabilistic model and an event sequence (S1430). The probabilistic model generating unit of the simulation model generating apparatus may receive the preceding and succeeding tables and the process model from the process model generating unit. The probabilistic model generator may calculate a transition point and a transition frequency between tasks based on the preceding and following tables. The probabilistic model generator may calculate a transition probability between transition operations based on a transition point and a transition frequency. The probabilistic model generator may generate a probabilistic model (eg, the probabilistic model of FIG. 4 ) based on the transition probability and the process model. The probabilistic model generator may generate an event sequence based on the probabilistic model. The probabilistic model generator may obtain an order between tasks based on the probabilistic model, and may generate an event sequence in such a way as to obtain a probability that tasks will be performed in the corresponding order. The probabilistic model generator may transmit the probabilistic model and the event sequence to the schedule predictor (eg, the schedule predictor 240 of FIG. 2 ).

The simulation model generating apparatus may check whether the log data includes the independent variable (S1440). The schedule predicting unit of the simulation model generating apparatus may receive log data from the log data obtaining unit. Also, the schedule predictor may receive the probabilistic model and the event sequence from the probabilistic model generator. The schedule predictor may check whether an independent variable is included in the log data.

When the log data includes the independent variable (Yes in S1440), the schedule predictor may predict the required work time based on the machine learning algorithm (S1450). The schedule predictor may predict the task required based on a pre-trained machine learning algorithm. For example, log data may be input to the machine learning model, and the operation time may be predicted based on independent variables included in the log data.

On the other hand, when the log data does not include the independent variable (No in S1440), the schedule predictor may predict the required work time based on the work time included in the log data (S1460). The schedule predictor may calculate an average of the work required time for each task based on the log data, and may calculate the task required time and a standard deviation of the corresponding task based on this. The schedule predictor may obtain a normal distribution function based on the average and standard deviation of the work required time. The schedule predictor may predict the required work time based on the normal distribution function.

The simulation model generator may generate a simulation model (S1470). The schedule prediction unit of the simulation model generator may predict the schedule time of the job based on the predicted job required time. The schedule predictor may generate a simulation model based on a schedule time, a probabilistic model, and an event sequence, and performs a simulation using the generated simulation model to output a schedule prediction result.

The methods according to the present invention may be implemented in the form of program instructions that can be executed by various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the computer-readable medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software.

Examples of computer-readable media include hardware devices specially configured to store and carry out program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as at least one software module to perform the operations of the present invention, and vice versa.

Although it has been described with reference to the above embodiments, it will be understood by those skilled in the art that various modifications and changes can be made to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. will be able

Claims (20)

obtaining log data;
generating a leading and trailing table based on the log data;
generating a process model based on the preceding and succeeding tables;
generating a probabilistic model based on the process model;
generating an event sequence based on the probabilistic model;
predicting a required work time based on the log data; and
and generating a simulation model based on the probabilistic model, the event sequence, and the time required for the task. The method according to claim 1,
The log data is
A method for generating a simulation model, comprising at least one of a case identifier (ID) of a task, a type of the task, and a duration of the task. The method according to claim 1,
The step of creating the preceding and trailing table comprises:
identifying a process pattern based on the log data;
identifying a precedence and trailing relationship between tasks based on the process pattern; and
and generating a precedence table based on the precedence relation. The method according to claim 1,
The step of generating the probabilistic model comprises:
obtaining a transition point and a transition frequency based on the preceding and subsequent tables;
obtaining a transition probability based on the transition point and the transition frequency; and
and generating the probabilistic model based on the transition probability. The method according to claim 1,
wherein the probabilistic model is a markov chain. The method according to claim 1,
The step of generating the event sequence comprises:
obtaining an inter-task order based on the probabilistic model;
obtaining a probability that tasks are performed in the obtained order; and
generating a sequence of events based on the order and probability. The method according to claim 1,
The step of estimating the time required for the operation is
A method of generating a simulation model, comprising the step of estimating the task required time by means of a pre-trained machine learning model. 8. The method of claim 7,
wherein the machine learning model is XGboost. The method according to claim 1,
The step of estimating the time required for the operation is
calculating an average of the time required for each task included in the log data;
calculating a standard deviation based on the mean;
obtaining a normal distribution function based on the mean and the standard deviation; and
Comprising the step of estimating the task required based on the normal distribution function, simulation model generation method. The method according to claim 1,
The step of generating the simulation model comprises:
predicting a schedule time of a job based on the required time for the job; and
and generating the simulation model based on the sequence of events and the schedule time of the probabilistic model. processor and
and a memory in which one or more instructions executed by the processor are stored,
the one or more instructions,
obtain log data;
create a leading and trailing table based on the log data;
generate a process model based on the preceding and subsequent tables;
generate a probabilistic model based on the process model;
generate an event sequence based on the probabilistic model;
predicting a required work time based on the log data; And,
and generating a simulation model based on the probabilistic model, the sequence of events, and the time required for the task. 12. The method of claim 11,
The log data is
A simulation model generating apparatus comprising at least one of a case ID (identifier) of a job, a type of the job, and a required time of the job. 12. The method of claim 11,
When creating the preceding and trailing tables,
the one or more instructions,
identify a process pattern based on the log data;
identifying a precedence and trailing relationship between tasks based on the process pattern; And,
and generating a precedence and trailing table based on the precedence and trailing relationship. 12. The method of claim 11,
When generating the probabilistic model,
the one or more instructions,
obtain a transition point and a transition frequency based on the preceding and subsequent tables;
obtain a transition probability based on the transition point and the transition frequency; And,
The simulation model generating apparatus is performed to generate the probabilistic model based on the transition probability. 11. The method of claim 10,
The probabilistic model is a markov chain (markov chain), simulation model generating apparatus. 12. The method of claim 11,
When generating the event sequence,
the one or more instructions,
obtaining an inter-task order based on the probabilistic model;
obtain a probability that tasks will be performed in the obtained order; And,
and generating a sequence of events based on the order and probability. 12. The method of claim 11,
When estimating the time required for the above operation,
the one or more instructions,
A simulation model generating apparatus that is performed to predict the required time of the task by means of a pre-trained machine learning model. 18. The method of claim 17,
The machine learning model is Xgboost, a simulation model generating device. 12. The method of claim 11,
When estimating the time required for the above operation,
the one or more instructions,
calculating an average of the time required for each task included in the log data;
calculate a standard deviation based on the mean;
obtain a normal distribution function based on the mean and the standard deviation; And,
A simulation model generating apparatus that is performed to predict a task required time based on the normal distribution function. 12. The method of claim 11,
When generating the simulation model,
the one or more instructions,
predicting a schedule time of a job based on the required time for the job; And,
and generating the simulation model based on the sequence of events and the schedule time of the probabilistic model.

Images