KR102238424B1 - System Modeling Method by Machine Learning using Big data - Google Patents

Abstract

The present invention relates to a system modeling method in which big data machine learning is embedded, in which a hypothetical model is established, and parameter values verified using machine learning are calculated using big data obtained from an actual system, and then applied to the hypothetical model It has its purpose.
An object of the present invention is to provide a system modeling method for completing a simulation model for a target system through machine learning on big data acquired through operation and observation of the target system in a hypothetical model defined through knowledge acquisition of a target system. In this regard, the process of defining a hypothetical model for the target system by identifying obtainable information related to the target system; Learning information necessary for a hypothetical model by using a machine learning algorithm with big data actually acquired from the target system in a functional block provided by a hypothetical model; And a process of completing a simulation model for a target system by applying information learned and verified through a machine learning process to a hypothetical model.

Description

System Modeling Method by Machine Learning using Big data}

The present invention relates to a technology for modeling and simulating a physical complex system (target system), and more specifically, machine learning to fit the model of big data acquired from an actual system, and verified The present invention relates to a system modeling method with built-in big data machine learning that calculates parameter values and applies them to a hypothetical model.

delete

delete

delete

delete

delete

In general, in order to perform motion/performance analysis or prediction for a real-world system, an abstracted model is created and executed to measure/observe measures of interest aspects such as motion/performance. In order to ensure the reliability of the analysis/prediction results of the target system through such a model, how well (accurately) the system is modeled becomes an important issue.

As a method of modeling, there is a method of creating an abstraction model using knowledge of the physical laws or rules of operation included in the target system. This is modeling and simulation that can express the causal relationship between the controlled input and the corresponding output as a model. It is a (M&S) based method. This method has the limitation that detailed information about the system to be modeled must be available. In addition, when a model is built in a modeling and simulation (M&S)-based method, it is essential to verify the model how well it reflects the actual system in order to ensure the validity of the model. When it is difficult to obtain data from a real-world system, since the validity of the model cannot be verified, there may be a problem that the reliability of the analysis/prediction results based on the model cannot be guaranteed.

As another modeling method for predicting and analyzing a system, there is a data modeling method in which the rules/patterns/functions contained in the corresponding system are derived by analyzing a large number of data acquired through operation and observation of the target system. The machine learning-based model, which is a representative method of data modeling, is a method that represents the correlation between one set of data and another data set. In the era of big data, more effective machine learning has become possible by utilizing a large number of data. . However, the data model built through machine learning can be predicted under the premise that the system will operate without any change in the future, but if the composition or operating laws of the system change, the future cannot be predicted with the previously learned model. There are limitations.

The term'big data' is widely spread as a means of predicting a diversified society. Big data refers to a set of data that exceeds the ability of common software tools to collect, manage, and process within acceptable elapsed time. Because such large amounts of data provide more insight than existing limited data, it is receiving a lot of interest in research in various fields such as science, engineering, defense, business, medicine, and politics. For this reason, modeling using big data is becoming an essential and important issue in the era of big data.

Modeling using big data can be defined as data modeling that focuses on showing the correlation of data. These approaches are classified into two types, data mining and machine learning, and are being studied. 1 shows a method of performing data modeling in two ways.

Data mining is one of the useful methods of data modeling as shown on the left of FIG. 1. By using data mining technology, users who want to predict system outcomes can not only analyze data patterns or attributes in one dimension, but also identify distribution functions of data in data patterns.

You can then use a method such as Goodness of Fitness (GOF) to verify the distribution function with real-world data, and then obtain a "random number generation" model. The finally obtained model can be used in the process of predicting future data patterns.

Meanwhile, machine learning can be another means of data modeling as shown in the right side of FIG. 1. Using machine learning algorithms such as artificial neural networks (ANNs) and genetic algorithms (GE), users can map associations between one set of data d1 and another set of data dn.

Similar to the data mining process, a data model can be obtained by passing the validation process of the actual map data using a general figure of merit such as RMSE (Root-Mean-Square Error). The future value of the data set "dn" can then be predicted using the given data set "d1".

This data modeling proceeds as a series of processes such as acquisition, modeling, verification and prediction. Data modeling has been widely used in various fields such as science, engineering, economy, and industry to predict the future behavior of a target system, and some researchers have argued that given enough information, the correlation is sufficient to make strong and informative predictions. .

Contrary to these expectations, however, data modeling doesn't mean that it will always be a powerful modeling approach. This method has several limitations. One of the typical limitations is that it can account for correlations between data, rather than representing a causal relationship between control inputs and their outputs.

The data model cannot cope with unexpected and changing situations in the system. That is, if the components or structure/behavior of the system are changed after learning the model, accurate prediction using the data model is impossible.

Another limitation is the inability to cope with unexpected events. In real systems, unexpected events can occur due to the complexity and uncertainty of the system. These are generally not included in the data set we can acquire, and when such an event occurs, a data model based on the original data set cannot accurately predict unexpected events.

In addition, similarly, there is a limitation that the prediction result is affected by the amount of data acquired from the target system.

Systems science-based simulation modeling is required to overcome these limitations of data modeling. Simulation modeling can be defined as a theory-based modeling method commonly used in the simulation field, and physical or operational laws existing in the target system are used to build the model.

This makes it possible to clearly represent the causal relationship between the control input and the corresponding output set, unlike data modeling. Nevertheless, simulation modeling alone cannot be a perfect solution for modeling complex systems in the era of big data.

For example, when it is difficult to obtain sufficient knowledge about the system, it is impossible to completely build a modeling and simulation (M&S) model that satisfies the objective. This is because the simulation modeling approach is based on prior knowledge of the target system, and its completion depends on how much you can understand about the system. Extensive physical and operational knowledge of the target system is required for accurate simulation modeling.

In addition, after the simulation model is created, it is necessary to check the validity of the model through model verification. If there is no data for verification in the actual system or it is difficult to obtain, it is difficult to check the validity of the model.

As such, there are clearly limitations to the two modeling methods.

2 illustrates this limitation through a simple example. Assuming that a system having x1, x2, x3, x4, and y as inputs and outputs shown at the top of FIG. 2 is modeled, the case of using the simulation modeling method is as shown in the right diagram. The system model can be constructed through mathematical knowledge of the target system, but it becomes effective when the output value output through the model is verified through the data acquired from the actual system.

On the contrary, in the case of machine learning based on the data on x1, x2, x3, x4, y obtained by operating the system as shown in the diagram on the left, a high-accuracy model that outputs y when x1, x2, x3, x4 values are input is used. It can be obtained, but if the system operation law at the top of Fig. 2 is changed (for example, when the system operation law ("X", multiply) at the upper right of Fig. 2 is changed by division), the previously acquired big data There is a limitation in that the machine learning model derived through the system cannot properly predict the output y of the system under study whose system operation laws have changed.

US Patent Publication No. 2017/0286572 "Digital twin of twinned physical system" (October 5, 2017)

In order to overcome the problems of the prior art described above, there is a need for a mutually collaborative method that can overcome the limitations of each approach by complementarily utilizing the advantages of the two modeling methods to enable more robust analysis/prediction support. Do. Therefore, in order to improve the problems of the prior art, the present invention establishes a hypothetical model, calculates parameter values verified using machine learning, and calculates the big data obtained from the actual system. Its purpose is to provide a system modeling method with built-in big data machine learning that can be applied to a hypothetical model (Gray Box).

The system modeling process with built-in big data machine learning to achieve the object of the present invention is through machine learning on big data acquired through operation and observation of the target system in a hypothetical model defined through knowledge acquisition of the target system. A system modeling method for completing a simulation model for a target system, comprising: a first step of defining a hypothetical model for a target system by identifying obtainable information related to the target system; A second process of learning information necessary for a hypothetical model through a machine learning algorithm on the big data actually obtained from the target system according to the functional blocks provided by the hypothetical model; And a third process of completing a simulation model for a target system by applying the information learned and verified through the machine learning process to a hypothetical model.

Here, the second process is characterized in that the plurality of functional blocks inside the hypothetical model perform machine learning for each functional block to calculate a verified parameter value, and provide it as information necessary for the hypothetical model.

The parameter value includes a variable value, a probability, a function or a graph, which is characterized in that it is placed in the function block.

In addition, as one form of the hypothetical model, a cellular automata model is used, and the state transition function of a cell required for the hypothetical model is learned using machine learning using the big data. It is characterized in that it is provided in a cell automata model.

In the system modeling method with built-in big data machine learning according to the present invention, the model itself becomes a verified model by embedding the contents of machine learning using real big data acquired through operation and observation of the target system. Can be used to overcome limitations that may be encountered when analyzing or predicting the system of interest.

In other words, by embedding machine learning into the system model, it reduces the degree of prior knowledge required for the target system and achieves the effect of model verification, while analyzing the system behavior according to system structure/rule changes that could not be done with machine learning alone. Prediction becomes possible.

In addition, the present invention can be easily applied to all systems based on the terrain grid, such as forest fires, traffic, and diseases.

In addition, it has the following advantages because it learns the state transition function of the cell automata model through big data and machine learning. First, as a feature of space variant, the state transition function is not collectively expressed for all cells, but can be expressed differently for individual cells. For example, it is possible to reflect the features of the terrain that vary depending on the location for each cell.

The second is a feature of a time variant, and the state transition function can reflect a feature that may vary depending on the time of day, such as morning or afternoon.

Third, it is important to reflect the stochastic feature through big data and machine learning because the driver's behavior can appear probabilistically rather than deterministic in traffic simulations, for example. .

In addition, the present invention may reflect the nonlinear characteristics of the system. In other words, since many systems in the real world have nonlinear characteristics in which the superposition between inputs and outputs is not established, it is a great advantage that the state transition function can take into account nonlinear characteristics. Finally, there is an effect that additional information such as weather conditions can be expressed through big data obtained from the actual system.

1 is a diagram for explaining a process of performing data modeling in two ways, data mining and machine learning according to the prior art,
2 is a diagram for explaining the limitations of the two modeling methods of FIG. 1,
3 is a modeling configuration diagram through application of big data machine learning to a hypothetical model for implementing the present invention,
4 is an overall flowchart of a system modeling process in which big data machine learning is embedded according to an embodiment of the present invention.
5 is a configuration diagram for explaining a specific embodiment of FIG. 3,
6 is a diagram for explaining the definition of a cell automata applied to the hypothetical model in FIG. 5 and a method of updating the state;
7 is a diagram for explaining a state transition function of each cell in the cell automata of FIG. 6.

The features and effects of the present invention described above will become more apparent through the following detailed description in connection with the accompanying drawings, and accordingly, those of ordinary skill in the technical field to which the present invention pertains can easily implement the technical idea of the present invention. I will be able to.

In the present invention, various modifications may be made and various forms may be applied, and specific embodiments will be exemplified and described in detail in the text.

However, this is not intended to limit the present invention to a specific form of presentation, it should be understood to include all changes, equivalents, and substitutes included in the spirit and scope of the present invention.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention.

The present invention is a big data machine that calculates validated parameter values and applies it to a hypothetical model (Gray Box) because machine learning is performed to fit the model of big data acquired from an actual system. It relates to a system modeling method with built-in learning.

The configuration and specific operation according to a preferred embodiment of the present invention will be described in detail with reference to FIGS. 3 to 6.

First, the present invention is a cooperative modeling method in which a machine learning method is applied to modeling and simulation-based modeling, and the operation required for the specification of a hypothetical model by using the machine learning method using the big data actually obtained from the target system of the present invention. /Predict function functions, parameters, etc.

3 and 4 are a system configuration diagram and a flow chart of a modeling process in which big data machine learning is embedded according to an embodiment of the present invention. Referring to this, a user according to the purpose of modeling and simulation of the target system 100 for modeling The modeling requirements of should be analyzed. (S110)

Since modeling is not a complete representation of the entire target system 100, but is a process of abstracting the system for a predetermined purpose, modeling a system suitable for the purpose is important.

For system modeling, first, a gray box model 110 is defined for the system by grasping the domain knowledge/experience and theories that can be acquired related to the target system. (S130)

Since the hypothetical model 110 itself cannot be simulated, a process of constructing a complete model by securing information such as operation functions and parameters required for model completion is required.

Information necessary for model completion is obtained by actually operating/observing the modeling target system 100 to obtain the big data 120, and information necessary for the hypothetical model 110 through machine learning on the acquired big data 120. Acquire. (S150)

That is, information necessary for the hypothetical model 110 may be learned from the big data 120 acquired through the operation and observation of an actual system using a machine learning algorithm such as an artificial neural network.

The white box model 130 for the target system is completed by applying the learned and verified information to the hypothetical model based on real data. (S170)

Finally, the operation/performance analysis and prediction of the target system of interest is performed through the simulation of the completed system model 130. (S190)

FIG. 5 is a diagram for explaining FIG. 3 as a specific embodiment. First, a hypothetical model 110 of the target system 100 is established through knowledge acquisition of the target system 100. The structure of the target system 100 Although a characteristic feature can be obtained with the hypothetical model 110, coefficients corresponding to "m, n, a, b, c, d, e, k", that is, actual values are not known.

Therefore, a plurality of functional blocks inside the hypothetical model 110 calculates a verified parameter value by performing machine learning using big data for each functional block, and the calculated parameter value is added to the hypothetical model 110. Provide necessary information.

More specifically, the coefficients can be learned through big data 120 (x1, x2, x3, x4, g1, g2, y) acquired through actual operation and observation of the target system 100. Machine learning is performed for each functional block through a learning method such as a neural network. That is, g() uses "g1, g2, y", g1() uses "x1, x2, g1", and g2() can be learned using "x3, x4, g2", respectively. have.

Here, the parameter values include not only functions, but also variable values, probabilities, or graphs, which are included in each functional block.

Through learning in this way, the verified values of the parameters (m, n, a, b, c, d, e, k) are calculated, and when applied to the hypothetical model 110, the system model 130 is completed, The operation/performance analysis and prediction of the target system 100 is performed through the simulation of the completed system model 130.

6 is a diagram for explaining the definition of a cell automata applied to the hypothetical model in FIG. 5 and a method of updating the state. As an embodiment of the hypothetical model 110, the hypothetical model 110 is a cell automata. (Cellular Automata) is made through the model 111, and the state transition function of a cell necessary for a hypothetical model is learned through machine learning using the big data 120, and this is a cell automata model 111 Will be provided to.

In more detail, the state transition function of the cell required for the hypothetical model 110 using machine learning such as artificial neural network modeling using the big data 120 collected by the target system 100 Function) is learned, and it is provided to a hypothetical model, the cell automata model 111.

That is, when the learned state transition functions and the cell automata model 111 are combined, the system model 130 (white box) is completed. In this mutually collaborative approach, it was found that improved modeling is possible in which the problem of simulation model is solved through machine learning through big data, and the problem of system structure and rule change is solved through hypothetical model through simulation modeling. Able to know.

Here, the cell automata model 111 is a discrete model handled in modeling such as mathematics, physics, complex systems, biology, and microstructures, and is defined in cells (cells) arranged in a regular grid form.

Each cell can have a finite number of states, and a grid is defined with a finite number of dimensions. For each cell, cells called neighbors are defined as a relationship to that cell, for example, neighbors can be defined as cells separated by one cell in all directions to one cell.

When the time t=0, the state of each cell is designated and this is called the initial state. The new generation is created from the previous generation by the state transition function, which is a mathematical function that specifies the cell's new state according to the state of each cell and its neighbors, that is, sets the behavioral rules of the cells.

In general, the rules of behavior are the same for each cell, do not change over time, and are applied to all cells of each generation at the same time.

When the cell automata and the state transition function will be described in detail with reference to FIG. 6, Cell(i, j) is a cell at a position (i, j), has a state s(i, j), and N is a cell (i , j) represents the neighboring cell pattern, and T represents the state transition function, that is, the agent's rule of action. The state of each cell can be defined according to the type of problem, such as traffic, water pollution, and fire spread.

7 is a diagram illustrating a state transition function of each cell in a cell automata, and it can be seen how the state transition function differs from a general model.

In general, the ideal state transition function reflects the state of neighboring cells to cause a state transition, but in the present invention, the state transition is received by receiving not only the state of the neighboring cells, but also the terrain information and external factors (weather conditions such as weather, temperature, etc.). Is made.

That is, it can be seen that more accurate prediction for the next cell is made by reflecting the characteristics of the terrain in the ideal model and even real-time weather information.

At this time, the state transition function is not the same for each cell or does not change over time as in the conventional method, but varies according to the change of the cell position and time, and the state transition rule is not deterministic and has uncertainty. If the state transition function is obtained using domain knowledge as in the past, validation is required, but if it is obtained through machine learning of big data, it is based on real data, so the problem of validation can be solved.

As described above, the system modeling method with built-in big data machine learning according to an embodiment of the present invention is not limited by this embodiment, although it has been described by the limited embodiments and drawings, and is not limited to the general technical field to which the present invention belongs. It goes without saying that various modifications and variations are possible within the scope of the technical spirit of the present invention and the scope of the claims to be described below by those of ordinary skill in the art.

100: target system 110: hypothetical model
111: cell automata model 120: big data
130: system model

Claims (6)

In a computing system that implements a model for a target system,
The computing system
A hypothetical model defined on the basis of structural information obtainable related to the target system through acquiring knowledge of the target system and including a plurality of functional blocks therein; And
At least one processor; Including,
The plurality of functional blocks are
At least one or more first functional blocks for inputting first data and outputting second data among big data obtained through actual operation and observation of the target system based on the structural information; And
At least one or more second functional blocks configured to input the second data of the big data and output the third data based on the structural information;
Including,
At least one or more of the at least one first functional block and the at least one second functional block is a machine learning functional block capable of machine learning,
The at least one or more processors
The first machine learning function to perform machine learning using a first machine learning function block included in the at least one or more first function blocks by receiving the first data as an input and outputting the second data among the big data. The second machine learning so that machine learning is performed by a second machine learning function block included in the at least one or more second function blocks by controlling a block or by receiving the second data as an input and the third data as an output. A computing system that controls a functional block and implements a model for the target system. The method of claim 1,
The hypothetical model and the plurality of functional blocks are defined based on obtainable domain knowledge, experience, and theory related to the target system. The method of claim 1,
The at least one or more processors
When machine learning is performed by the first machine learning function block, a verified first parameter value is obtained and the first parameter value is used as information to identify the first machine learning function block, or the second machine learning function A computing system for implementing a model for a target system, obtaining a verified second parameter value when machine learning is performed by a block and using the second parameter value as information for identifying the second machine learning function block. The method of claim 3,
The first parameter value or the second parameter value includes a variable value, probability, function, or graph input to the first machine learning function block or the second machine learning function block. Computing system. The method of claim 1,
The first data is data expressed as an input of the hypothetical model based on the structural information, the second data is data expressed as an internal variable of the hypothetical model based on the structural information, and the third A computing system for implementing a model for a target system, wherein data is data expressed as an output of the hypothetical model based on the structural information. In a computing system that implements a model for a target system,
The computing system
A hypothetical model defined on the basis of structural information obtainable related to the target system through acquiring knowledge of the target system and including a plurality of functional blocks therein; And
At least one processor; Including,
The plurality of functional blocks are
At least one or more first functional blocks for inputting first data and outputting second data among big data obtained through actual operation and observation of the target system based on the structural information; And
At least one or more second functional blocks configured to input second data of the big data and output third data based on the structural information;
Including,
At least one of the at least one or more first-stage functional blocks and the at least one or more two-stage functional blocks is a machine learning functional block capable of machine learning,
The at least one or more processors
Receive new input data for the target system,
Inputting the new input data into the hypothetical model and controlling the hypothetical model to execute an inference process of the hypothetical model,
A computing system for implementing a model for a target system, providing the output of the hypothetical model as a result of inference of the hypothetical model with respect to the output of the target system for the new input data.

Images