KR20190078850A - Method for estimation on online multivariate time series using ensemble dynamic transfer models and system thereof - Google Patents

Abstract

Provided are a real-time multivariate time series prediction method through a dynamic transition ensemble model, and a system thereof. The real-time multivariate time series prediction method through a dynamic transition ensemble model comprises the following steps. A real-time multivariate time series prediction system selects a predetermined number of a plurality of dynamic transition models to define a dynamic transition ensemble model based on the selected selection models. As the real-time multivariate time series prediction system receives real-time data while updating weights of each of the selection models included in the dynamic transition ensemble model or model coefficients of each of the selection models, and performs the real-time prediction.

Description

[0001] The present invention relates to a method and system for predicting real time multivariate time series through a dynamic transition ensemble model,

The present invention relates to a method and system for real-time multivariate time series prediction using a dynamic transition ensemble model, and more particularly, to a method and system for real-time multivariate time series prediction using an ensemble model using a plurality of dynamic transition models Method and system thereof.

Attempts have been made to predict specific events (e.g., errors or failures) in real time through various existing APM solutions or CCTV network other time series prediction models.

However, in a multivariate environment where there are a number of independent factors affecting system performance, conventional technologies are not capable of predicting through real-time analysis, but are merely monitoring.

In addition, the VAR (Vector Autoregressive) model, which is known as a conventional multivariate time series prediction model, is a prediction model using various assumptions, and it is difficult to cope with various situations due to the feature that a single model predicts variables at one time It is difficult to predict one target variable using multivariate data.

Accordingly, the present invention provides a method and system for enhancing the accuracy of prediction using an ensemble model combining a plurality of dynamic transition models.

The present invention also provides a method and system for improving predictive power by setting a target response variable among many variables using real time multivariate time series data.

In addition, it provides a method and system for selecting accurate models by offline learning and real-time accurate prediction in online.

According to an aspect of the present invention, a real time multivariate time series prediction method using a dynamic transition ensemble model is a method of real time multivariate time series prediction by defining a dynamic transition ensemble model based on selected selection models by selecting a predetermined number of a plurality of dynamic transition models The real time multivariate time series prediction system receives real time data and performs prediction in real time while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models .

The real-time multivariate time series prediction method using the dynamic transition ensemble model may further include the step of generating the plurality of dynamic transition models through the learning data by the real-time multivariate time series prediction system.

Wherein the step of generating the plurality of dynamic transition models through the learning data comprises linearly combining the autoregressive influence of the dependent variable Y and the influence of the independent variable x at the present time on the dependent variable Y, And generating models.

The dynamic transfer models may be generated by the following equations.

[Mathematical Expression]

는 자기회귀계수,

는 회귀모델 상수,

는 독립변수 x에 대한 1차 회귀계수,

는 오차항, m은 독립변수의 개수를 나타낸다. Where Y is the dependent variable, B is the backshift operator, d is the difference factor, p is the autoregressive order, I is the display function,

Is the autoregressive coefficient,

Regression model constants,

Is the first-order regression coefficient for the independent variable x,

Is the error term, and m is the number of independent variables.

개(여기서, q는 독립변수 선택개수)의 동적전이 모형들을 생성하는 것을 특징으로 할 수 있다.Wherein the generating of the plurality of dynamic transition models by the real-time multivariate time series prediction system through learning data comprises:

(Where q is an independent variable number of choices).

Wherein the step of defining a dynamic transition ensemble model based on the selected selection models by selecting a predetermined number of the plurality of dynamic transition models generated by the real time multivariate time series prediction system includes: As the selection models.

The dynamic transition ensemble model may be defined by the following equation.

Where w is the weight of each choice model.

Wherein the real time multivariate time series prediction system performs prediction in real time while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models using the following equation And updating the weight value.

Wherein the real time multivariate time series prediction system performs prediction in real time while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models by using a predetermined window size Updating model coefficients of each of the selected indices through an RLS method while updating the model coefficients of the selected indices.

Wherein the real time multivariate time series prediction system performs prediction in real time while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models, t-1), the weighting value is obtained.

According to an aspect of the present invention, there is provided a method for predicting a real time multivariate time series through a dynamic transition ensemble model, the method comprising: generating a plurality of dynamic transition models using learning data; Selecting a predetermined number of the plurality of dynamic transition models and using the dynamic transition ensemble model based on the selected selection models, receiving the real-time data and updating the weight of each of the selection models or the model coefficient of each of the selection models And performing prediction in real time.

The method may be implemented by a computer program installed in a data processing apparatus and recorded in a recording medium.

According to an aspect of the present invention, there is provided a real-time multivariate time series prediction system using a dynamic transition ensemble model, comprising: a DB storing information on each of a plurality of dynamic transition models; a selection model selecting a predetermined number of the plurality of dynamic transition models; An ensemble model module for defining a dynamic transition ensemble model based on the real-time data, a weight of each of the selection models included in the dynamic transition ensemble model after receiving real-time data or a model coefficient of each of the selection models, And a control module for performing the control.

The real time multivariate time series prediction system using the dynamic transition ensemble model may further include a dynamic transition model module for performing learning through learning data to generate the plurality of dynamic transition models.

According to an aspect of the present invention, there is provided a real time multivariate time series prediction system using a dynamic transition ensemble model, the system including a processor and a memory for storing a computer program executed by the processor, The above method can be performed.

According to the technical idea of the present invention, the accuracy of prediction can be enhanced by using an ensemble model combining a plurality of dynamic transition models. Also, there is an effect that the real-time prediction performance can be improved while updating the weight and / or model coefficient of each of the plurality of dynamic transition models in real time.

In addition, the model is selected through off-line learning, and accurate prediction can be performed in real time on-line.

Also, there is an effect that prediction power can be improved by setting a target response variable among many variables using real time multivariate time series data. Therefore, not only can it be used for predicting a failure in real time in a conventional application performance management solution or a CCTV network, but also real time multivariate time series data generated in a server network system such as log data or performance data generated in various databases or servers It can be applied in industries that require analysis, prediction, and use. If an appropriate threshold value can be set, it can be applied to the detection of anomaly, and it can be applied to fields such as an intrusion detection on a network, or the like, by using data generated in the medical industry.

BRIEF DESCRIPTION OF THE DRAWINGS A brief description of each drawing is provided to more fully understand the drawings recited in the description of the invention.
FIG. 1 shows an example of a real time multivariate time series prediction system using a dynamic transition ensemble model according to the technical idea of the present invention.
FIG. 2 is a schematic flow chart for implementing a real time multivariate time series prediction method using a dynamic transition ensemble model according to an embodiment of the present invention.
FIG. 3 and FIG. 4 are diagrams showing examples of learning data according to an embodiment of the present invention.
Figure 5 illustrates an example of a selection model selected from a plurality of dynamic transition models in accordance with an embodiment of the present invention.
FIG. 6 exemplarily shows a change in weight of each selected shape in an ensemble model according to an embodiment of the present invention.
7 is a diagram showing a prediction result according to the technical idea of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention is capable of various modifications and various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the detailed description. It is to be understood, however, that the invention is not to be limited to the specific embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another.

The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise.

In this specification, the terms "comprises" or "having" and the like refer to the presence of stated features, integers, steps, operations, elements, components, or combinations thereof, But do not preclude the presence or addition of features, numbers, steps, operations, components, parts, or combinations thereof.

Also, in this specification, when any one element 'transmits' data to another element, the element may transmit the data directly to the other element, or may be transmitted through at least one other element And may transmit the data to the other component. Conversely, when one element 'directly transmits' data to another element, it means that the data is transmitted to the other element without passing through another element in the element.

Hereinafter, the present invention will be described in detail with reference to the embodiments of the present invention with reference to the accompanying drawings. Like reference symbols in the drawings denote like elements.

FIG. 1 shows an example of a real time multivariate time series prediction system using a dynamic transition ensemble model according to the technical idea of the present invention.

Referring to FIG. 1, a real time multivariate time series prediction system (hereinafter, a 'prediction system') 100 using a dynamic transition ensemble model is implemented to realize a real time multivariate time series prediction method using a dynamic transition ensemble model according to the technical idea of the present invention .

The prediction system 100 may be a client installed in a terminal of a predetermined user or a system installed in a server. The prediction system 100 is installed in a predetermined data processing apparatus, and the software (e.g., application) implemented to implement the technical idea of the present invention and the hardware of the data processing apparatus for driving the software are organically combined Lt; / RTI >

The prediction system 100 may be a system for predicting time series data of a specific variable in a multivariate environment in which a plurality of independent variables exists according to the technical idea of the present invention. Such a prediction system 100 is advantageous in that a conventional multivariate time series prediction system (e.g., Var, etc.) can use one or a few specific variables (e.g., Dependent variable).

Hereinafter, the learning data and models exemplified in this specification are examples applied to the application performance management system, and the scope of the present invention is not limited thereto.

The prediction system 100 includes a control module 110 and an ensemble model module 120.

The prediction system 100 may further include a predetermined DB 130 and / or a dynamic transition model module 140. According to an embodiment of the present invention, some of the above-mentioned components may not necessarily correspond to the components necessary for implementation of the present invention, and according to an embodiment, But may include more components.

The prediction system 100 may include hardware resources and / or software necessary to implement the technical idea of the present invention, and does not necessarily mean one physical component or a single device . That is, the prediction system 100 may mean a logical combination of hardware and / or software provided to implement the technical idea of the present invention. If necessary, the prediction system 100 may be installed in a separate apparatus to perform respective functions And may be embodied as a set of logical structures for realizing the technical idea of the present invention.

In this specification, a module may mean a functional and structural combination of hardware for carrying out the technical idea of the present invention and software for driving the hardware. For example, the module may refer to a logical unit of a predetermined code and a hardware resource for executing the predetermined code, and it does not necessarily mean a physically connected code or a kind of hardware But can be easily deduced to the average expert in the field of the present invention.

The control module 110 may perform functions and / or functions of other components included in the prediction system 100, such as an ensemble model module 120, a DB 130, and / or a dynamic transition model module 140, Or resources.

The prediction system 100 according to the technical idea of the present invention has many assumptions about multivariate time series prediction and it is difficult to predict one specific variable by using multivariate data. In order to solve the problem that is not robust to various situations, And a dynamic transition ensemble model combining models.

For this, the ensemble model module 120 can select a predetermined number of the plurality of predefined dynamic transition models. And we can define selected dynamic transition models, ie, dynamic transition ensemble models based on selection models. This dynamic transition ensemble model can be defined by the following equation.

[Equation 1]

는 k 번째 선택모형을 의미할 수 있다.Where w is the weight of each selection model,

Is the k-th choice model.

The ensemble model module 120 can select K dynamic transition models from among a plurality of dynamic transition models as a selection model. Of course, the ensemble model module 120 may select the K selection models in the order of models having superior prediction performance, and may select various schemes.

Information about a plurality of dynamic transition models may be stored in the DB 130. [

In addition, the DB 130 stores learning data for learning the plurality of dynamic transition models, information on dynamic transition models according to learning results (i.e., model coefficients and errors for defining dynamic transition models), and The information on the performance measurement results of the dynamic transition models of the model can be further stored.

The dynamic transfer models may be multivariate time series models constructed through off-line learning. That is, each dynamic transition model can operate as one multivariate time series model.

The dynamic transfer models can be learned through learning data, and coefficients for defining a dynamic transfer model can be defined through the learning result.

The dynamic transformation model module 140 may generate a plurality of such dynamic transformation models. It is needless to say that information on the generated dynamic transition models and related information (e.g., learning data, performance, etc.) can be stored in the DB 130.

)들의 영향을 선형 결합하여 상기 동적전이 모형들을 생성할 수 있다.The dynamic transfer model module 140 calculates an autoregressive influence portion and an independent variable

) Can be linearly combined to generate the dynamic transfer models.

These dynamic transfer models can be generated by the following equations.

&Quot; (2) "

으로 표현될 수 있다. Here,

. ≪ / RTI >

)들의 영향은

로 표현될 수 있다.In addition,

The influence of

. ≪ / RTI >

,

는 자기회귀 계수(autoregression coefficient),

는 회귀모델 상수,

는 독립변수에 대한 1차 회귀 계수(simple linear regression coefficient),

는 오차, m은 독립변수의 개수를 의미할 수 있다.In addition, Y is the dependent variable, B is the backshift operator, d is the difference factor, p is the autoregressive order, I is the indicator function,

,

Is an autoregression coefficient,

Regression model constants,

Is the simple linear regression coefficient for independent variables,

Is the error, and m is the number of independent variables.

According to one embodiment, the data as shown in FIG. 3 can be the learning data.

FIG. 3 shows an example of learning data according to an embodiment of the present invention. The learning data may be learning data used in an application performance management solution.

As shown in FIG. 3, the response speed of the learning data can be the dependent variable Y and the remaining variables (e.g., Totla_mem, Free_mem, Used_mem, Usage, sys_cpu, user_cpu, count, http, javad, jdbc) Can all be independent variables (x). Therefore, it may be a dynamic transition ensemble model according to the technical idea of the present invention for predicting the response speed when the independent variables are changed in a time-wise manner with each value using each of these learning data.

The dynamic transfer model module 140 may generate a plurality of dynamic transfer models using Equation (2).

For example, the dynamic transfer modeling module 140,

&Quot; (3) "

(Where q is an independent variable number of choices).

는 독립변수가 q개인 경우 전체 독립변수의 개수(m) 중에서 서로 다른 1개를 뽑을 수 있는 조합의 수부터 서로 다른 q개까지 뽑을 수 있는 조합의 수를 의미할 수 있다. 즉, [수학식 3]은 종속변수에 영향을 미치는 독립변수가 q개 이하라고 가정할 경우, 생성될 수 있는 모든 서로 다른 수학식 2의 개수가 될 수 있다.That is, in the case of the possible autoregressive order (p), possible differential order (d) and number of independent variables affecting the dependent variable, the number of all possible different equations (2) here

Is the number of combinations that can be selected from the number of combinations that can be drawn from one different among the total number (m) of independent variables when q is independent variable. That is, [Equation 3] can be the number of all different Equation 2 that can be generated assuming that the number of independent variables that affect the dependent variable is q or less.

When a plurality of different dynamic transition models (for example, the number of Equation 3) are generated and learned, the learned dynamic transition models can be stored in the DB 130, respectively.

According to one embodiment, the dynamic transfer model module 140 generates a dynamic transfer model using some (e.g., 3000) of learning data (e.g., 4000) Can be used to update the generated dynamic transfer model. A predetermined measurement value (e.g., RMSE: Root Meas Square Error) capable of evaluating the performance of each model may be stored in the DB 130 as well.

For example, the data shown in FIG. 5 represent the identification numbers of dynamic transition models sorted in ascending order based on performance measures (e.g., RMSE). That is, it can mean that the 1628th dynamic transition model has the highest performance measure (for example, 0.3679042).

Then, the ensemble model module 120 can select a predetermined number (for example, K) of the different dynamic transition models stored in the DB 130 as a model for use in the dynamic transition ensemble model, that is, a selection model. Of course, the ensemble model module 120 can select K dynamic transformation models as good as the performance evaluation results of the respective dynamic transformation models. Here, K may be arbitrarily selected, or it may be determined through a plurality of iterative simulations that the number of training data and K suitable for various situations are determined.

Then, the dynamic transition ensemble model defined by the ensemble model module 120 may be expressed by Equation (1).

If the dynamic transition ensemble model is defined as described above, the control module 110 receives real-time data and updates the weight of each of the selection models included in the dynamic transition ensemble model and / or the model coefficient of each of the selection models Prediction can be performed in real time. In other words, time series prediction can be performed in real time as the ensemble model becomes dynamic transition.

등)는 선택모형을 정의하는 파라미터 자체가 실시간으로 동적 전이됨을 의미할 수 있다. The fact that the control module 110 updates the weights of each of the selection models may mean that the influence or performance of each selection model is dynamic transition in real time. The model coefficients of each of the selection models (eg,

Etc.) may mean that the parameter itself defining the selection model is dynamic transition in real time.

As described above, the control module 110 has an effect of improving prediction performance by updating the dynamic transition ensemble model while real-time dynamic transition occurs as real-time data is input.

The control module 110 may update the weights of each of the selection models using the following equation.

&Quot; (4) "

는 동적전이 모형에 의해 예측된 예측값을 의미할 수 있으며, 수학식 4의 왼쪽 식은 현재 시점의 다음시점부터 예측하고자 하는 시간 타임 개수(T)까지의 실제 데이터(

)와 예측값(

)의 차이의 제곱의 합에 반비례하여 해당 동적전이 모형의 가중치가 연산됨을 의미할 수 있다. 또한 수학식 4의 오른쪽 식은 선택모형들 각각의 가중치의 합은 1이 됨을 의미할 수 있으며 각각의 가중치는 전체 가중치의 합에서 차지하는 해당 가중치의 비율을 나타냄을 의미할 수 있다.here

The left equation of Equation (4) represents the predicted value predicted by the dynamic transition model, and the left equation of Equation (4) represents the actual data from the next time point of the current point to the time time number T to be predicted

) And the predicted value (

), And the weight of the corresponding dynamic transfer model is calculated in inverse proportion to the sum of the squares of the differences. Also, the right-hand side of Equation (4) can mean that the sum of the weights of each of the selection models is 1, and each weight can represent the ratio of the corresponding weights to the sum of all the weights.

Meanwhile, the control module 110 can update the model coefficients of each of the selection models. At this time, as is well known, the control module 110 performs forward updating using a predetermined window size, The model coefficients of each of the selection models can be updated.

을 이용하여 해당

번째 선택모형의 가중치를 구해야하지만 온라인(실시간)으로 순차적으로 데이터가 입력되는 상황이므로

이 존재하지 않게 된다. 따라서 이때에는 (t-1)의 시점에서의 오차 즉,

를 이용하여 가중치

를 구할 수도 있다.According to Equation (5), if the control module 110 predicts the next time point immediately after the current time point,

Using

We need to find the weight of the second selection model. However, since the data is sequentially input in the online (real time)

Is not present. Therefore, at this time, the error at the time point of (t-1)

Lt; / RTI >

.

를 이용하여 가중치

를 구할 수도 있다.Similarly, when the present time is t, the weight (e.g., t + n) at a later time point is an error at the time point of (tn)

Lt; / RTI >

.

The real time multivariate time series prediction method using the dynamic transition ensemble model according to the technical idea of the present invention can be summarized as shown in FIG.

FIG. 2 is a schematic flow chart for implementing a real time multivariate time series prediction method using a dynamic transition ensemble model according to an embodiment of the present invention.

Referring to FIG. 2, the prediction system 100 according to the technical idea of the present invention can receive learning data and perform learning in off-line (S100, S110).

As a result, a plurality of dynamic transition models (e.g., Equation 3) may be generated as described above (S120). Each dynamic transfer model can be defined by equation (2).

Then, the prediction system 100 selects K of the plurality of dynamic transition models to construct a dynamic transition ensemble model (S130). The constructed dynamic transition ensemble model can be expressed by Equation (1).

In operation S140, the prediction system 100 may obtain the input data in real time (S140), and may perform the prediction while updating the weights and / or model coefficients of each selection model in real time (S150).

An embodiment according to the technical idea of the present invention will be described with reference to FIG. 3 to FIG.

FIG. 3 shows an example of learning data according to an embodiment of the present invention, FIG. 5 shows an example of a selection model selected from a plurality of dynamic transition models according to an embodiment of the present invention, FIG. 6 shows a weight change of each of the selection models in the ensemble model according to the embodiment of the present invention, and FIG. 7 is a diagram showing a prediction result according to the technical idea of the present invention.

According to an embodiment of the present invention, predictions for application to a talent management solution of an application are performed using the learning data of the kind shown in FIG. 3 as described above. The dependent variable to be predicted is the response speed.

4000 pieces of actual learning data are used, and the learning data thus used is as shown in FIG.

Among the 4000 learning data, the prediction system 100 constructs a plurality of dynamic transition models using 3000 learning data, and updates the respective dynamic transition models using the remaining 1000 learning data.

Then, K (for example, 40) selection models are selected using the performance values (for example, RMSE) of the respective dynamic transition models as shown in FIG. FIG. 5 shows performance values of the selected model and the selected model according to the prediction performance of t + 6.

We define the dynamic transition ensemble model using selected K selection models, and perform predictions by updating weights and model coefficients.

As shown in FIG. 6, it is confirmed that the weighting of K (40) selection models is changed as shown in FIG. 6.

Also, FIG. 7 shows the prediction result of t + 6 in the above environment, where the black solid line is the actual data and the red solid line represents the predicted value through the dynamic transition ensemble model according to the embodiment of the present invention. As shown in FIG. 7, a relatively excellent prediction performance is exhibited.

The real time multivariate time series prediction method using the dynamic transition ensemble model according to the technical idea of the present invention analyzes and predicts by using real time multivariate time series data generated in the server network system such as log data or performance data generated in various databases or servers It may be applicable in the required industry. In addition, if an appropriate threshold can be set, it can be applied to the detection of anomaly, and it can be applied to fields such as intrusion detection of a network, or the like, by using data generated in the medical industry.

Meanwhile, the real-time multivariate time series prediction method using the dynamic transition ensemble model according to the embodiment of the present invention can be implemented as a computer-readable program command and stored in a computer-readable recording medium. The control program and the target program according to the present invention can also be stored in a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored.

Program instructions to be recorded on a recording medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of software.

Examples of the computer-readable recording medium include magnetic media such as a hard disk, a floppy disk and a magnetic tape, optical media such as CD-ROM and DVD, a floptical disk, And hardware devices that are specially configured to store and execute program instructions such as magneto-optical media and ROM, RAM, flash memory, and the like. The computer readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

Examples of program instructions include machine language code such as those produced by a compiler, as well as devices for processing information electronically using an interpreter or the like, for example, a high-level language code that can be executed by a computer.

The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

It is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents. .

Claims (15)

Selecting a predetermined number of the plurality of dynamic transition models and defining a dynamic transition ensemble model based on the selected selection models;
Wherein the real time multivariate time series prediction system receives real time data and performs prediction while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models, A Real - Time Multivariate Time Series Prediction Method Using Model.
The method of claim 1, wherein the real time multivariate time series prediction method using the dynamic transition ensemble model comprises:
Wherein the real time multivariate time series prediction system further comprises generating the plurality of dynamic transition models through training data.
3. The method of claim 2, wherein the real-time multivariate time series prediction system generates a plurality of dynamic transition models through learning data,
Generating a dynamic transition model by linearly combining the autoregressive influence of the dependent variable Y and the influence of the independent variable x at the present time on the dependent variable Y to generate a real time multivariate time series prediction model using the dynamic transition ensemble model.
4. The method of claim 3, wherein the dynamic transition models are generated by the following equation.
[Mathematical Expression]

Where Y is the dependent variable, B is the backshift operator, d is the difference factor, p is the autoregressive order, I is the display function,

Is the autoregressive coefficient,

Regression model constants,

Is the first-order regression coefficient for the independent variable x,

Is the error term, and m is the number of independent variables.
5. The method of claim 4, wherein the real time multivariate time series prediction system generates the plurality of dynamic transition models through learning data,

(Where q is the number of independent variables selected) is generated by using a dynamic transition ensemble model.
The method as claimed in claim 1, wherein the real-time multivariate time series prediction system includes a plurality of dynamic transition models, and the dynamic transition ensemble model is defined based on the selected selection models,
And selecting K, which is excellent in prediction performance among the plurality of dynamic transition models, as the selection models.
The method of claim 1, wherein the dynamic transient ensemble model is defined by the following equation:

Where w is the weight of each choice model.
The method of claim 1, wherein the real-time multivariate time series prediction system performs prediction in real time while updating a weight of each of the selection models included in the dynamic transition ensemble model or model coefficients of each of the selection models,
Wherein the weights are updated using the following equation: < EMI ID = 15.0 >

The method of claim 1, wherein the real-time multivariate time series prediction system performs prediction in real time while updating a weight of each of the selection models included in the dynamic transition ensemble model or model coefficients of each of the selection models,
And updating the model coefficients of each of the selected indices through an RLS method while performing a forward update using a predetermined window size. The present invention also provides a method for predicting real time multivariate time series through a dynamic transition ensemble model.
The method of claim 1, wherein the real-time multivariate time series prediction system performs prediction in real time while updating a weight of each of the selection models included in the dynamic transition ensemble model or model coefficients of each of the selection models,
Wherein a weight is obtained by using a predicted error at a previous time point (t-1) in a current time point t, and a real time multivariate time series prediction method using a dynamic transition ensemble model.
A real time multivariate time series prediction system generating the plurality of dynamic transition models through learning data;
Selecting a predetermined number of the plurality of dynamic transition models generated by the real time multivariate time series prediction system and using the dynamic transition ensemble model based on the selected selection models, A method for predicting real time multivariate time series through a dynamic transition ensemble model, comprising the step of performing prediction while updating the model coefficients of each of the selection models.
A computer program installed in a data processing apparatus and recorded on a recording medium for performing the method according to any one of claims 1 to 11.
A DB storing information on each of the plurality of dynamic transition models;
An ensemble model module for defining a dynamic transition ensemble model based on selected selection models by selecting a predetermined number of the plurality of dynamic transition models;
A real time multivariate ensemble model with a real time data and a control module for performing prediction while updating the weight of each of the selection models included in the dynamic transition ensemble model or the model coefficient of each of the selection models, Time Series Prediction System.
14. The system of claim 13, wherein the real-time multivariate time series prediction system using the dynamic transition ensemble model comprises:
A real time multivariate time series prediction system using a dynamic transition ensemble model further comprising a dynamic transition model module for performing learning through learning data to generate the plurality of dynamic transition models.
A processor; And
A memory for storing a computer program executed by the processor,
The computer program, when executed by the processor, causes the method according to any one of claims 1 to 10 to be performed.

Images