KR20240015453A - Method and apparatus for predicting critical state of asset price bubble using improved log periodic power law (lppl) algorithm - Google Patents

Abstract

In these embodiments, in order to optimize the parameters of the LPPL model, initial LPPL parameters are generated through price peak extraction, hyperparameters that control the evolution strategy of the genetic algorithm, which is a nonlinear optimization algorithm, are adjusted, and a genetic algorithm with improved hyperparameters is used. After optimizing the LPPL parameters by applying , we provide an asset bubble criticality prediction method and device that accurately predicts the bubble critical point through the LPPL model based on the optimized LPPL parameters.

Description

Asset bubble criticality prediction method and device using improved LPPL (Log Periodic Power Law) algorithm {METHOD AND APPARATUS FOR PREDICTING CRITICAL STATE OF ASSET PRICE BUBBLE USING IMPROVED LOG PERIODIC POWER LAW (LPPL) ALGORITHM}

The technical field to which the present invention belongs relates to a method and device for predicting the asset bubble critical state by estimating the initial parameters of the Log Periodic Power Law (LPPL) model and optimizing the parameters with a non-linear genetic algorithm.

The content described in this section simply provides background information for this embodiment and does not constitute prior art.

The bursting of the financial asset bubble threatens to cause losses reaching trillions of dollars in a very short period of time. Accordingly, predicting the collapse of financial assets affects various stakeholders involved.

When collapse is defined as a change in the state of a phenomenon at a critical point, it is very difficult to predict the exact collapse time of various phenomena, but it can be used as an important means of effectively handling future crises. LPPL (Log Periodic Power Law), used as a tool for predicting collapse, was first developed in statistical physics and has proven to be useful in predicting bubbles. This was started by focusing on the physical properties (super exponential growth + log frequency) revealed in the precursor phenomenon before collapse. The idea was derived from the fact that as a critical point approaches, the volatility of a certain phenomenon becomes more frequent and the cycle becomes shorter. Using this approach, financial crises can be analyzed based on time series price data of financial assets.

Using LPPL, a more improved LPPL algorithm can be designed to suit the various characteristics of the asset market (e.g., financial market, real estate market, crypto market, etc.) to be analyzed.

“Forecasting Financial Crashes: Revisit to Log-Periodic Power Law “, Kwangwon Ahn, (2018.08.01) "Can government stabilize the housing market? The evidence from South Korea ", Kwangwon Ahn, (2020.07.15) “Boost and Burst: Bubbles in the Bitcoin Market”, Kwangwon Ahn, (2020.06.15)

In order to optimize the parameters of the LPPL model, embodiments of the present invention generate initial LPPL parameters through price peak extraction, adjust hyperparameters of a genetic algorithm, which is a nonlinear optimization algorithm, and apply a genetic algorithm with improved hyperparameters. The main purpose is to optimize the LPPL parameters and then accurately predict the bubble critical point through the LPPL model based on the optimized LPPL parameters.

Other unspecified objects of the present invention can be additionally considered within the scope that can be easily inferred from the following detailed description and its effects.

According to one aspect of the present embodiment, in the asset bubble threshold prediction method, defining an LPPL model with a plurality of LPPL parameters according to LPPL (Log Periodic Power Law) theory; selecting a peak within a circulation period in the LPPL model to calculate an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model; Calculating optimized LPPL parameters using a genetic algorithm based on the initial LPPL parameters; adjusting hyper-parameters that control the evolution strategy of the genetic algorithm; And providing an asset bubble criticality prediction method including calculating the bubble burst point from time series price data through an LPPL model applying the optimized LPPL parameters.

The plurality of LPPL parameters include a price parameter, a distance parameter, a wave parameter, a time parameter, a phase parameter, a frequency parameter, and an exponential parameter, and the LPPL model includes the price parameter, the distance parameter, and the wave parameter according to the LPPL theory. It can be expressed as a relationship between the parameters, the time parameter, the phase parameter, the frequency parameter, and the exponential parameter.

The step of calculating the initial LPPL parameter includes detecting a peak group having a plurality of peaks within the circulation period using the time series price data, and differentially applying peak weights to the plurality of peaks in consideration of the current point in time. And, screen the plurality of peaks to which the peak weight is applied, select the first peak group with the largest value, and calculate the initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the first peak group. You can.

The step of calculating the optimized LPPL parameters includes applying a genetic algorithm to the initial LPPL parameters to derive an optimized value through repeated selection, breeding, mutation, and crossing considering the evolutionary process of the population, thereby generating the optimized LPPL parameters. Parameters can be calculated.

The step of calculating the initial LPPL parameter includes screening a plurality of peaks to which the peak weight is applied, selecting a first peak group with the largest value, and calculating the LPPL parameter of the LPPL model based on the first peak group. Calculate a first initial LPPL parameter to be applied, screen a plurality of peaks to which the peak weight is applied, select a second peak group with a second largest value, and select the LPPL parameter of the LPPL model based on the second peak group. A second initial LPPL parameter to be applied to can be calculated.

The step of calculating the optimized LPPL parameters may be performed by starting the evolution of the population of individuals from the first initial LPPL parameters and temporarily applying the second initial parameters to the selection, breeding, mutation, and crossover. .

In the step of adjusting the hyper-parameters that control the evolution strategy of the genetic algorithm, the hyper-parameters may be changed using the second initial parameters.

The step of adjusting the hyperparameters that control the evolution strategy of the genetic algorithm includes population population options, fitness scaling options, selection options, breeding options, mutation options, crossover options, stopping criteria options, or these that affect the evolution strategy. The combination can be changed.

According to another aspect of the present embodiment, in the asset bubble threshold prediction device, an LPPL model defined with a plurality of LPPL parameters according to LPPL (Log Periodic Power Law) theory; a first initial parameter calculation unit that selects a peak within a circulation period in the LPPL model and calculates an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model; a parameter optimization unit that calculates optimized LPPL parameters using a genetic algorithm based on the initial LPPL parameters; a hyper-parameter control unit that adjusts hyper-parameters that control the evolution strategy of the genetic algorithm; and a bubble burst prediction unit that calculates a bubble burst point from time series price data through the LPPL model applying the optimized LPPL parameters.

The plurality of LPPL parameters include a price parameter, a distance parameter, a wave parameter, a time parameter, a phase parameter, a frequency parameter, and an exponential parameter, and the LPPL model includes the price parameter, the distance parameter, and the wave parameter according to the LPPL theory. It can be expressed as a relationship between the parameters, the time parameter, the phase parameter, the frequency parameter, and the exponential parameter.

The first initial parameter calculation unit detects a peak group having a plurality of peaks within the circulation period using the time series price data, and differentially applies peak weights to the plurality of peaks in consideration of the current point in time, By screening a plurality of peaks to which the peak weight is applied, the first peak group with the largest value is selected, and an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model can be calculated based on the first peak group. .

The parameter optimization unit may calculate the optimized LPPL parameters by applying a genetic algorithm to the initial LPPL parameters to derive optimized values through repeated selection, breeding, mutation, and crossover in consideration of the evolution process of the population. .

The asset bubble threshold prediction device includes a second initial parameter calculation unit accessible to the LPPL model; and a peak group adjustment unit connected to the first initial parameter calculation unit and the second initial parameter calculation unit.

The peak group adjuster screens the plurality of peaks to which the peak weight is applied and selects the first peak group with the largest value, and the first initial parameter calculation unit selects the LPPL of the LPPL model based on the first peak group. A first initial LPPL parameter to be applied to the parameter may be calculated.

The peak group adjuster screens a plurality of peaks to which the peak weight is applied and selects a second peak group with a second largest value, and the second initial parameter calculation unit selects the LPPL of the LPPL model based on the second peak group. A second initial LPPL parameter to be applied to the parameter may be calculated.

The parameter optimization unit starts the evolution of the population of entities from the first initial LPPL parameter, temporarily applies and changes the second initial parameter to the selection, breeding, mutation, and intersection, and the hyperparameter adjuster The hyper parameter can be changed using the second initial parameter.

The hyper-parameter control unit may change population options, adaptive scaling options, selection options, breeding options, mutation options, intersection options, stopping criteria options, or a combination thereof that affect the evolution strategy.

According to another aspect of this embodiment, in a computer-readable storage medium storing a computer program executable by a computer, a computer-readable storage medium storing a computer program capable of executing an asset bubble criticality prediction method is provided.

According to another aspect of the present embodiment, a computer program recorded on a computer-readable storage medium comprising computer program instructions executable by a processor, wherein when the computer program instructions are executed by the processor, a method for predicting asset bubble threshold It provides a computer program characterized in that it performs.

As described above, according to embodiments of the present invention, the initial LPPL parameters of the LPPL model are generated through price peak extraction, the hyperparameters that control the evolution strategy of the genetic algorithm, which is a nonlinear optimization algorithm, are adjusted, and the hyperparameters are There is an effect of optimizing LPPL parameters by applying an improved genetic algorithm.

The LPPL model designed based on optimized LPPL parameters has the effect of improving prediction accuracy at the bubble critical point.

Even if the effects are not explicitly mentioned here, the effects described in the following specification and their potential effects expected by the technical features of the present invention are treated as if described in the specification of the present invention.

Figure 1 is a diagram illustrating an asset bubble threshold prediction device according to an embodiment of the present invention.
Figure 2 is a flowchart illustrating an asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 3 is a flowchart illustrating the step of calculating the LPPL parameter of the asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 4 is a flowchart illustrating the steps of optimizing the LPPL parameter of the asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 5 is a diagram illustrating LPPL parameters calculated by the asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 6 is a diagram illustrating hyper parameters adjusted by the asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 7 is a diagram illustrating optimized LPPL parameters derived by an asset bubble threshold prediction method according to another embodiment of the present invention.
Figure 8 is a diagram illustrating the relative error results for bubble threshold detection of the LPPL model applying the optimized LPPL parameters derived by the asset bubble threshold prediction method according to another embodiment of the present invention.
Figures 9 to 11 are diagrams illustrating the sensitivity of performance error indicators between LPPL parameters of the LPPL model applied by the asset bubble threshold prediction method according to another embodiment of the present invention.
Figures 12 and 13 are diagrams showing asset bubble threshold prediction accuracy according to simulation results of embodiments of the present invention.
Figure 14 is a block diagram illustrating a computing device practicing embodiments of the present invention.

Hereinafter, in describing the present invention, if it is determined that related known functions may unnecessarily obscure the gist of the present invention as they are obvious to those skilled in the art, the detailed description will be omitted, and some embodiments of the present invention will be described. It will be described in detail through exemplary drawings.

LPPL (Log Periodic Power Law) theory is a methodology that informs the state change at a critical point, that is, the point of collapse. LPPL theory considers super exponential growth and log frequency properties, which correspond to the physical properties in the precursor phenomenon before collapse.

The exponential growth characteristic is, for example, a phenomenon in which the price increases by 5%, 10%, 20%, 40%, etc. every month, and can be expressed as an exponential function. Internal participants promote future returns with positive feedback, but this phenomenon cannot be sustained, so it collapses at a critical point.

The logarithmic frequency property indicates that fluctuations become more frequent as a critical point approaches. This means that the buy-sell cycle of internal participants is becoming more frequent.

Through the LPPL model, financial crises, etc. can be identified as predictable factors, and the interaction between internal factors causes collapse due to system instability.

The LPPL model can be expressed mathematically as follows.

The LPPL model considers exponential growth and logarithmic frequency to determine the price parameter (A), distance parameter (B), wave parameter (C), time parameter (t _c ), phase parameter (Φ), frequency parameter (ω), and It is expressed as a relational expression of the exponent parameter (β).

The LPPL model is expressed as the sum of the price parameter and the first relational expression. The first relational expression is expressed as the product of the distance parameter and the second relational expression and the third relational expression. The second relational expression subtracts time from the time parameter and applies the exponential parameter. The third relation is expressed as the sum of the first and fourth relations. The fourth relation is expressed as the product of the wave parameter and the cosine function. As an independent variable of the cosine function, the frequency parameter is multiplied by the logarithmic function and the phase parameter is added. The value obtained by subtracting time from the time parameter is applied as the independent variable of the logarithmic function.

Figure 1 is a diagram illustrating an asset bubble threshold prediction device according to an embodiment of the present invention.

Referring to FIG. 1, the asset bubble threshold prediction device 10 includes an LPPL model 15, a first initial parameter calculation unit 100, a parameter optimization unit 200, a hyper parameter adjustment unit 400, and a bubble collapse prediction unit. Includes 300.

The LPPL model (15) is defined with a plurality of LPPL parameters according to the Log Periodic Power Law (LPPL) theory. The plurality of LPPL parameters may include a price parameter, a distance parameter, a wave parameter, a time parameter, a phase parameter, a frequency parameter, and an exponential parameter. The LPPL model can be expressed as a relationship between price parameters, distance parameters, wave parameters, time parameters, phase parameters, frequency parameters, and exponential parameters according to LPPL theory.

The first initial parameter calculation unit 100 selects a peak within a circulation period in the LPPL model and calculates an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model. The first initial parameter calculation unit 100 detects a peak group having multiple peaks within a circulation period using time series price data, differentially applies peak weights to the multiple peaks in consideration of the current time, and applies peak weights to the peaks. By screening a plurality of weighted peaks, the first peak group with the largest value can be selected, and an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model can be calculated based on the first peak group.

The parameter optimization unit 200 calculates optimized LPPL parameters using a genetic algorithm based on the initial LPPL parameters. The parameter optimization unit 200 can calculate optimized LPPL parameters by applying a genetic algorithm to the initial LPPL parameters to derive optimized values through repeated selection, breeding, mutation, and crossover in consideration of the evolutionary process of the population. .

The hyper parameter control unit 400 adjusts hyper parameters that control the evolution strategy of the genetic algorithm.

The bubble burst prediction unit 300 calculates the bubble burst point from time series price data through an LPPL model applying optimized LPPL parameters.

The asset bubble threshold prediction device 10 may include a second initial parameter calculation unit 150 that can access the LPPL model. The asset bubble threshold prediction device 10 may include a peak group adjustment unit 500 connected to the first initial parameter calculation unit 100 and the second initial parameter calculation unit 150.

The peak group control unit 500 screens a plurality of peaks to which peak weights are applied and selects the first peak group with the largest value, and the first initial parameter calculation unit 100 models the LPPL based on the first peak group. The first initial LPPL parameter to be applied to the LPPL parameter of can be calculated.

The peak group control unit 500 screens a plurality of peaks to which peak weights are applied and selects the second peak group with the second largest value, and the second initial parameter calculation unit 150 models the LPPL based on the second peak group. A second initial LPPL parameter to be applied to the LPPL parameter of can be calculated.

The parameter optimization unit 200 starts the evolution of the population of entities from the first initial LPPL parameters, temporarily applies and changes the second initial parameters for selection, breeding, mutation, and crossover, and the hyperparameter control unit 400 2 Hyper parameters can be changed using the initial parameters.

The hyperparameter control unit 400 may change population options, adaptive scaling options, selection options, breeding options, mutation options, crossover options, stopping criteria options, or combinations thereof that affect the evolution strategy.

Among the various components illustratively shown in FIG. 1, some components may be omitted or other components may be additionally included.

Components included in the asset bubble criticality prediction device are shown separately in FIG. 1, but a plurality of components may be combined and implemented as at least one module. Components are connected to a communication path that connects software modules or hardware modules within the device and operate organically with each other. These components communicate using one or more communication buses or signal lines.

The asset bubble criticality prediction device may be implemented in a logic circuit using hardware, firmware, software, or a combination thereof, and may also be implemented using a general-purpose or special-purpose computer. The device may be implemented using hardwired devices, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc. Additionally, the device may be implemented as a System on Chip (SoC) including one or more processors and a controller.

The asset bubble criticality prediction device may be mounted on a computing device or server equipped with hardware elements in the form of software, hardware, or a combination thereof. A computing device or server includes all or part of a communication device such as a communication modem for communicating with various devices or a wired or wireless communication network, a memory for storing data to execute a program, and a microprocessor for executing a program to perform calculations and commands. It can refer to a variety of devices, including:

Figure 2 is a flowchart illustrating an asset bubble threshold prediction method according to another embodiment of the present invention.

The asset bubble criticality prediction method may be performed by a computing device or an asset bubble criticality prediction apparatus.

Referring to FIG. 2, in step S210, a step of defining an LPPL model with a plurality of LPPL parameters according to LPPL (Log Periodic Power Law) theory may be performed.

In step S220, a peak within the circulation period in the LPPL model may be selected to calculate an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model.

In step S230, an optimized LPPL parameter may be calculated using a genetic algorithm based on the initial LPPL parameter.

In step S240, a step may be performed to adjust hyperparameters that control the evolution strategy of the genetic algorithm.

In step S250, the time of bubble collapse can be calculated from time series price data through the LPPL model applying the optimized LPPL parameters.

Figure 3 is a flowchart illustrating the step of calculating the LPPL parameter of the asset bubble threshold prediction method according to another embodiment of the present invention.

Referring to FIG. 3, the date and price are input as time series data in step S310. In step S320, the repetition variable i is initialized.

In step S330, a peak group having multiple peaks within a circulation period is detected using time series price data. In step S340, peak weights are differentially applied to a plurality of peaks considering the current point in time.

In step S350, a plurality of peaks to which peak weights are applied are screened and the first peak group with the largest value is selected. In step S360, an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model is calculated based on the first peak group. Considering that the period is 2π, solutions to equations for a plurality of parameters can be found by specifying a plurality of specific values based on the peak group.

In step S370, LPPL parameters are optimized based on a genetic algorithm. The population is re-estimated using a non-linear algorithm. Apply the initial LPPL parameters as the initial population. In the process of re-estimation, the limiting range of LPPL parameters can be considered. Whenever changed values of each parameter (child generation data) are generated, they are added to the population. Afterwards, RMSE is calculated for each parameter vector, and if this value has superiority considering the upper ratio, it is designated as the parent generation of the next generation and the same process is repeated.

In step S380, it is determined whether the repetition variable exceeds the maximum repetition variable, and in step S390, the final LPPL parameter is output.

In the process of calculating the initial LPPL parameters, a plurality of peaks to which peak weights are applied are screened to select the first peak group with the largest value, and the first initial LPPL to be applied to the LPPL parameters of the LPPL model based on the first peak group. Calculate parameters.

In the process of calculating the initial LPPL parameters, a plurality of peaks to which peak weights are applied are screened to select a second peak group with the second largest value, and a second initial LPPL to be applied to the LPPL parameters of the LPPL model based on the second peak group. Parameters can be calculated.

Figure 4 is a flowchart illustrating the steps of optimizing the LPPL parameter of the asset bubble threshold prediction method according to another embodiment of the present invention.

Referring to FIG. 4, hyperparameters of the genetic algorithm are set in step S410. Adjusting the hyperparameters that control the evolutionary strategy of a genetic algorithm changes the population options, fitness scaling options, selection options, breeding options, mutation options, crossover options, stopping criteria options, or combinations thereof that affect the evolution strategy. You can.

In step S420, a plurality of entity groups are created. In step S430, population 1 is evolved. Migration is performed in step S440. Afterwards, population population 2 is evolved and migration is performed again. Repeat these actions.

Extract the elite in step S450. In step S460, it is determined whether the elite has changed, and in step S470, whether the performance error index has changed. The steps are repeated depending on whether the elite has changed and whether the performance error index (e.g. RMSE) has changed.

The step of calculating the optimized LPPL parameters is to calculate the optimized LPPL parameters by applying a genetic algorithm to the initial LPPL parameters to derive optimized values through repeated selection, breeding, mutation, and crossing considering the evolutionary process of the population. can do.

The step of calculating the optimized LPPL parameters starts with the evolution of the population of individuals from the first initial LPPL parameters and can be changed by temporarily applying the second initial parameters to selection, breeding, mutation, and crossover. In the step of adjusting the hyper parameters that control the evolution strategy of the genetic algorithm, the hyper parameters can be changed using the second initial parameter. Some or all of the second initial parameters can be applied during selection, mutation, and crossover processes.

Figure 5 is a diagram illustrating LPPL parameters calculated by the asset bubble threshold prediction method according to another embodiment of the present invention.

The plurality of LPPL parameters include a price parameter (A), a distance parameter (B), a wave parameter (C), a time parameter (t _c ), a phase parameter (Φ), a frequency parameter (ω), and an exponential parameter (β). do.

The price parameter (A) refers to the price at the critical point, the price higher than the current price. The distance parameter (B) is negative and represents a measure of the distance to the threshold price. B increases in the time before collapse when C is close to 0. The wave parameter (C) controls the size of waves around an exponential trend. Waves are set considering that they cannot overpower the trend. The time parameter (t _c ) refers to the time when collapse is most likely. The phase parameter (Φ) is a parameter related to phase adjustment. The frequency parameter (ω) refers to the frequency of fluctuations during an asset bubble. A value that is too small is an incorrect wave, and a value that is too large is considered random noise. The exponential parameter (β) refers to the exponent of power law growth. A negative number indicates an infinite critical price, a positive number close to 0 indicates no trend, and a positive number close to 1 indicates a low risk rate.

Figure 6 is a diagram illustrating hyper parameters adjusted by the asset bubble threshold prediction method according to another embodiment of the present invention.

Hyperparameters that control the evolution strategy of genetic algorithms include population population options, fitness scaling options, selection options, breeding options, mutation options, crossover options, and stopping criterion options that affect the evolution strategy.

Population options can be set to various data types, and the number of decision variables (solutions) generated can be adjusted. Adaptive scaling options can be set based on rank, proportional, upward, or linear shift. Selection options can be set considering probability distribution selection, uniform selection, roulette selection, tournament selection, elite preservation selection, etc. Breeding options can be set considering the number of elites, the mating rate to the next generation, etc. Mutation options can be set considering Gaussian mutation, uniform distribution mutation, mutation based on a specific value, etc. Crossing options can be set considering random crossing, crossing based on one decision variable, crossing based on two points, crossing by randomly weighting variables, crossing by weighting variables with low objective functions, crossing by arithmetic mean, etc. The stopping criteria option can be set considering the number of households, time, and determined optimal solution criteria.

Figure 7 is a diagram illustrating the optimized LPPL parameters derived by the asset bubble criticality prediction method according to another embodiment of the present invention, and Figure 8 is a diagram illustrating the optimized LPPL parameters derived by the asset bubble criticality prediction method according to another embodiment of the present invention. This figure illustrates the relative error results for bubble threshold detection of the LPPL model applying LPPL parameters.

Through the relative error analysis of the best fit, it can be confirmed that bubble burst cases can be accurately captured in the US index, Japanese index, Hong Kong index, and Chinese index.

Figures 9 to 11 are diagrams illustrating the sensitivity of performance error indicators between LPPL parameters of the LPPL model applied by the asset bubble threshold prediction method according to another embodiment of the present invention.

As a result of confirming the interrelationship between the time parameter (t _c ), phase parameter (Φ), frequency parameter (ω), and exponential parameter (β) through sensitivity analysis of RMSE (Root Mean Squared Error), the frequency parameter (ω) It showed high sensitivity to small fluctuations. This indicates the need to apply a genetic algorithm to the LPPL model. If the initial value of the frequency parameter is close to the local minimum, a local optimization error may occur. However, this problem can be overcome with a genetic algorithm, and the processing speed can be improved by adjusting the hyperparameters of the genetic algorithm.

Figures 12 and 13 are diagrams showing asset bubble threshold prediction accuracy according to simulation results of embodiments of the present invention.

Looking at the simulation results for the U.S. index, Japanese index, Hong Kong index, and China index, peaks within the cycle are detected, weights are applied considering the current point, initial parameters are estimated, and hyperparameters are adjusted using a genetic algorithm. After optimizing the parameters, it can be confirmed that the prediction accuracy of the bubble collapse time predicted through the LPPL model designed based on this is improved. In addition, the speed of processing bubble threshold prediction can be improved by adjusting the hyperparameters of the genetic algorithm.

Figure 14 is a block diagram illustrating a computing device practicing embodiments of the present invention.

Computing device 1010 includes at least one processor 1020, a computer-readable storage medium 1030, and a communication bus 1070.

The processor 1020 may control the computing device 1010 to operate. For example, the processor 1020 may execute one or more programs stored in the computer-readable storage medium 1030. One or more programs may include one or more computer-executable instructions, which, when executed by processor 1020, may be configured to cause computing device 1010 to perform operations according to example embodiments. there is.

The computer-readable storage medium 1030 is configured to store computer-executable instructions or program code, program data, and/or other suitable form of information. The program 1040 stored in the computer-readable storage medium 1030 includes a set of instructions executable by the processor 1020. In one embodiment, computer-readable storage medium 1030 includes memory (volatile memory, such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, It may be flash memory devices, another form of storage medium that can be accessed by computing device 1010 and store desired information, or a suitable combination thereof.

Communication bus 1070 interconnects various other components of computing device 1010, including processor 1020 and computer-readable storage medium 1040.

Computing device 1010 may also include one or more input/output interfaces 1050 and one or more communication interfaces 1060 that provide an interface for one or more input/output devices. The input/output interface 1050 and communication interface 1060 are connected to the communication bus 1070. Input/output devices (not shown) may be connected to other components of computing device 1010 through input/output interface 1050.

In FIG. 2, each process is described as being executed sequentially, but this is only an illustrative explanation, and those skilled in the art can change the order shown in FIG. 2 and execute it without departing from the essential characteristics of the embodiments of the present invention. Alternatively, it may be applied through various modifications and modifications, such as executing one or more processes in parallel or adding other processes.

Operations according to the present embodiments may be implemented in the form of program instructions that can be performed through various computer means and recorded on a computer-readable medium. Computer-readable media refers to any media that participates in providing instructions to a processor for execution. Computer-readable media may include program instructions, data files, data structures, or combinations thereof. For example, there may be magnetic media, optical recording media, memory, etc. A computer program may be distributed over networked computer systems so that computer-readable code can be stored and executed in a distributed manner. Functional programs, codes, and code segments for implementing this embodiment can be easily deduced by programmers in the technical field to which this embodiment belongs.

These embodiments are intended to explain the technical idea of the present embodiment, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

Claims (16)

In the asset bubble criticality prediction method by the asset bubble criticality prediction device,
Defining an LPPL model with a plurality of LPPL parameters according to Log Periodic Power Law (LPPL) theory;
selecting a peak within a circulation period in the LPPL model to calculate an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model;
Calculating optimized LPPL parameters using a genetic algorithm based on the initial LPPL parameters;
adjusting hyper-parameters that control the evolution strategy of the genetic algorithm; and
An asset bubble criticality prediction method comprising calculating the bubble burst point from time series price data through an LPPL model applying the optimized LPPL parameters. According to paragraph 1,
The plurality of LPPL parameters include a price parameter, a distance parameter, a wave parameter, a time parameter, a phase parameter, a frequency parameter, and an exponential parameter,
The LPPL model is an asset bubble threshold prediction, characterized in that it is expressed as a relational expression of the price parameter, the distance parameter, the wave parameter, the time parameter, the phase parameter, the frequency parameter, and the exponential parameter according to the LPPL theory. method. According to paragraph 1,
The step of calculating the initial LPPL parameter is,
Detecting a peak group having a plurality of peaks within the circulation period using the time series price data,
Peak weights are differentially applied to the plurality of peaks in consideration of the current point in time,
Screening a plurality of peaks to which the peak weight is applied and selecting the first peak group with the largest value,
An asset bubble threshold prediction method, characterized in that calculating an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the first peak group. According to paragraph 3,
The step of calculating the optimized LPPL parameters is,
An asset bubble threshold, characterized in that the optimized LPPL parameters are calculated by applying a genetic algorithm to the initial LPPL parameters to derive optimized values through repeated selection, breeding, mutation, and crossover in consideration of the evolutionary process of the population of individuals. Prediction method. According to clause 4,
The step of calculating the initial LPPL parameter is,
Screening a plurality of peaks to which the peak weight is applied and selecting the first peak group with the largest value,
Calculating a first initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the first peak group,
Screening a plurality of peaks to which the peak weight is applied and selecting a second peak group with the second largest value,
An asset bubble threshold prediction method, characterized in that for calculating a second initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the second peak group. According to clause 5,
The step of calculating the optimized LPPL parameters starts the evolution of the population of individuals from the first initial LPPL parameters, and changes the second initial parameters by temporarily applying the selection, the breeding, the mutation, and the crossing,
The step of adjusting the hyper-parameters that control the evolution strategy of the genetic algorithm is an asset bubble threshold prediction method, characterized in that the hyper-parameters are changed using the second initial parameters. According to paragraph 3,
The step of adjusting the hyperparameters that control the evolution strategy of the genetic algorithm includes population population options, fitness scaling options, selection options, breeding options, mutation options, crossover options, stopping criteria options, or these that affect the evolution strategy. An asset bubble criticality prediction method characterized by changing the combination of . In the asset bubble criticality prediction device,
LPPL model defined with multiple LPPL parameters according to Log Periodic Power Law (LPPL) theory;
a first initial parameter calculation unit that selects a peak within a circulation period in the LPPL model and calculates an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model;
a parameter optimization unit that calculates optimized LPPL parameters using a genetic algorithm based on the initial LPPL parameters;
a hyper-parameter control unit that adjusts hyper-parameters that control the evolution strategy of the genetic algorithm; and
An asset bubble criticality prediction device including a bubble burst prediction unit that calculates the bubble burst time from time series price data through the LPPL model to which the optimized LPPL parameters are applied. According to clause 8,
The plurality of LPPL parameters include a price parameter, a distance parameter, a wave parameter, a time parameter, a phase parameter, a frequency parameter, and an exponential parameter,
The LPPL model is an asset bubble threshold prediction, characterized in that it is expressed as a relational expression of the price parameter, the distance parameter, the wave parameter, the time parameter, the phase parameter, the frequency parameter, and the exponential parameter according to the LPPL theory. Device. According to clause 8,
The first initial parameter calculation unit,
Detecting a peak group having a plurality of peaks within the circulation period using the time series price data,
Peak weights are differentially applied to the plurality of peaks in consideration of the current point in time,
Screening a plurality of peaks to which the peak weight is applied and selecting the first peak group with the largest value,
An asset bubble threshold prediction device, characterized in that for calculating an initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the first peak group. According to clause 10,
The parameter optimization unit,
An asset bubble threshold, characterized in that the optimized LPPL parameters are calculated by applying a genetic algorithm to the initial LPPL parameters to derive optimized values through repeated selection, breeding, mutation, and crossover in consideration of the evolutionary process of the population of individuals. Prediction device. According to clause 11,
a second initial parameter calculation unit accessible to the LPPL model; and
It includes a peak group adjustment unit connected to the first initial parameter calculation unit and the second initial parameter calculation unit,
The peak group adjuster screens a plurality of peaks to which the peak weight is applied and selects the first peak group with the largest value,
The first initial parameter calculation unit calculates a first initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the first peak group,
The peak group adjuster screens a plurality of peaks to which the peak weight is applied and selects a second peak group with a second largest value,
The second initial parameter calculation unit calculates a second initial LPPL parameter to be applied to the LPPL parameter of the LPPL model based on the second peak group. According to clause 12,
The parameter optimization unit starts the evolution of a population of individuals from the first initial LPPL parameters and changes the second initial parameters by temporarily applying them to the selection, breeding, mutation, and crossover,
The hyper-parameter control unit is an asset bubble threshold prediction device, characterized in that the hyper-parameter is changed using the second initial parameter. According to clause 8,
The hyper-parameter control unit is an asset bubble characterized in that it changes a population option, a fitness scaling option, a selection option, a breeding option, a mutation option, a crossover option, a stopping criterion option, or a combination thereof that affects the evolution strategy. Critical prediction device. A computer-readable storage medium storing a computer program executable by a computer,
A computer-readable storage medium recording a computer program capable of executing the method according to any one of claims 1 to 7. A computer program recorded on a computer-readable storage medium containing computer program instructions executable by a processor, wherein when the computer program instructions are executed by the processor,
A computer program characterized in that it performs the method according to any one of claims 1 to 7.

Images