JP2003108759A - Simulation method of market information and system therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a simulation method of market information and system therefor for generating more exact risk management information without impairing characteristics of the respective yield curves as shortening calculation time with simple structure. SOLUTION: Simulation is performed based on the market information, in the simulation for generating the risk management information, main component data is generated by analyzing main components of the respective pieces of fluctuated market information by a main component analyzing part 10 and the risk management information is generated by executing full Monte Carlo simulation by approximate values by the yield curves of the respective pieces of fluctuated market information based on the main component data by a full Monte Carlo simulation executing part 20 in the simulation for generating the risk management information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for simulating market information, that is, a simulation method and system for simulating market information to generate risk management information.

[0002]

2. Description of the Related Art As the need for asset management and risk management has increased, a model has been considered in which numerical values such as VaR and EaR representing the risk of financial assets are generated from market information as risk management information. For example, the major VaR models in existence today include the Variance-Covarian method beginning with RiskMetrics.
(ce method), a Monte Carlo method that changes the application from the derivative product model, and a historical method that is an empirical analysis using past data. There are other lattice methods and many improved types and derived types for them, but these three are becoming established as typical methods.

The variance / covariance method has spread rapidly since JP Morgan released the related information under the name RiskMetrics (registered trademark) free of charge as a method of market risk management aimed at compliance with the second BIS regulation. Has become a method. The Monte Carlo method was also introduced in the field of market risk management because of the objection that the variance / covariance method has limitations in handling option trading. or,
The historical method is characterized by empirical analysis using past data, and is the opposite of the above two methods as a VaR calculation method. The Monte Carlo method and the historical method are sometimes referred to as a simulation method by combining the two because of their repeated calculation characteristics.

Among these, in the simulation by the Monte Carlo method, the VaR model handles a few hundred kinds of risk factors in a typical case, and a financial institution handling individual stocks handles about 4000 series of daily data. In that case, a fully adopted system may require several hours of calculation time. Further, in the Monte Carlo method, if the position input is made once, it may be difficult to correct even weekdays because the calculation time is long. Of course, using large-scale hardware will alleviate the problem, but it will be cost-effective. For this reason, recently, a new calculation method has been developed, and the calculation speed has been greatly reduced, and the system has been downsized, and thus has been widely adopted.

One of them is the Δ-Γ-Θ Monte Carlo method. This simple configuration is shown in FIG. Market information entered (risk factor: exchange rate in Fig. 5,
Although interest rates and stock prices are shown, there are actually many more risk factors, such as risk factors for determining interest rates, but they are not shown in Fig. 5 for simplification.) The moment value (Δ, Γ, Θ, ...) Which is calculated by the calculation unit 51 and approximates this portfolio value is calculated by the moment calculation unit 52 and is input to the Monte Carlo simulation execution unit 53. Compared to repeating risk factor calculations, the calculation time is greatly reduced.

[0006]

However, the VaR
In the model, it is necessary to calculate the profit and loss sensitivity of each position for each risk factor in advance, and it is necessary to decompose the risk as required for position data and map it to each risk factor. Most products are decomposed into components such as interest rates, exchange rates, and stock prices, and the two-leg method, which is also described in the BIS document, is a typical example. It is necessary to consider the characteristics of the VaR model together with the knowledge of For example, in the case of interest rates, inter-series correlation (multico) will have an adverse effect if the risk factor is increased.

Therefore, the yield curve assumed in the model by the Monte Carlo method, which is frequently used for calculating VaR, cannot be considered as a yield curve that can be very realistic as shown in FIG. There are also models that have improved on this issue and are used to calculate various derivatives and potential exposures, but they are not well equipped and inflexible and lack the vital and vital objectivity for risk management models. Therefore, it is rarely adopted as a VaR model.

In view of the above-mentioned conventional problems, the present invention is a market information for generating more accurate risk management information without spoiling the characteristics of each yield curve while significantly reducing the calculation time with a simple structure. The simulation method and system thereof are provided.

[0009]

In order to solve this problem, the simulation method of the present invention is a simulation method for generating risk management information by performing simulation on the basis of market information, and changing each market information. Principal component analysis is performed to generate principal component data, and full Monte Carlo simulation is performed by the approximate value of the yield curve of each fluctuating market information based on the principal component data to generate risk management information. Is characterized by. Here, the market information includes exchange rates, interest rates, and stock prices. Further, the principal component data includes at least first to third moment data.

Further, the simulation system of the present invention is a simulation system for performing simulation based on market information to generate risk management information, and performing principal component analysis of each fluctuating market information to obtain the principal component. Principal component analysis means for generating data, and full Monte Carlo simulation means for generating risk management information by executing full Monte Carlo simulation with an approximate value of the yield curve of each of the varying market information based on the principal component data. It is characterized by having. Here, the market information is
Including exchange rates, interest rates, and stock prices. Further, the principal component analyzing means generates principal component data including at least first to third moment data.

Further, the storage medium of the present invention is a storage medium for storing a simulation program for simulating based on market information and generating risk management information in a computer-readable manner, wherein the simulation program changes. The principal component analysis of each market information is performed to generate the principal component data, and the full Monte Carlo simulation is performed by the approximate value of the yield curve of each fluctuating market information based on the principal component data. And a full Monte Carlo simulation module for generating risk management information.

[0012]

DETAILED DESCRIPTION OF THE INVENTION Referring to the accompanying drawings,
An embodiment of the present invention will be described.

<Example of Configuration of Monte Carlo Simulation System of Present Embodiment> FIG. 1 is a block diagram of the Monte Carlo simulation system of the present embodiment.

In FIG. 1, a principal component analysis unit 10 performs principal component analysis on each market information (risk factor). Principal component analysis is a method of synthesizing (compressing) several factors having a correlation to obtain some components and obtaining the total strength and characteristics thereof. In Figure 1,
For example, the characteristic of a risk factor having a correlation is comprehensively obtained from each input of a plurality of input exchange rates, interest rates, stock prices, etc. However, the present invention is not limited to this, and as an example of principal component analysis in asset management, there is one that calculates the moment of a risk factor). Regarding the method of selecting variables in principal component analysis, there are various papers on which basic rules have been studied (for example, Yuichi Mori: Okayama University of Science), and refer to them.

Reference numeral 20 denotes a full Monte Carlo simulation execution unit that performs full Monte Carlo simulation based on the principal component data generated by the principal component analysis unit 10. Full Monte Carlo simulation will not be described in detail.

In the simulation with the configuration of FIG. 1, the risk factor of each grid point is extracted from the yield curve far from the actual one as shown in FIG. More accurate risk management information (VaR and E
aR, etc.) can be generated. In addition, the newly selected principal component by the principal component analysis has a small correlation with other data (those having a correlation are combined into one), and as a result, the correlated fluctuation of the calculation result is suppressed. Therefore, the value converges faster and the calculation time is shortened.

FIG. 2 is a diagram showing a hardware configuration example of a computer for realizing the Monte Carlo simulation of the present embodiment.

Reference numeral 21 is a CPU for arithmetic / control, and 22 is a CP.
A ROM that executes a fixed program executed by U21 and fixed parameters, and a RAM 23 that is used for temporarily storing the processing program of the CPU21 and data. In this example, a principal component analysis module and a full Monte Carlo simulation module are provided. Program area 23a to be stored
And a data area 23b having a principal component analysis value storage unit, VaR and EaR calculation value storage unit, and the like. still,
The program may be read into the RAM 23 from the external storage unit 24 such as a phorropy or a CD-ROM and executed.

Reference numeral 24 is, for example, a floppies or a CD-ROM.
An external storage unit such as the above stores each data used in the simulation, various market data in this example, as a database 24a, and stores a program downloaded and executed in the RAM 23 in the program area 24b. . Reference numeral 25 denotes an input interface, which interfaces the input from the pointing device such as the mouse 25a and the data input keyboard 215b to the system. An output interface 26 is used for displaying data from the system on the display unit 26a and for printing on the printer 26b. Reference numeral 27 is a communication unit for communication connection with the outside.

Although the example of FIG. 2 shows an example in which the system of the present embodiment is configured by a general-purpose personal computer, the simulation executed by this system is highly dimensional and can be repeated at least tens of thousands of times. Since it is desired to process the calculation in real time, it is preferable to configure a dedicated computer system. For example, one in which a plurality of computers are connected in parallel to perform parallel processing and one in which a plurality of CPUs are provided in one computer to perform parallel processing are conceivable.

<Example of Software Processing of Monte Carlo Simulation System of Present Embodiment> FIG. 3 is a diagram showing an outline of an example of software processing of the Monte Carlo simulation system of the present embodiment.

First, in step S31, market information (risk factors) necessary for the simulation executed in this system is collected. In step S32, a principal component analysis is performed on this risk factor independently or while merging a plurality of risk factors to generate a more uncorrelated input variable. Next, step S3
In step 3, full Monte Carlo processing is executed from the variables input based on the generated principal component data. Step S
At 34, VaR, EaR, etc. calculated by the full Monte Carlo process are output.

FIG. 4 is an example of an internal yield curve based on the principal component analysis, which is not a data that is far from the actual state as shown in FIG. 6 but a value that accurately reflects the actual state. Therefore, a more accurate and fast convergence full Monte Carlo is realized.

[0024]

According to the present invention, a market information simulation method and system for generating more accurate risk management information without compromising the characteristics of each yield curve while significantly reducing the calculation time with a simple configuration. Can be provided.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a Monte Carlo simulation system according to the present embodiment.

FIG. 2 is a block diagram showing an example of a hardware configuration for realizing the Monte Carlo simulation system of the present embodiment.

FIG. 3 is a flowchart showing an example of simulation processing of the Monte Carlo simulation system according to the present embodiment.

FIG. 4 is a diagram showing an example of an internal yield curve according to the present embodiment.

FIG. 5 is a block diagram of a conventional Δ-Γ-Θ Monte Carlo simulation system.

FIG. 6 is a diagram showing an example of a conventional internal yield curve.

Claims (7)

[Claims] 1. A simulation method for generating risk management information by performing simulation based on market information, wherein principal component data is generated by performing principal component analysis of each fluctuating market information, and the principal component data is generated. A simulation method, characterized in that risk management information is generated by executing a full Monte Carlo simulation with an approximate value of a yield curve of each of the fluctuating market information based on the above. 2. The simulation method according to claim 1, wherein the market information includes exchange rates, interest rates, and stock prices. 3. The simulation method according to claim 1, wherein the principal component data includes at least first to third moment data. 4. A simulation system for performing simulation based on market information to generate risk management information, wherein principal component analysis is performed for principal component analysis of each fluctuating market information to generate principal component data. And a full Monte Carlo simulation means for generating risk management information by executing a full Monte Carlo simulation with an approximate value of the yield curve of each of the changing market information based on the principal component data. ·system. 5. The simulation system according to claim 4, wherein the market information includes exchange rates, interest rates, and stock prices. 6. The simulation system according to claim 4, wherein the principal component analysis means generates principal component data including at least first to third moment data. 7. A storage medium for computer-readable storage of a simulation program for simulating market information and generating risk management information, wherein the simulation program is a main component of each fluctuating market information. A principal component analysis module that performs an analysis to generate principal component data, and a full Monte Carlo simulation is performed by an approximate value of the yield curve of each of the changing market information based on the principal component data to generate risk management information. And a full Monte Carlo simulation module.