JP5579051B2 - Information processing apparatus, information processing method, and program - Google Patents

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program.

Banks are exposed to interest rate risk due to their high share of financial assets and high external funding. Basel II requires banks to develop a banking account interest rate risk management system, and the FSA's “overall supervision guidelines” include a total of 20 capital in assets and liabilities due to interest rate shocks and changes in interest rates. Banks where economic value declines by more than 50% will be defined as outlier banks, and the relevant banks will receive orders from the authorities for business improvement as necessary. For this reason, banks need to perform risk management such as ALM (Asset Liability Management) appropriately.
Indicators for measuring interest rate risk generally include duration and outlier ratio. When calculating these, it is necessary to predict the cash flows of financial assets and liabilities.

Liquidity deposits are the primary means of raising funds that account for the bulk of bank liabilities, and the interest rate risk held by banks will be measured differently depending on the method for predicting the cash flow of liquid deposits. The impact of this on the liquidity deposits is important.
However, liquid deposits are not fixed in terms of maturity like fixed deposits and bonds, and are exposed to the risk that the balance will change in response to depositor requests. Balance fluctuations include seasonal factors such as pension payments and bonus payments, macro factors, and deposits and withdrawals due to individual depositor factors. The portion of the liquid deposit that remains in the bank for a long time without being withdrawn is called the core deposit. Since not only the balance but also the deposit interest rate is not fixed, it is difficult to estimate cash flow, and it is difficult to predict the cash flow of liquid deposits appropriately.

In addition, the FSA's “Overall Supervisory Guidelines” provide a reasonable model for the definition of core deposits for the bank's internal management to model depositor behavior, and to appropriately identify the core deposit amount and assign it to the due date. In the case of implementation, the guideline indicates that the definition should be followed, and various approximate core deposit management models have been proposed.
As one of them, a core deposit internal management model called AA-Kijima model is disclosed (see Non-Patent Document 1). The AA-Kijima model is a model that estimates the parameters by the regime switching method based on the balance transition, and assumes that the deposit balance moves in geometric Brownian motion (the deposit balance follows a lognormal distribution). VaR (Volume at Risk) Is calculated.
Regime switching method (when used for analysis of liquid deposits) is considered that there are several trends (phases) in the transition of deposit balance, and the rate of change in balance for each trend is determined from past deposit balance transition data. It is a technique to seek. As phases, balance rising period (time when deposit balance was rising significantly), balance stable period (time when deposit balance was almost flat or slightly increasing / decreasing), balance falling period (deposit balance was falling) Time).

  VaR is an index that indicates a conservative estimate of the balance at a future time, and means the level of the deposit balance such that the probability that the deposit balance falls below a certain level is 100 (1-α)%. It is an index calculated based on the same concept as Value at Risk. In other words, VaR is the worst deposit balance that can occur at a confidence level of 100α% when considering a stress scenario in which core deposits continue to flow out. VaR is calculated by using the balance change rate μ and volatility σ as the worst deposit balance at each time point such as one year, two years later, and the like. Then, the maturity ladder is developed based on the calculated VaR, and the outlier ratio is calculated.

“Internal model for bank account interest rate risk management (AA-Kijima Model)”, Securities Analyst Journal, P79-92, April 2007

  However, in the regime switching method, the parameters (μ, σ) are not estimated in a form associated with the explanatory variables, and therefore, it is not easy to use in other banking practices except for the purpose of calculating the risk index.

  The present invention has been made in view of such a problem, and an object thereof is to predict a cash flow more appropriately by estimating a parameter in a form associated with an explanatory variable.

  Therefore, an information processing apparatus according to the present invention is an information processing apparatus having a storage unit, and includes deposit balance information indicating a value of a deposit balance provided for each date information, and a plurality of types of variables provided for each date information. Reading means for reading explanatory variable information indicating the value of each variable and explanatory variable definition information defining a plurality of types of variables from the storage unit, calculating a balance change rate of the deposit balance information, and using date information as a key Generating means for generating integrated information obtained by integrating the calculated balance change rate and the explanatory variable information; generating a plurality of conditions including variables defined in the explanatory variable definition information; and Compare the explanatory variable information of the information, classify the balance change rate of the integrated information into each node corresponding to each of the plurality of conditions, and information on the balance change rate belonging to the node and the condition corresponding to the node And having a classification means for storing in the storage unit in association with.

  Here, the “reading unit” corresponds to, for example, a data reading unit 21 described later. The “generating unit” corresponds to, for example, an analysis data creating unit 22 described later. The “classification unit” corresponds to, for example, a balance change rate classification unit 23 described later.

  According to the present invention, it is possible to predict the cash flow more appropriately by estimating the parameter in association with the explanatory variable.

It is a figure which shows an example of a liquid deposit risk management system. It is a figure which shows an example of the hardware constitutions of information processing apparatus. It is a figure which shows an example of a function structure of information processing apparatus. It is a figure which shows an example of the flowchart which concerns on the process in information processing apparatus. It is a figure which shows an example of the item added after the item of a time series data, and a process. It is a figure which shows an example of the item added after the item of a time series data, and a process. It is a figure which shows an example of the item added after the item of a time series data, and a process. It is a figure which shows an example of the data for analysis. It is a figure which shows an example of the actual value of the data for analysis of a balance change rate. It is a figure which shows an example of the actual value of the data for volatility analysis. It is a figure which shows an example of the method of determining the conditions corresponding to the node of a tree. It is a figure which shows an example of explanatory variable definition data. It is a figure which shows an example of the screen displayed on a display. It is a figure which shows an example of VaR. It is a figure which shows an example of a maturity ladder. It is a figure which shows an example of exposure. It is a figure which shows an example of input data. It is a figure which shows an example of a calculation result. It is a figure which shows an example of a calculation result. It is a figure which shows an example of market interest rate data and deposit interest rate data. It is a figure which shows an example of a calculation result. It is a figure which shows an example regarding the content in connection with prediction of a deposit balance. It is a figure which shows an example of a tree. It is a figure which shows an example of a processing program. It is a figure which shows an example of the calculation result of deposit balance prediction.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram illustrating an example of a liquid deposit risk management system according to the present embodiment. In the liquid deposit risk management system, parameters that can be used for a plurality of types of applications (calculation of risk index for liquid deposits, prediction of deposit balance, etc.) are calculated, stored, and managed. The liquid deposit risk management system includes an information processing device 100 and a storage device 200. The information processing apparatus 100 and the storage device 200 are connected to be communicable via the network 300.
The information processing apparatus 100 is an example of a computer, acquires various types of data from the storage device 200, reads various types of data from the recording medium 400, and uses the acquired data and the read data as a risk index for liquid deposits (described later). Various parameters for calculating the duration, outlier ratio, etc.).
The storage device 200 is an example of a computer, and is a server in a bank or a server such as the Bank of Japan, Ministry of Economy, Trade and Industry, Ministry of Internal Affairs and Communications, and stores various data. In FIG. 1, only one storage device 200 is shown for convenience of explanation, but a plurality of storage devices 200 may be provided.

  FIG. 2 is a diagram illustrating an example of a hardware configuration of the information processing apparatus 100. The information processing apparatus 100 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, an HD (Hard Disk) 14, an input device 15, a display device 16, and a recording medium drive device 17. And the interface device 18.

  The CPU 11 reads out a program from the ROM 12 or the HD 14 as necessary, stores the program in the RAM 13, and executes the program, thereby providing a part of a function to be described later or executing a process shown in a flowchart to be described later. Or The ROM 12 stores a program that is read first when the information processing apparatus 100 is powered on. The RAM 13 is a main memory of the information processing apparatus 100 and temporarily stores various data used by the CPU 11. The HD 14 stores various data such as a table.

  The input device 15 is configured to include a keyboard, a mouse, and the like operated by an operator (or user) of the information processing apparatus 100, and various kinds of information are input to the information processing apparatus 100 in response to an operation (user operation) of the keyboard, mouse, and the like. Enter operation information. The display device 16 includes a display or the like used by the user of the information processing device 100, and displays various information (or screens) and the like on the display. The recording medium drive device 17 reads and writes data on the recording medium 400 inserted into the recording medium drive device. The interface device 18 connects the information processing device 100 to the network 300 or the like.

Note that various types of data such as the above-described program may be provided to the information processing apparatus 100 via a recording medium 400 such as a CD-ROM, or may be downloaded via the network 300 or the like. For example, when a program is provided from the recording medium 400, the program is stored in the HD 14 from the recording medium 400 via the recording medium drive device 17 in response to a user operation that instructs installation of the program.
In addition, the storage device 200 has the same hardware configuration as the information processing device 100.

FIG. 3 is a diagram illustrating an example of a functional configuration of the information processing apparatus 100. In FIG. 3, an overview of the functions of the information processing apparatus 100 will be described, and details thereof will be described with reference to FIG.
The data reading unit 21 reads time series data (an example of explanatory variable information) including deposit balance data (an example of deposit balance information) from the HD 14. The HD 14 stores time-series data acquired from the storage device 200 and various data acquired and stored from the recording medium 400 in advance.
The analysis data creation unit 22 calculates the balance change rate indicating the change rate of the deposit balance from the time series data read by the data reading unit 21 based on the deposit balance data included in the time series data. Based on the balance change rate, the volatility indicating the ratio of the fluctuation range of the deposit balance is calculated, and analysis data (analysis data which is an example of integrated information) is created (generated).

The balance change rate classification unit 23 constructs a tree such as a decision tree or CART explaining the balance change rate of the analysis data. The volatility classification unit 24 constructs a tree such as a decision tree or CART that explains the volatility of the analysis data.
The balance change rate determining unit 25 uses a parameter (stress change rate μs at the time of stress) used to calculate a deposit balance VaRα (t) indicating a prospect for the minimum level of the deposit balance based on the constructed balance change rate tree. ). The volatility determination unit 26 determines a parameter (volatility σs under stress) used for calculation of VaRα (t) based on the constructed volatility tree.

The VaRα (t) calculation unit 27 calculates the deposit balance VaRα (t) at the time of stress using the parameters (μs, σs) determined by the balance change rate determination unit 25 and the volatility determination unit 26.
The maturity ladder calculation unit 28 calculates a maturity ladder X _i that indicates the amount of cash flow that flows out for each term. The exposure calculation unit 29 calculates an exposure Y _i indicating an amount obtained by adjusting the maturity ladder X _i in consideration of the following rate (β) of the deposit interest rate to the market interest rate.
The risk index calculation unit 30 calculates the risk index of liquid deposits such as average duration D and outlier ratio.
The deposit balance prediction unit 31 outputs information for predicting the deposit balance based on the constructed balance change rate tree and the latest time-series data.

FIG. 4 is a diagram illustrating an example of a flowchart relating to processing in the information processing apparatus 100.
First, the data reading unit 21 reads various data stored in the HD 14 (SQ10, SQ12). Here, the HD 14 includes time series data (1) acquired from a bank's core system and the like, and time series data (2 ) And are stored.
For example, the time series data (1) is deposit balance data for each category such as deposit type, depositor, amount hierarchy, and the like. In addition, for example, time series data (2) includes deposit interest rate (including interest rates such as fixed deposits) data, market data (competed rate data, unsupported rate data, etc.), macro index data (industrial production index data). , GDP data, etc.). Each time series data is data for each date information such as day, week, month, quarter, year.
The HD 14 stores explanatory variable definition data (an example of explanatory variable definition information), which will be described later, acquired from the recording medium 400 or downloaded via the network 300.

ここで、図５Ａに、時系列データ（１）の項目の一例と、ＳＱ１４の処理の後に追加される項目の一例とを示す。 The analysis data creation unit 22 creates analysis data from various data read by the data reading unit 21 (SQ14 to SQ24).
In SQ14, the analysis data creation unit 22 calculates the balance change rate R _i based on the deposit balance data (V ₀ ,..., V _N ) included in the time series data (1) as the explained variable data. Calculate using the following formula. Note that the calculated balance change rate R _i (balance change rate data) is held in a storage device (an example of a storage unit) such as the RAM 13 or the HD 14.

Here, FIG. 5A shows an example of items of time-series data (1) and an example of items added after the processing of SQ14.

ここで、ｍは、ボラティリティσｊの推定に用いるデータ数である。
なお、図５Ａでは残高変化率の項目が追加されているが、ここでは、残高変化率の項目に代えて、ボラティリティの項目が追加される。また、ここでの計算で追加される項目の例については図示を省略する。 The analysis data creation unit 22 reads the balance change rate R _i held in the storage device as the explained variable data, and calculates the volatility σj using the following equation. The calculated volatility σj (volatility data) is stored in the storage device.

Here, m is the number of data used for estimating the volatility σj.
In FIG. 5A, a balance change rate item is added, but here, a volatility item is added instead of the balance change rate item. Further, illustration of items added in the calculation here is omitted.

  Here, the analysis data creation unit 22 may perform a seasonal adjustment process on the value calculated as described above (SQ16). As the seasonal adjustment process, for example, a general processing program “X-12-ARIMA” can be employed.

In SQ18, the analysis data creation unit 22 uses, as explanatory variable data, the deposit type (current / ordinary / savings / notification / extra tax payment) included in the time-series data (1), depositor (corporate / individual / public money, etc.) ) Calculate interest rate difference, rate of change, rate of change, etc. from data such as monetary hierarchy and data base month. For example, the interest rate difference is a value obtained by subtracting the ordinary deposit interest rate from the fixed deposit interest rate, and the change range is a value obtained by subtracting the ordinary interest rate at Tn−1 from the ordinary interest rate at the reference date Tn. Details are omitted. The calculated value (explanatory variable A data) is stored in the storage device.
Here, FIG. 5B shows an example of items of time-series data (1) and an example of items added after the processing of SQ18.

In SQ20, the analysis data creation unit 22 uses the interest rate difference from the deposit interest rate (including interest rates such as time deposits) data, market data, macro index data, etc. included in the time series data (2) as the explanatory variable data. , Change rate, change width, etc. are calculated. The calculated value (explanatory variable B data) is held in the storage device.
FIG. 5C shows an example of items of time-series data (2) and an example of items added after the processing of SQ20.
Here, the analysis data creation unit 22 may perform a seasonal adjustment process on the value (explanatory variable B data) calculated as described above (SQ22).
Note that the order of processing is not limited to the order described above, and need not be fixed.

In SQ24, the analysis data creation unit 22 reads the explained variable data, the explained variable A data, and the explained variable B data from the storage device, and reads the explained variable data, the explained variable A data, and the explained variable B data as the reference date, Analytical data is created by integrating with the base month as a key. The generated analysis data is stored in the storage device.
FIG. 6 shows an example of analysis data. In this example, the balance change rate analysis data is shown, but the volatility analysis data is the same. FIG. 7 shows an example of actual values of balance change rate analysis data, and FIG. 8 shows an example of actual values of volatility analysis data.

In SQ26, the balance change rate classification unit 23 reads the analysis data and the explanatory variable definition data from the storage device, and creates a tree that explains the balance change rate included in the analysis data.
An outline of tree creation will be described with reference to FIGS. 9 and 10. FIG. 9 is a diagram illustrating an example of a method for determining a condition corresponding to a node of a tree. FIG. 10 is a diagram illustrating an example of explanatory variable definition data. In the explanatory variable definition data, a plurality of types of variables are defined.
First, the balance change rate classifying unit 23 does not satisfy the condition 1 distribution and the balance change rate distribution that satisfies the condition 1 (the combination of the first item of the explanatory variable definition data and the first value of the first item). The balance change rate distribution of the condition (not condition 1) is obtained. Subsequently, the balance change rate classification unit 23 calculates a square error SE (square error) of both distributions using the following equation.

ｍｋ：ターミナルノードｋに該当するデータ数
Ｋ：ターミナルノードの数
Ｒｉ：ターミナルノードｋに該当するデータｉにおける残高変化率

mk: number of data corresponding to terminal node k K: number of terminal nodes Ri: balance change rate in data i corresponding to terminal node k

  In the present embodiment, a description will be given using a square error, but an absolute deviation may be employed instead of the square error. In this case, the balance change rate classification unit 23 calculates the absolute error AD of the two distributions using the following equation.

ｍｋ：ターミナルノードｋに該当するデータ数
Ｋ：ターミナルノードの数
Ｒｉ：ターミナルノードｋに該当するデータｉにおける残高変化率

mk: number of data corresponding to terminal node k K: number of terminal nodes Ri: balance change rate in data i corresponding to terminal node k

Subsequently, the balance change rate classification unit 23 satisfies the condition 2 and the distribution of balance change rates that satisfy the condition 2 (the combination of the first item of the explanatory variable definition data and the second value of the first item). The balance change rate distribution under no condition (not condition 2) is obtained. Subsequently, the balance change rate classification unit 23 similarly calculates a square error between the distributions of the two.
Then, the balance change rate classification unit 23 sets all values (balance change rate data) belonging to each of the first to nth items for all conditions (first to nth items of the explanatory variable definition data). When the square error is calculated for all of the two groups, the conditions (condition A and not condition A) that minimize the square error are specified. Based on this, the balance change rate classification unit 23 generates a node corresponding to the condition A and a node corresponding to the not condition A.

That is, the balance change rate classification unit 23 calculates the square error by grouping the data belonging to the variable name into two groups for all the variable names in FIG. 10 and for each variable name. There are a plurality of grouping conditions for each variable name, but the balance change rate classification unit 23 calculates a square error for all groupings (that is, all conditions), and the calculated squared error is minimized. Is identified.
When the variable name indicates a categorical variable, the balance change rate classification unit 23 groups the data into two based on a combination (combination) of data included in the category (an example of a variable name “MONTH” described later) Corresponding). When the variable name indicates a continuous variable, the balance change rate classification unit 23 groups the data into two groups using any value belonging to the variable name as a threshold (a variable name “ordinary deposit interest rate” described later). corresponding to the example of “t”).

Further, the balance change rate classification unit 23 repeats specifying the condition in the same manner, and further divides the terminal node corresponding to the condition A or the node corresponding to the not condition A into two distributions. Then, the conditions (condition B and not condition B) that minimize the square error are specified. Based on that, the balance change rate classification unit 23 generates, for example, two nodes (a node corresponding to the condition B and a node corresponding to the not condition B) from the nodes corresponding to the condition A.
In this way, the balance change rate classification unit 23 repeats specification of conditions and generation of nodes, and when it is determined that the number of terminal nodes exceeds a certain value, the processing for specifying conditions is terminated. That is, the balance change rate classification unit 23 generates two nodes from the node corresponding to the condition A to calculate the square error, and generates two nodes from the node corresponding to the not condition A to generate the square error. After calculation, a condition (condition B and not condition B) that minimizes the square error is specified, and a node corresponding to the condition B and a node corresponding to the not condition B are selected from the nodes having the smallest square error. And generate

  That is, the balance change rate classification unit 23 repeatedly searches for a condition and a not condition in which the two distributions differ most significantly when the condition is met and when the condition is not met (that is, the square of the two distributions). A plurality of conditions are generated by repeatedly searching for a condition and a not condition when the error is minimized or the absolute error of the two distributions is minimized. Then, the balance change rate classification unit 23 compares the generated plurality of conditions with the time series data of the analysis data, and sequentially classifies the balance change rate of the analysis data to each node corresponding to each of the plurality of conditions. Then, information on the balance change rate belonging to the node (average of the explained variable in the data belonging to the node, standard deviation, number of data, etc.) and the condition corresponding to this node are set as node data (node information), and each node data is By storing the associated tree information in the storage device, a tree explaining the balance change rate is created. Note that the method of storing is not limited to the above-described method. For example, the node data may be stored in the storage device in the order of generation (for each node).

FIG. 11 shows an example of the screen (output screen 500) displayed on the display showing the processing result (tree information and the like) by the balance change rate classification unit 23.
For example, the terminal node 4 has “industrial production index t−1 <= 100.75”, “interest rate difference 3t−1> 0.04”, “Δ time deposit interest rate 1> −0.12”, “time deposit interest rate”. 7t-1 <= 1.78 ”and“ Δ the next day end of month t-1> −0.20 ”.

Here, regarding the search for the first condition, the balance change rate classification unit 23 first reads the variable name “MONTH” of the item “1” with reference to the explanatory variable definition data. The type of the variable name “MONTH” is a categorical variable, and the possible values are “January” to “December”. A categorical variable is a flag that represents an attribute of data, and even if it is expressed as a numerical value in the form, it is a variable that has no particular meaning in the magnitude relationship. Therefore, the balance change rate classification unit 23 performs “January”, “February”,..., “December”, “January + February”, “January + March”,. + December ”,“ January + February + March ”,..., All combinations of conditions and not conditions are generated, and the square error of the balance change rate distribution is calculated for all conditions and not conditions. Similarly, the square error of each condition is calculated for the variable name “QUATER” of the item “2”.
Then, the balance change rate classification unit 23 reads the variable name “ordinary deposit interest rate t” of the item “3” with reference to the explanatory variable definition data. The type of the variable name “ordinary deposit interest rate t” is a continuous variable, and possible values are each data of the ordinary deposit interest rate 1 included in the analysis data of the deposit balance change rate (“0.1” shown in FIG. "0.092", "0.08", etc.). Therefore, the balance change rate classification unit 23 uses all values of “0.1”, “0.092”, “0.08”... As threshold values, and is below or above the threshold value (or below or above the threshold value). The conditions and not conditions are generated by the above, and the square error of the distribution of the balance change rate is calculated for all the conditions and not conditions.

In this way, the balance change rate classification unit 23 calculates the square of the balance change rate distribution for the conditions and not conditions of all the items of the explanatory variable definition data from the item “1” to the item “116” shown in FIG. When the error is calculated, the condition in which the square error becomes the smallest and the not condition (in this example, the industrial production index t-1 <= “100.75” and the industrial production index t-1> “100.75” ).
Then, the balance change rate classifying unit 23 obtains, for each node, a standard deviation indicating how much the explanatory variable varies in the number of nodes, an average of the balance change rate, and a data number N indicating the number of data. It is stored in the storage device as tree information.
Here, the balance change rate classification unit 23 calculates an average and a standard deviation using the following equations.

ｍｋ：あるノードｋに属するデータ数
Ｒｉ：あるノードｋに属するデータｉにおける残高変化率

mk: number of data belonging to a certain node k Ri: balance change rate in data i belonging to a certain node k

For example, in the output screen 500, when “industrial production index t−1 <= 100.75” (node 2), the average balance change rate is “0.005 (0.5%)”, and the standard deviation is It is indicated that “0.011” and the number of data N is “97”.
In other words, when the value of the industrial production index one month ago exceeds “100.75” (the so-called threshold) and when it does not, it shows that the situation of changes in deposit balances is the most different, When the threshold value is exceeded, it is indicated that the rate of change in the balance is often in the range of “−0.4% ± 0.3%”.

  Next, the balance change rate classification unit 23 refers to the explanatory variable definition data, and repeats the same processing for data that satisfies the searched condition. That is, in the example of FIG. 11, the balance change rate classification unit 23 refers to the explanatory variable definition data for 97 data included in the node 2, and performs the same process, that is, the item “1” illustrated in FIG. ”To“ 116 ”for all item conditions and not conditions of the explanatory variable definition data, when calculating the square error of the distribution of balance change rate, the condition and not condition that minimizes the square error are specified. This process is repeated sequentially. Thus, in the example of FIG. 11, the interest rate difference 3t-1> “0.04”, Δtime deposit interest rate 1> “− 0.12”, and time deposit interest rate 7t−1 <= “1.78”. Each condition is identified sequentially.

Note that even a variable once used as a condition may be used as a condition with another threshold or combination (combination) to cause a node to branch. In this manner, the balance change rate classification unit 23 stores the tree information including the balance change rate information, the condition information, and the like in the storage device, and creates a tree that explains the balance change rate.
As a result, what kind of value the balance change rate can take in what kind of condition, what corresponds to the aspect in the regime switching method is as a node of the tree as data in a format associated with the explanatory variable Stored.

  Moreover, in SQ26, the volatility classification part 24 performs the process similar to the balance change rate classification part 23 about volatility. Then, the volatility classification unit 24 creates a tree explaining the volatility, and information on the tree including volatility information (average of the explained variable in the data belonging to the node, standard deviation, number of data, etc.), condition information, and the like. Is stored in the storage device.

Here, in the model in this field, there is a strong demand for using a deposit interest rate or the like as an explanatory variable in calculating a parameter (balance change rate μ, volatility σ, etc.). For example, the deposit interest rate is generally constant, but sometimes changes greatly (jumps) with changes in the policy interest rate, so it is difficult to calculate parameters appropriately in regression analysis or the like. A similar problem occurs when categorical variables such as depositors and reference months are used as explanatory variables. Therefore, in the present embodiment, a configuration is employed in which parameters are calculated by generating a tree (decision tree) that can be appropriately handled even when the relationship between the explanatory variable and the explained variable is not necessarily linear.
In addition, according to the tree created as described above, under what conditions (situations), it becomes possible to model from a practical point of view whether the deposit balance falls or the volatility increases. . For example, it becomes possible to obtain the relationship between the product balance, depositor, amount hierarchy, interest rate (interest rate difference), seasonality, macro index, and the deposit balance.
In other words, the explanatory variable data and the explanatory variable definition data are more significant because the user's pre-selected data can be used as appropriate.

In SQ28, the balance change rate determination unit 25 determines the worst case value of the balance change rate from the information of the balance change rate tree. More specifically, the balance change rate determination unit 25 calculates, based on the node data of each node, for example, the negative value of the average value of the balance change rate of the node having the largest absolute value of the average value of the balance change rate of each node. It is determined as the balance change rate μs during stress. That is, the balance change rate determination unit 25 obtains the balance change rate μs during stress using the following formula based on the tree information. In the example of FIG. 11, “−6.8% (average of balance change rate of terminal node 2)” is determined.

In SQ28, the volatility determination unit 26 obtains the worst-case value (volatility σs at the time of stress) from the following equation, for example.

In SQ30, the VaRα (t) calculation unit 27 calculates the deposit balance VaRα (t) at the time of stress (VaR calculation process). FIG. 12A shows an example of the relationship between the deposit balance, time, and VaRα (t).
First, the VaRα (t) calculation unit 27 develops a date t that can be used for calculation of cash flow or the like from the input reference date to the expiration date of the output data. That is, the VaRα (t) calculation unit 27 sets Δt = T _{i + 1} −T _i so as to follow the output data interval (input output data interval).

ここで、ひと月を１／１２年とすると、Ｔ₀は、「０」（基準日）であり、Ｔ_Nは、出力データの満期（年）であるので、Ｔ_iは、基準日からの期間（年率）で表される。
図１３に、ＶａＲ算出処理などで用いられる予め入力される入力データの一例を示す。本実施形態では、入力装置１５、記録媒体ドライブ装置１７、インターフェース装置１８等を介して、基準日、信頼水準、入力データ間隔、出力データ間隔、出力データ満期の情報が予め記憶デバイスに格納されている。

Here, if a month is 1/12, T ₀ is “0” (reference date), and T _N is the maturity (year) of the output data, so T _i is the period from the reference date. (Annual rate)
FIG. 13 shows an example of input data input in advance used in VaR calculation processing or the like. In the present embodiment, information on the reference date, the confidence level, the input data interval, the output data interval, and the output data expiration date is stored in advance in the storage device via the input device 15, the recording medium drive device 17, the interface device 18, and the like. Yes.

Ｖ₀：基準日残高
μｓ：ストレス時の残高変化率
σｓ：ストレス時のボラティリティ Next, the VaRα (t) calculation unit 27 reads the parameters (μs, σs) calculated in SQ28, values input in advance, and the like from the storage device, and calculates VaRα (t). Here, when the deposit balance changes according to the geometric Brownian motion, the deposit balance V _{t at} time t is considered to follow a probability distribution of the following equation (where εt follows a normal distribution with an expected value of 0 and a variance t). It is a random variable.)

V ₀ : Balance on the reference day μs: Rate of change in balance during stress σs: Volatility during stress

Φ^-1（ｚ）：標準正規分布累積確率の逆関数 Then, the VaRα (t) calculation unit 27 calculates VaRα (t) of the confidence level α by the following formula based on the above formula.

Φ ^-1 (z): Inverse function of standard normal cumulative probability

FIG. 14 shows an example of a calculation result by the VaRα (t) calculation unit 27. Note that the calculation result by the VaRα (t) calculation unit 27 is stored in the storage device.
In this example, VaRα (t) calculated from the reference date “2010/9/30” to the output data maturity “10 years (2020/9/30)” for each output data interval “monthly (one month)”. The value of is shown. Note that the items (items 1 to 5) shown in FIG. 14 may be configured to use the items of analysis data as they are, or may be configured to use items specified by the user among the items of analysis data.

In SQ32, maturation utility ladder calculation unit 28 calculates the maturation utility ladder X _i (Machu utility ladder calculation process). FIG. 12B shows an example of the relationship between the deposit balance, time, VaRα (t), and the maturity ladder X _i . The maturity ladder X _i corresponds to a decrease (cash out flow) of the deposit balance, and indicates a difference of VaRα (t) for each term.
More specifically, the maturity ladder calculating unit 28 calculates the maturity ladder X _i using the following equation based on VaRα (t).

  FIG. 15 shows an example of a calculation result by the maturity ladder calculation unit 28. The calculation result by the maturity ladder calculating unit 28 is stored in the storage device.

In SQ32, the exposure calculation unit 29 calculates an exposure Y _i (exposure calculation process). FIG. 12C shows an example of the relationship between the maturity ladder X _i , the time, and the exposure Y _i . Exposure Y _i, the portion follows the market interest rates remained 0 years, the equivalent of fixed rate does not follow in which dropped to the original term.
More specifically, the exposure calculation unit 29 calculates the exposure Y _i using the following formula based on the maturity ladder X _i .

β：金利の追随率

β: Following rate of interest

Here, the exposure calculation unit 29 reads out the market interest rate data x _i , the deposit interest rate data y _{i, and the} like that are input in advance and stored in the storage device from the storage device, and uses the following formula to obtain the following rate β, intercept α and correlation coefficient R are calculated. Here, market interest rate data x _i , deposit interest rate data y _{i and the} like are stored in advance in a storage device via the input device 15, the recording medium drive device 17, the interface device 18, and the like. The follow rate β may be calculated in advance. FIG. 16 shows an example of market interest rate data x _i and deposit interest rate data y _i .

ここで、エクスポージャー算出部２９は、Ｓ_xx、Ｓ_syを、下記の式を用いて算出する。

Here, the exposure calculation unit 29 calculates S _xx and S _sy using the following equations.

なお、上述の内容は、ｙ_i＝α＋β・ｘ_iという回帰分析を最小二乗法により求める計算に相当する。
図１７に、エクスポージャー算出部２９による計算結果の一例を示す。なお、エクスポージャー算出部２９による計算結果は、記憶デバイスに格納される。

Note that the above description corresponds to the calculation for obtaining the regression analysis y _i = α + β · x _i by the least square method.
In FIG. 17, an example of the calculation result by the exposure calculation part 29 is shown. The calculation result by the exposure calculation unit 29 is stored in the storage device.

In SQ34, the risk index calculation unit 30 calculates a duration D, which is an example of a liquid deposit risk index, based on the exposure Y _i using, for example, the following equation.

Further, in SQ34, the risk index calculating unit 30, based on the exposure Y _i, and calculates the outlier ratio, which is an example of risk indicators of liquid deposits.
More specifically, the risk index calculation unit 30 determines (1) an interest rate shock due to a parallel movement of 200 basis points (BP: 1BP = 0.01%) up and down, or (2) a holding period of one year (240 business days). ), Interest rate risk of bank accounts such as interest-related assets / liabilities in banking accounts and derivative transactions, and Tier 1 capital and Tier 2 capital due to interest rate fluctuations with a probability of 1% up and down with an observation period of at least 5 years Find the ratio to the sum of.

In SQ36, the deposit balance prediction unit 31 reads the latest data of the time series data (2) (deposit interest rate data, market data, macro index data, etc.) from the storage device.
In SQ38, the deposit balance prediction unit 31 obtains a predicted value of the deposit balance using information on the deposit balance tree. FIG. 18A shows an image of deposit balance prediction.
More specifically, the deposit balance prediction unit 31 determines to which node in the tree the latest read data belongs, and the average value, standard deviation of the balance change rate of the node to which the latest data belongs, Information obtained by processing the number of data as it is or processing into expected values, graphs, tables, etc. is output. Therefore, the user can predict the deposit balance by looking at the output content.

FIG. 18B shows an example of a tree created in SQ26. Here, among the nodes of the tree, the terminal node is referred to as a terminal node and is indicated by a double line frame.
In this example, first, based on the index 1 included in the most recent data, a branch is made depending on a threshold value of 1 or more (or more or less). Next, based on the index 2 included in the latest data, the process branches depending on whether any of the values in parentheses is met. In the present embodiment, the index 1 is an example of a continuous variable, and the index 2 is an example of a nominal variable (category variable). Note that N is the number of data belonging to the node, and indicates the number of past data samples used for creating a tree.

That is, the deposit balance prediction unit 31 branches based on the index 1 and the index 2 and sets μ of the corresponding terminal node as the μ predicted value. At this time, the deposit balance prediction unit 31 determines the μ predicted value according to the example of the processing program shown in FIG. 18C (processing example 600).
Similarly, the deposit balance prediction unit 31 may obtain the predicted value using information on the volatility tree.

FIG. 19 shows an example of output data of the deposit balance prediction unit 31. The predicted values μ and σ are values determined according to the processing program shown in FIG. 18C. Further, the deposit balance increase / decrease is a value obtained by subtracting the reference date balance V ₀ from the deposit balance Vα (t) predicted by the following formula. Here, for t and α, values according to a period to be predicted and a reliability level are set via the input device 15, the recording medium drive device 17, the interface device 18, and the like, and stored in a storage device, for example. In FIG. 19, when t = 1W (1 week), 1M (1 month), and α = 15.9% and 84.1%, the amount of increase or decrease in deposit balance at each period and confidence level is obtained. Indicates.

As described above, according to the above-described configuration, the deposit balance change rate μ and the volatility σ are estimated in association with explanatory variables such as the macro index and the deposit interest rate. That is, it becomes possible to analyze and explain the movement of the deposit balance change rate μ and the volatility σ based on the macro index and the deposit interest rate. Further, according to the above-described configuration, the deposit balance can be predicted from the calculated parameters. In addition, since the explanatory variables can be changed as appropriate, the calculated parameters are not merely for calculating the risk index, but can be applied to banking practice.
According to the embodiment described above, it is possible to predict the cash flow more appropriately by estimating the parameter in association with the explanatory variable.

  The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to such specific embodiments, and various modifications can be made within the scope of the gist of the present invention described in the claims.・ Change is possible.

11 CPU
12 ROM
13 RAM
14 HD
15 Input device

Claims (7)

An information processing apparatus having a storage unit,
Deposit balance information indicating the value of the deposit balance provided for each date information, explanatory variable information indicating the values of each variable of a plurality of types of variables provided for each date information, and explanatory variable definition information defining a plurality of types of variables Reading means for reading from the storage unit;
A generating unit that calculates a balance change rate of the deposit balance information, and generates integrated information that integrates the calculated balance change rate and the explanatory variable information using date information as a key;
Generating a plurality of conditions including variables defined in the explanatory variable definition information, comparing the plurality of conditions with the explanatory variable information of the integrated information, and for each node corresponding to each of the plurality of conditions Classifying means for classifying the balance change rate of the integrated information, and storing the information on the balance change rate belonging to the node and the condition corresponding to the node in the storage unit;
An information processing apparatus comprising:   The classification means includes a condition for each combination of a variable defined in the explanatory variable definition information and a value predefined for the variable or explanatory variable information corresponding to the variable included in the integrated information, and A not condition is generated, and when the balance change rate of the integrated information is divided into either a condition or a not condition, a search is made for a condition and a not condition in which the two distributions of the balance change rate are most different, and the condition and the not condition The information processing apparatus according to claim 1, wherein each of the plurality of conditions is set as a condition corresponding to a node, and searching for a next condition is repeated to generate the plurality of conditions. The generating means calculates, for each of the calculated balance change rates, volatility indicating a ratio of the fluctuation range of the deposit balance information, and using the date information as a key, integrated information obtained by integrating the calculated volatility and the explanatory variable information. Generate more,
The classification means generates a plurality of conditions including variables defined in the explanatory variable definition information, compares the plurality of conditions with the explanatory variable information of the integrated information, and corresponds to each of the plurality of conditions. 3. The information processing apparatus according to claim 2, wherein the volatility of the integrated information is classified into each node, and information on the volatility belonging to the node and a condition corresponding to the node are associated and stored in the storage unit. . Based on the information on the balance change rate belonging to each node, the negative value of the average value of the balance change rate of the node having the largest absolute value of the average value of the balance change rate of each node is determined as the balance change rate at the time of stress, Determining means for determining the average value of the volatility of the node having the largest average value of the volatility of each node as the volatility at the time of stress based on the information of the volatility belonging to each node;
VaR calculating means for calculating VaR indicating the value of the deposit balance at the time of stress using the balance change rate at the time of stress and the volatility at the time of stress determined by the determining means;
A maturity ladder calculating means for calculating a maturity ladder indicating the amount of cash flow flowing out every year using the VaR calculated by the VaR calculating means;
Exposure calculation means for calculating an exposure indicating an amount obtained by adjusting the maturity ladder calculated by the maturity ladder calculation means in consideration of a follow-up rate of the deposit interest rate to the market interest rate;
Index calculation means for calculating an index of interest rate risk using the exposure calculated by the exposure calculation means;
The information processing apparatus according to claim 3, further comprising:   The condition of each node stored in the storage unit is compared with the explanatory variable information of the date information with the integrated information, and the balance change rate of the integrated information corresponding to the date information indicates which of the plurality of nodes The information processing apparatus according to claim 2, further comprising an output unit that specifies whether the node belongs to and outputs information on a balance change rate that belongs to the specified node. An information processing method executed by an information processing apparatus having a storage unit,
Deposit balance information indicating the value of the deposit balance provided for each date information, explanatory variable information indicating the values of each variable of a plurality of types of variables provided for each date information, and explanatory variable definition information defining a plurality of types of variables Reading from the storage unit;
Generating a balance change rate of the deposit balance information, and generating integrated information that integrates the calculated balance change rate and the explanatory variable information using date information as a key;
Generating a plurality of conditions including variables defined in the explanatory variable definition information, comparing the plurality of conditions with the explanatory variable information of the integrated information, and for each node corresponding to each of the plurality of conditions Classifying the balance change rate of the integrated information, and storing the information on the balance change rate belonging to the node and the condition corresponding to the node in the storage unit;
An information processing method characterized by comprising: A computer having a storage unit;
Deposit balance information indicating the value of the deposit balance provided for each date information, explanatory variable information indicating the values of each variable of a plurality of types of variables provided for each date information, and explanatory variable definition information defining a plurality of types of variables Reading means for reading from the storage unit;
A generating unit that calculates a balance change rate of the deposit balance information, and generates integrated information that integrates the calculated balance change rate and the explanatory variable information using date information as a key;
Generating a plurality of conditions including variables defined in the explanatory variable definition information, comparing the plurality of conditions with the explanatory variable information of the integrated information, and for each node corresponding to each of the plurality of conditions Classifying means for classifying the balance change rate of the integrated information, and storing the information on the balance change rate belonging to the node and the condition corresponding to the node in the storage unit;
Program to make it work.

Images