Executive retirement benefit amount diagnostic system

Abstract

To provide a system that diagnoses the appropriateness of planned retirement benefit payment amounts for executives from the perspective of tax denial risk using a predictive model derived by statistical analysis or machine learning.SOLUTION: An executive retirement benefit amount diagnostic system of the present invention comprises: an input unit for accepting basic information about an executive to be diagnosed; a storage unit for storing a prediction model derived in advance by a statistical method or machine learning; a retirement benefit amount prediction unit for inputting the basic information into the prediction model to predict an appropriate retirement benefit amount for the executive to be diagnosed; a payment amount diagnostic unit for calculating the amount equivalent to a threshold for the occurrence of excessive denial risk for tax purposes and diagnosing the appropriateness of the planned retirement benefit payment amount from the perspective of excessive denial risk by comparing it with the planned retirement benefit payment amount for the executive included in the accepted basic information; a diagnostic result output unit for outputting the diagnostic result; and a control unit.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an executive retirement allowance diagnostic system, and more particularly to an executive retirement allowance diagnostic system that diagnoses the appropriateness of planned executive retirement allowance payments from the perspective of tax denial risk. [Background technology]

[0002] Executive retirement benefits, also known as executive retirement allowances, are an important matter for companies in terms of personnel policy and financial management. Companies generally have executive retirement allowance regulations in place, which calculate the retirement and death allowances of their executives based on factors such as their years of service, merit multiplier, and monthly salary at the time of retirement.

[0003] Executive retirement benefits can be included as a tax deduction if they are paid after a resolution at a general shareholders' meeting and are within a reasonable amount. However, if the tax authorities determine that the amount is excessive in light of the payment status of other companies in the same industry, the disproportionately high portion will be denied as a tax deduction.

[0004] On the other hand, with regard to the average merit multiplier method, which is usually adopted by tax authorities as a basis for denying executive retirement benefits, the sampling process of similar corporations in the same industry that forms the basis for calculating the multiplier is not disclosed, making it difficult for taxpayer companies to properly determine whether the executive retirement benefits they plan to pay are an amount that will not be denied by the tax authorities as excessive.

[0005] In the past, systems that simulate retirement benefits (Patent Document 1) and technology that calculates insurance amounts to prepare for economic crises due to the death of a company's executive have been proposed to support corporate financial planning. However, as mentioned above, these systems do not take into account the need to predict whether the calculated retirement benefits are appropriate amounts that will not be deemed excessive for tax purposes. Therefore, there is a need for a system that can properly diagnose whether the amount of retirement benefits to be paid to executives is within a range that is deemed reasonable. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2002-92535 [Patent Document 2] Japanese Patent Application Publication No. 10-254970 Summary of the Invention [Problem to be solved by the invention]

[0007] The present invention has been made in consideration of the above-mentioned conventional problems, and an object of the present invention is to provide an executive retirement allowance diagnosis system that diagnoses the appropriateness of the planned amount of executive retirement allowance payment from the perspective of the risk of excessive denial by tax authorities. [Means for solving the problem]

[0008] In order to achieve the above object, one aspect of the present invention provides a system for diagnosing the appropriateness of an executive's planned retirement allowance from the perspective of tax denial risk. The computer system comprises an input unit that receives basic information for diagnosing whether the planned retirement allowance of the executive being diagnosed is within a range of amounts appropriate for tax purposes; a memory unit that stores the received basic information and a prediction model derived in advance by statistical methods or machine learning, which is used to predict an appropriate retirement allowance for the executive; a retirement allowance prediction unit that inputs the received basic information into the prediction model to predict an appropriate retirement allowance for the executive; and a calculation unit that calculates an amount equivalent to the threshold for the occurrence of excessive tax denial risk based on the appropriate retirement allowance for the executive predicted by the retirement allowance prediction unit and the prediction error of the prediction model, and calculates the appropriate retirement allowance for the executive included in the basic information. a payment amount diagnosis unit that determines whether a planned retirement allowance payment is equal to or less than the calculated excessive denial limit, and diagnoses that the risk of excessive denial is low and that the planned retirement allowance payment is appropriate if the planned retirement allowance payment is equal to or less than the amount equivalent to the threshold for the occurrence of the excessive denial risk, and diagnoses that the risk of excessive denial is high and that the planned retirement allowance payment is inappropriate if the planned retirement allowance payment exceeds the amount equivalent to the threshold for the occurrence of the excessive denial risk; a diagnosis result output unit that outputs the results of the diagnosis to a display means in a predetermined display format; and a control unit that controls the input unit, the memory unit, the retirement allowance amount prediction unit, the payment amount diagnosis unit, the diagnosis result output unit, and the entire system, and the payment amount diagnosis unit sets, as the amount equivalent to the threshold for the occurrence of the excessive denial risk, an amount obtained by multiplying the prediction error of the prediction model by a predetermined coefficient and adding the result to the predicted amount of retirement allowance appropriate to the executive.

[0009] The basic information further includes as data items the position of the executive to be diagnosed, years in office, monthly remuneration at time of retirement, and corporate information of the executive's place of employment, and it is preferable that the corporate information of the executive's place of employment includes as data items one or more of the type of corporation, number of employees, annual sales, and capital. The computer system may further include a model setting unit that sets and changes the type of the prediction model and the data items to be used based on user input. It is preferable that the predictive model derived by the statistical method is a regression model derived by multiple regression analysis using actual executive retirement benefit payment information of a statistically significant number of corporations collected in advance, with at least two or more data items included in the basic information as explanatory variables. The predictive model derived by the machine learning is a regression model derived by deep learning using a statistically significant number of corporate executive retirement benefit payment actual data collected in advance, and it is preferable that at least two of the data items included in the basic information are used as an input layer, the predicted value of the retirement benefit amount is used as an output layer, and an intermediate layer includes one or more data items from the basic information that are not set in the input layer or the output layer. The control unit can store the received basic information and the results of the diagnosis in the storage unit for each officer to be diagnosed. [Effects of the Invention]

[0010] According to the present invention, the appropriateness of planned retirement benefits for executives can be easily assessed from the perspective of the risk of excessive denial for tax purposes. Therefore, appropriate retirement benefit preparation plans and tax treatments can be proposed to clients. Furthermore, the executive retirement benefit assessment system according to the present invention eliminates arbitrary sampling and uses a prediction model optimized by statistical analysis or machine learning using publicly available data, thereby providing fair and highly accurate prediction results. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a block diagram showing the configuration of an executive retirement allowance diagnostic system according to one embodiment of the present invention. [Figure 2] 1 is a block diagram showing the functional configuration of an executive retirement pay diagnostic system according to one embodiment of the present invention. [Figure 3]10 is a diagram showing an example of an initial screen generated by the input unit of the executive retirement allowance diagnostic system according to this embodiment. FIG. [Figure 4] 10 is a diagram showing an example of an input screen for user registration generated by the input unit of the executive retirement allowance diagnostic system according to this embodiment. FIG. [Figure 5] FIG. 10 is a diagram showing an example of a data table made up of data items entered from an input screen for user registration. [Figure 6] 10 is a diagram showing an example of a screen for inputting diagnosis subject information generated by the input unit of the executive retirement allowance diagnosis system according to this embodiment. FIG. [Figure 7] FIG. 10 is a diagram showing an example of a diagnosis subject data table including data items input from a diagnosis subject information input screen. [Figure 8] This is a diagram explaining a retirement benefit prediction model that uses deep learning. [Figure 9] FIG. 10 is a diagram showing an example of a diagnostic result display screen output by a diagnostic result output unit according to the present embodiment. [Figure 10] 10 is a diagram showing an example of a model management screen generated by the input unit of the executive retirement allowance diagnostic system in this embodiment. FIG. [Figure 11] FIG. 10 is a diagram showing an example of an input screen for setting modeling conditions generated by a model setting unit in this embodiment. [Figure 12] 3 is a flowchart showing the operation of the executive retirement allowance diagnostic system according to one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, specific examples of embodiments of the present invention will be described in detail with reference to the drawings.

[0013] FIG. 1 is a block diagram showing the configuration of an executive retirement allowance diagnostic system according to one embodiment of the present invention.

[0014] The executive retirement allowance diagnosis system 10 according to one embodiment of the present invention shown in Figure 1 is a system configured with a general computer, and includes a control device 11, a storage device 12, an input device 13, a display device 14, and an output device 15. It also includes a communication device 16 that establishes a network connection with an external terminal device 20, etc.

[0015] The control device 11 has a central processing unit (CPU), as well as ROM and RAM (not shown), and by reading and executing an OS (Operating System), specified programs and data, it realizes and controls each part (each functional part) that constitutes the executive retirement pay diagnosis system 10 of the present invention described below.

[0016] The storage device 12 stores various data and programs and constitutes a storage unit including a storage area (hereinafter referred to as a storage unit) according to the data content described below. The storage device 12 is preferably a non-volatile storage device such as an SSD (Solid State Drive) or HDD (Hard Disk Drive). Furthermore, it does not have to be a single storage device, and may be composed of multiple storage devices connected to each other via a communication network. In other words, the storage device 12 may be built into the executive retirement allowance diagnosis system 10 or may be composed of a data server connected via a communication network.

[0017] The input device 13 is a device for receiving various inputs including instructions from the user, and is composed of a keyboard, a pointing device such as a mouse, a touchpad, etc. The input device 13 may also be equipped with an interface for connecting to peripheral devices, and configured to be able to acquire data stored in a computer-readable storage medium.

[0018] The display device 14 is a display device that visually displays to the user the input screen generated by the control device 11, as well as the acquired information and calculation results. The display device 14 can also be configured to share some of the functions with the input device 13 as a touch panel combined with a touch pad that allows touch input.

[0019] The output device 15 is a printer that writes on paper media or a data writing device that writes on a computer-readable recording medium, and is configured as an external device connected to the computer main body or is configured in a form that is included within the computer.

[0020] The communication device 16 is configured with a communication device or the like for network connection to an external terminal device 20, etc., and functions as a communication interface. The communication device 16 is, for example, a communication card for a wired LAN or a wireless LAN (Local Area Network), Bluetooth (registered trademark), or WiFi (registered trademark), or may also be a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various types of communication.

[0021] The terminal device 20 is an external computer terminal or mobile information terminal connected via a network to the executive retirement allowance diagnostic system 10, and provides (transmits) basic information for predicting an appropriate (reasonable) retirement allowance for the executive being diagnosed using the executive retirement allowance diagnostic system 10. The terminal device 20 may be configured to receive the diagnosis results from the executive retirement allowance diagnostic system 10.

[0022] The executive retirement pay diagnosis system 10 according to one embodiment of the present invention is not limited to being configured as a single computer, but may also be configured as a network-connected cloud computing system.

[0023] Hereinafter, the functional configuration of an executive retirement allowance diagnosis system according to one embodiment of the present invention will be described with reference to FIGS.

[0024] FIG. 2 is a block diagram showing the functional configuration of the executive retirement allowance diagnostic system according to one embodiment of the present invention.

[0025] As shown in Figure 2, the executive retirement pay diagnosis system 10 according to one embodiment of the present invention includes a model setting unit 120, a retirement pay prediction unit 130, a payment amount diagnosis unit 140, and a diagnosis result output unit 150, which are functional units realized by executing an executive retirement pay diagnosis program on the CPU of the control unit 100 configured by the control device 11.

[0026] In addition, the executive retirement allowance diagnosis system 10 comprises an input unit 110 that accepts input information from the input device 13 or the terminal device 20, an output interface (I / F) unit 160 that outputs (transmits) the diagnosis results to the output device 15 or the terminal device 20, and a memory unit 200 that includes a prediction model storage unit 210, a prediction data storage unit 220, and a diagnosis-related information storage unit 230 that are assigned according to the contents to be stored, and the control unit 100 controls the above-mentioned functional units, input unit, output unit, memory unit, and the system as a whole (connecting lines connecting the control unit to each functional unit are not shown).

[0027] The input unit 110 generates an initial screen for accepting input that instructs the system (executive retirement pay diagnosis system) 10 to perform certain processes, and displays it on the display device 14 or a network-connected terminal device 20.The input unit 110 receives instructions and / or information (data) entered by the user on the displayed initial screen via the input device 13 (or communication device 16) and transmits them to a specified output destination (the relevant functional unit or control unit).

[0028] Specifically, the input unit 110 receives instruction inputs on an initial screen to move (progress) to a stage for registering information about a user (individual or corporation) who wishes to be diagnosed, a stage for inputting information about the person to be diagnosed, and a stage for inputting setting information such as modeling conditions for a prediction model to be used in the system 10, and transmits the received instruction inputs to the control unit 100. The control unit 100 executes a program corresponding to the instruction input, thereby operating each corresponding functional unit and causing the specified processing to proceed.

[0029] FIG. 3 is a diagram showing an example of an initial screen generated by the input unit of the executive retirement allowance diagnostic system according to this embodiment.

[0030] When the input unit 110 receives an instruction input from the user on the initial screen 300, for example, an instruction to move to a screen for inputting information about the corporation or the like to which the user belongs, that is, an instruction (click) on the image portion (button display) of "1. User Registration" shown in Figure 3, the input unit 110 transmits the instruction input to the control unit 100.

[0031] In addition, when the input unit 110 receives an instruction input to move to a screen for inputting information such as the position of the executive to be diagnosed, the number of years in office, and the planned amount of retirement benefits to be paid, i.e., an input instructing the image portion of "2. Input of information about the executive to be diagnosed" shown in Figure 3, it transmits the instruction input to the retirement benefits prediction unit 130 via the control unit 100.

[0032] Similarly, when the input unit 110 receives an instruction input from the user on the initial screen 300 to move to a screen for managing the predictive models used in the system 10, i.e., an input instructing the image portion of "3. Model Management" shown in FIG. 3, the input unit 110 transmits the instruction input to the control unit 100 and the model setting unit 120. Note that users who can manage predictive models are limited to those with pre-registered administrator authority, and it is desirable that the configuration requires personal authentication on the screen to which the user is directed. Managing predictive models includes selecting the type of predictive model to be used in the system 10, setting the data items to be used in the selected predictive model, and setting a data table for deriving model parameters.

[0033] When the input unit 110 receives input from the user on the initial screen 300 to point to (click on) the image portion (button display) of "1. User Registration" shown in Figure 3, it generates a screen for inputting information about the user (individual or corporation) who wishes to use the system 10, and displays it on the display device 14 or a network-connected terminal device 20.

[0034] 4 is a diagram showing an example of an input screen for user registration generated by the input unit of the executive retirement allowance diagnostic system according to this embodiment. In the following explanation, we will explain the case where the input screen is displayed on the display device 14, but the same applies when it is displayed on the network-connected terminal device 20.

[0035] When the name, type (corporation, family-owned company, etc.), capital, number of employees, industry, annual sales, and location of the corporation or other entity that will be the user are entered via the input device 13 into multiple text boxes (401 to 407) corresponding to the input items displayed on the input screen 400 shown in Figure 4, the input unit 110 creates a data table A consisting of data items corresponding to each text box (401 to 407) and stores it in the diagnosis-related information storage unit 230 of the memory unit 200.

[0036] Each text box (401 to 407) corresponds to an input field (cell) of the spreadsheet software, and information (data) entered into some of the text boxes (401, 402) is linked to data table B, which corresponds to the input of diagnostic subject information, which will be described later with reference to Figure 6, and the information entered into some of the text boxes (401, 402) is reflected in the corresponding items (name, type, etc. of the corporation, etc.) of data table B.

[0037] 5 is a diagram showing an example of a data table made up of data items entered on an input screen for user registration. In the description herein, data table A will be referred to as "company information data table A."

[0038] When the "Register" button 410 displayed on the input screen 400 shown in FIG. 4 is instructed (clicked), the input unit 110 determines that the input is complete and stores the created company information data table A in the diagnosis-related information storage unit 230. At this time, the input unit 110 associates a "user ID" for identification with the saved company information data table A and stores it. The user ID is displayed, for example, in a user ID display field 420 on the input screen 400 shown in FIG. 4. At the same time, the user ID may be output (transmitted) to the output device 15 or the terminal device 20 via the output interface unit 160.

[0039] When the input unit 110 receives input from the user on the initial screen 300 by pointing (clicking) on ​​the image portion (button display) of "2. Input information about the person to be diagnosed" shown in Figure 3, it generates a screen for inputting information such as the position of the executive to be diagnosed, the number of years in office, and the planned amount of retirement benefits to be paid, and displays this on the display device 14.

[0040] FIG. 6 is a diagram showing an example of a screen for inputting diagnosis subject information generated by the input unit of the executive retirement allowance diagnosis system according to this embodiment.

[0041] 6, when the "user ID" assigned at the time of user registration is entered in text box 601, control unit 100 authenticates the entered user ID, reads out company information data table A corresponding to the authenticated user ID from diagnosis-related information storage unit 230 in memory unit 200, and displays it in a predetermined format so that the user can confirm it. On input screen 600 shown in FIG. 6, the "name" and "type" of the corporation or the like are displayed in text boxes (602, 603).

[0042] The screen 600 for inputting information about the person to be diagnosed displays text boxes for inputting information about the executive career and remuneration of the executive to be diagnosed. In this embodiment, the input items related to the executive career and remuneration of the executive to be diagnosed are the executive name, job title, years in office, monthly remuneration at time of retirement, and planned retirement allowance, and text boxes (604 to 608) are provided for each item. However, the input items are not limited to these.

[0043] After the information of the officer to be diagnosed is entered via the input device 13 (or communication device 16) into the text boxes (604 to 608) corresponding to each input item on the input screen 600, when a "Register" button 610 displayed on the input screen 600 shown in Fig. 6 is instructed (clicked), the input unit 110 determines that the input is complete, creates a data table B consisting of data items corresponding to each text box (604 to 608) using the user ID as an identifier, and saves it in the diagnosis-related information saving unit 230 of the memory unit 200. Hereinafter, data table B will be referred to as "diagnosis target data table B" in the description of this specification.

[0044] Fig. 7 is a diagram showing an example of a diagnosis subject data table consisting of data items entered from a diagnosis subject information input screen. The diagnosis subject data table B shown in Fig. 7 forms a link table with the company information data table A shown in Fig. 5, with the user ID as a common identifier.

[0045] The information on the user company entered through the input screen 400 for user registration and the information on the executive to be diagnosed entered through the input screen 600 for the information on the person to be diagnosed correspond to the basic information used by the executive retirement allowance amount diagnosis system 10 to diagnose whether the planned amount of retirement allowance to be paid to the executive to be diagnosed is within the range of amounts that are appropriate for tax purposes.

[0046] When the "Proceed to diagnosis" button 620 displayed on the input screen 600 shown in Figure 6 is clicked, the control unit 100 activates the retirement allowance prediction unit 130, and predicts the appropriate retirement allowance amount for the executive to be diagnosed using the basic information registered in the company information data table A and the diagnosis target data table B, and stores the predicted retirement allowance amount (hereinafter referred to as the retirement allowance prediction amount) in the prediction data storage unit 220. Here, "reasonable retirement allowance" means the amount of retirement allowance that the director being diagnosed and a director of the same position who belongs to a similar corporation in the same industry are expected to receive when the differences in various factors that normally exist and their individual peculiarities are ignored and the amount is averaged out.

[0047] The retirement allowance prediction unit 130 can use a regression model derived by statistical analysis or machine learning using previously collected information on the actual retirement allowance payments for executives of a statistically significant number of corporations as a prediction model for predicting the appropriate retirement allowance amount for the executive to be diagnosed. The previously derived prediction model and its prediction error are stored in the prediction model storage unit 210 of the memory unit 200. The step of setting up the regression model to be used for prediction will be described in detail later.

[0048] The retirement allowance prediction unit 130 inputs data registered in the data items designated as explanatory variables (or input variables) of the prediction model from the company information data table A and the diagnosed person data table B into a prediction model derived by statistical analysis or machine learning, and predicts (calculates) the amount of retirement allowance corresponding to (i.e., appropriate) the basic information of the executive to be diagnosed.

[0049] As an example, when applying a prediction model derived using statistical techniques, the retirement allowance prediction unit 130 calculates the predicted value of the retirement allowance amount (predicted retirement allowance amount), which is the dependent variable, in the multiple regression equation shown in the following equation (1): for example, in the processing stage related to the "model management" described below, when at least one of "number of employees," "annual sales," or "capital" from company information data table A and "years of executive tenure" from diagnostic target data table B are set as explanatory variables, the multiple regression equation consisting of these explanatory variables is input with data registered in the cells of the corresponding data items in the data column associated with the user ID of the company to which the executive to be diagnosed belongs from both data tables (A, B) shown in Figures 5 and 7.

[0050]

number

[0051] Specifically, if the "number of employees" in the company information data table A is set as the explanatory variable X1 and the "years of tenure as director" in the diagnosis target data table B is set as the explanatory variable X2, then in formula (1), X3 to X i are omitted. In addition, the partial regression coefficients (β1, β2) and constant β0 corresponding to these explanatory variables (X1, X2) are calculated in advance at the stage of deriving the model and stored in the prediction model storage unit 210. However, the number of explanatory variables and data items are not limited to the above two items, and for example, a process may be performed to determine the explanatory variables to be used in the prediction calculation based on the evaluation results of multicollinearity for three or more registered data items.

[0052] In the present embodiment described above, a regression equation derived by linear regression among statistical techniques is used as the prediction model, but the present invention is not limited to this.

[0053] As another example, when applying a predictive model derived by machine learning, in addition to linear regression, decision trees, random forests, support vector machines, neural network models, and even deep learning may be used.

[0054] Figure 8 is a diagram for explaining a retirement benefit prediction model that applies deep learning.

[0055] When deriving a retirement benefit prediction model using deep learning, for example, initially, the model is constructed using at least three layers: input, intermediate, and output layers, with 90% of the actual data used as training data and the remaining 10% used as test data. Subsequently, the number of intermediate layers and the ratio of training data to test data can be revised depending on the prediction error.

[0056] For example, when applying deep learning to derive a prediction model for retirement benefits, the explanatory variables (X1, X2) set in the above multiple regression equation are used as the input layer (x1, x2), and in addition to the explanatory variables (X1, X2), the intermediate layer is made up of the parameter variables (a0, a1, a2), and the objective variable (y (where the possible values ​​are y1, ..., y n A multilayer neural network (see Figure 8) is configured with the predicted value of the retirement allowance amount as the output layer, and using a statistically significant number of corporate executive retirement allowance payment records (configured in a data set) collected in advance, the predicted value (predicted amount) of the retirement allowance amount calculated by the configured multilayer neural network and the actual payment amount are input into the squared error function shown in the following formula (2), and the weight w that minimizes the error function value En is calculated. ij The model is optimized by finding it using the error propagation method. For the sake of convenience, a three-layered neural network is shown in Figure 8, but deep learning can be configured by adding more layers to the intermediate layer.

[0057]

number

[0058] In addition, the predicted value y k The relationship between the input layer (x1, x2) and the hidden layer (a0, a1, a2) is expressed by the following series of Equations 3.

[0059]

number

[0060] When the payment amount diagnosis unit 140 receives the retirement payment amount (predicted retirement payment amount) appropriate for the executive to be diagnosed, as predicted by the retirement payment prediction unit 130, it adds the amount obtained by multiplying the prediction error of the prediction model used for the prediction (standard error in the case of a regression model derived using statistical methods) by a predetermined coefficient to the received predicted retirement payment amount, calculates an amount equivalent to the threshold for the occurrence of excessive denial risk, and compares the calculated amount equivalent to the threshold for the occurrence of excessive denial risk with the amount registered in the ``Planned Retirement Payment Amount'' field in the diagnosis target data table B.

[0061] The system then determines whether the planned amount of retirement allowance to be paid for the executive officer being assessed is equal to or less than the amount equivalent to the threshold for the occurrence of excessive denial risk. If the planned amount of retirement allowance to be paid is equal to or less than the amount equivalent to the threshold for the occurrence of excessive denial risk, it diagnoses that the risk of excessive denial is low and the planned amount of retirement allowance to be appropriate, and if the planned amount of retirement allowance to be paid exceeds the amount equivalent to the threshold for the occurrence of excessive denial risk, it diagnoses that the risk of excessive denial is high and the planned amount of retirement allowance to be inappropriate.

[0062] The coefficient by which the prediction error of the prediction model is multiplied to calculate the amount equivalent to the threshold for the occurrence of excessive denial risk is determined based on past records of excessive denials by tax authorities and precedents. For example, in the case of a prediction model derived using statistical methods (multiple regression model), the coefficient is set within the range of 1.0 to 1.96.

[0063] The diagnostic result output unit 150 outputs the diagnostic results from the payment amount diagnostic unit 140 to the display means in a predetermined display format. In this embodiment, the calculated retirement benefit forecast amount is output to the display device 14 in the form of a probability distribution graph.

[0064] Figure 9 is a diagram showing an example of a diagnostic result display screen output by the diagnostic result output unit according to this embodiment. The diagnostic result display screen 900 shown in Figure 9 shows a distribution chart in which the estimated retirement allowance amount appropriate for the executive to be diagnosed, as predicted by a statistically derived prediction model, is used as a representative value, and the amount corresponding to the threshold for the risk of excessive denial calculated based on the standard error of the prediction model, is plotted overlaid with the planned retirement allowance payment amount stored in the diagnosis target data table B. The diagnostic result display screen 900 also displays the user ID and name of the corporation or other entity in a text display field 901, and the result of the tax denial risk assessment performed by the payment amount assessment unit 140 for the diagnosis target in a text display field 902. However, this is merely an example, and the display format of the diagnostic results is not limited to the example shown in Figure 9.

[0065] When the input unit 110 receives input from the user on the initial screen 300 to point (click) on the image portion (button display) of "3. Model Management" shown in Figure 3, it generates a screen for inputting setting information such as input variables used to register executive retirement pay survey data (actual executive retirement pay payment information) and derive a prediction model, and displays the screen on the display device 14.

[0066] FIG. 10 is a diagram showing an example of a model management screen generated by the input unit of the executive retirement allowance diagnostic system in this embodiment.

[0067] The model management screen 1000 shown in Figure 10 displays a field for selecting work items and an input field 1001 for an administrator ID, and is configured so that only users with pre-registered administrator privileges can proceed to the next stage.

[0068] When the administrator ID is entered into the input field 1001 on the model management screen 1000 shown in FIG. 10, and "Modeling condition setting" is selected as the work item, and then an instruction to execute the operation (clicking the "OK" button 1010) is received, the control unit 100 authenticates the administrator ID, and if the authentication is successful, activates the model setting unit 120.

[0069] The model setting unit 120 generates an input screen for the administrator to set the derivation or learning conditions of a model corresponding to a statistical model or a machine learning model, and causes the display device 14 to display the screen.

[0070] 11 is a diagram showing an example of an input screen for setting modeling conditions generated by the model setting unit in this embodiment. In this embodiment, the case of a statistical model will be described.

[0071] 11 displays an input screen 1100 for selecting data items that affect the data item that serves as the objective variable (in this embodiment, the predicted retirement benefit amount), i.e., data items to be set as explanatory variables in the prediction model (in this embodiment, the case of a multiple regression model is described) used to calculate the predicted retirement benefit amount. For example, the administrator checks (points to) the boxes for "number of employees," "annual sales," "job title," and "years in office" from the data items displayed in the explanatory variable selection field D.

[0072] The input screen 1100 also displays a field E for selecting the type of prediction model. A scroll button 1110 may be provided in the selection field E, allowing the type of prediction model to be used to be selected by scrolling. Thereafter, when the "OK" button 1120 is clicked, the model setting unit 120 generates a model parameter derivation data table (not shown) in which data for the data items selected in the explanatory variable selection field D are arranged from the company information data table A and the diagnosis recipient data table B stored in the diagnosis-related information storage unit 230.

[0073] Therefore, by specifying the data items displayed in the explanatory variable selection field D, it is possible to change or add data items, and a rearranged data table for deriving model parameters is created by clicking the "Update" button 1130. Note that it is preferable to collect and register in the data table for deriving model parameters not only data registered by the user, but also a dataset consisting of actual executive retirement pay information for a statistically significant number of corporations available in the private sector (a dataset consisting of the data items included in both data tables described above) and data on executive retirement pay amounts obtained from official publications.

[0074] After creating a data table for deriving model parameters, when the model setting unit 120 receives a command specifying the type of prediction model, it obtains the expression or formula (abbreviated as prediction model formula) of the specified prediction model from the prediction model storage unit 210, and performs statistical processing using the data contained in the corresponding data item of the created data table for deriving model parameters to calculate or optimize coefficients for each explanatory variable of the prediction model.

[0075] Thereafter, the coefficients of the explanatory variables included in the prediction model formula (partial regression coefficients in the case of multiple regression) and prediction errors (standard errors in the case of multiple regression), which are derived using the data table for deriving model parameters, are saved or overwritten in the prediction model saving unit 210. In the case of a machine learning model (deep learning), the learning process may be performed by adding a process for optimizing the data items themselves (i.e., combinations of data items) specified in the intermediate layer of the prediction model.

[0076] On the model management screen shown in Figure 10, "Prediction model management" and "Registered information management" correspond to processes for modifying or deleting the created prediction model formula and registered user information, respectively, and detailed explanations are omitted.

[0077] By configuring the model management input screens 1000 and 1100 as described above, the administrator can appropriately set the diagnostic data items and prediction models by reflecting the acquired performance data. Note that the administrator can also configure the input screen 1100 in Fig. 11 so that he or she can start over from the selection of explanatory variables and / or prediction models by instructing (clicking) the "Reset" button 1140.

[0078] The data processing flow of the executive retirement allowance diagnosis system according to one embodiment of the present invention will be described in detail below with reference to FIGS. 1 to 3 and 12. FIG.

[0079] In the executive retirement pay diagnosis system 10 of this embodiment, the control unit 100 executes the executive retirement pay diagnosis program and proceeds through a series of steps from step S100 to step S370 described below based on instructions input by the user, so that user operations are simplified to instructing (clicking) on ​​the selection icons (buttons) displayed on the screen by the above-mentioned executive retirement pay diagnosis system 10 and entering numerical values ​​into the input fields.

[0080] Therefore, users can estimate (standard) executive retirement pay amounts appropriate to the executives being assessed based on statistical analysis and assess the appropriateness of the planned executive retirement pay amounts from the perspective of the risk of excessive denial for tax purposes, without needing specialized knowledge or programming skills regarding executive retirement pay. Furthermore, the executive retirement pay assessment system 10 according to this embodiment can select effective data items for training a prediction model from all data items included in the acquired assessment subject data and company information data.

[0081] FIG. 12 is a flowchart showing the operation of the executive retirement allowance diagnostic system according to one embodiment of the present invention.

[0082] As shown in Figure 12, when the executive retirement allowance diagnostic system 10 is started, the control unit 100 of the executive retirement allowance diagnostic system 10 executes the executive retirement allowance diagnostic program, causes the input unit 110 to generate an initial screen (see Figure 3), and displays it on the display device 14 (step S100). In the following explanation, detailed explanations of the functions and operations of the individual functional units of the system 10 that execute the processing at each stage are omitted, as they are as described above. The explanation will be given assuming that input of instructions and the like based on the screen display is handled by the display device 14, which is equipped with a touchpad and also serves as the pointing input function of the input device 13.

[0083] When "1. User Registration" is selected from the "Work Items" displayed on the initial screen 300 (step S110), the input unit 110 generates and displays an input screen (see FIG. 4) for user information registration (step S120). The name, type, capital, number of employees, industry, annual sales, and location of the corporation or the like are entered into the text box-style input fields (input cells) displayed on the input screen 400, and when the "Register" button 410 displayed on the input screen 400 is selected (clicked), the input unit 110 creates a data row consisting of the data items entered in each input field and registers it in the company information data table A stored in the diagnosis-related information storage unit 230 of the memory unit 200 (step S130). After that, the process returns to step S100 and waits for the next instruction to be input.

[0084] The user information may be input via an input screen displayed on a network-connected terminal device 20. In this case, the executive retirement allowance diagnosis system 10 is network-connected to the terminal device 20 via the communication device 16 and is configured to be able to communicate data with the terminal device 20.

[0085] When "2. Enter diagnostic subject information" is selected from the "Work items" displayed on the initial screen 300 (step S110), the input unit 110 generates an input screen 600 (see FIG. 6) for entering diagnostic subject information and displays it on the display device 14 (step S220). Thereafter, each piece of information about the officer to be diagnosed is entered into the text box-style input fields (input cells) displayed on the input screen 600, and when the "Register" button 610 is selected, the input unit 110 creates a data row consisting of the entered data items and registers it in the diagnostic subject information data table B stored in the diagnosis-related information storage unit 230 of the memory unit 200 (step S230).

[0086] The retirement allowance prediction unit 130 uses the user's company information data and diagnosis target data registered in the company information data table A and diagnosis target data table B to predict (calculate) the appropriate retirement allowance amount (retirement allowance prediction amount) that the director to be diagnosed and a director with the same position who belongs to a similar corporation in the same industry will receive, using a predetermined prediction model (step S240).

[0087] The retirement allowance prediction unit 130 outputs the predicted retirement allowance amount to the payment amount diagnosis unit 140, and the tax audit prediction unit 140 adds the received predicted retirement allowance amount to the amount obtained by multiplying the prediction error of the prediction model used for the prediction by a predetermined coefficient, to calculate an amount equivalent to the threshold for the occurrence of excessive denial risk (step S250).

[0088] The payment amount diagnosis unit 140 compares the calculated amount equivalent to the threshold for the occurrence of excessive denial risk with the planned retirement allowance payment amount registered in diagnosis target data table B to diagnose the appropriateness of the excessive denial risk (step S260). That is, if the planned retirement allowance payment amount is equal to or less than the amount equivalent to the threshold for the occurrence of excessive denial risk, the system diagnoses that the risk of excessive denial is low and the planned retirement allowance payment amount is appropriate. If the planned retirement allowance payment amount exceeds the amount equivalent to the threshold for the occurrence of excessive denial risk, the system diagnoses that the risk of excessive denial is high and the planned retirement allowance payment amount is inappropriate. The diagnosis result output unit 150 then outputs the diagnosis results from the payment amount diagnosis unit 140 to display means in a predetermined display format.

[0089] When "3. Model Management" among the "Work Items" displayed on the initial screen 300 is selected (step S110), the input unit 110 generates and displays a model management screen (see FIG. 10) (step S320). The desired work is selected from the work items displayed on the model management screen 1000 (step S330). When receiving an input selecting "Modeling Condition Setting," the model setting unit 120 selects the type of predictive model and generates and displays an input screen 1100 (see FIG. 11) for setting the derivation or learning conditions for the selected predictive model (step S340).

[0090] When the input unit 110 receives an input for selecting a data item to be set as an explanatory variable of the prediction model, it transmits a command specifying a data table for deriving model parameters and a prediction model to the model setting unit 120 (step S350).

[0091] The model setting unit 120 uses the model parameter derivation data table to optimize the coefficients for each explanatory variable of the prediction model (or the weights for each node) through statistical analysis or machine learning (step S360). The optimization is achieved by the model setting unit 120 applying the model parameter derivation data table to the prediction model and adjusting the coefficients or weights so as to minimize the error with the actual values. Note that an example of a screen for confirming the adjustment results is omitted.

[0092] Thereafter, the control unit 100 returns to step S100 and displays the initial screen (see FIG. 3). The user selects whether to end the task or perform another task using buttons displayed on the initial screen 300 (step S370). To end the task, the user instructs (clicks) the "End" button 310. When another task is selected, the control unit 100 continues to execute the corresponding process.

[0093] The predictions made using the above-mentioned data processing and statistical analysis techniques can be realized by a computer-executable program written in a programming language such as C++, JavaScript (registered trademark), R language, or Python, and can be stored and distributed on a computer-readable recording medium such as ROM, EEPROM, EPROM, flash memory, CD-ROM, CD-RW, DVD, SD card, or USB memory.

[0094] As described above, the executive retirement allowance diagnosis system of the present invention uses statistical analysis or machine learning to determine the appropriateness of an executive's planned retirement allowance payment from the perspective of the risk of excessive denial for tax purposes. This significantly reduces the manpower, time, and other costs required for preparing documents necessary to estimate the appropriateness of an executive's retirement allowance, i.e., the risk of an executive's retirement allowance being overstated. Furthermore, the executive retirement allowance diagnosis support system of the present invention can create and present supporting data that can counter the average merit multiplier method used by tax authorities in tax litigation and other cases, which is difficult for taxpayers to grasp, regarding excessive determinations of executive retirement allowances.

[0095] Although the embodiments of the present invention have been described in detail above with reference to the drawings, the present invention is not limited to the above-described embodiments and can be modified in various ways without departing from the technical scope of the present invention. [Explanation of symbols]

[0096] 10. Executive Retirement Pay Diagnostic System 11 Control device 12 Storage device 13 Input Devices 14 Display device 15 Output Devices 16. Communications equipment 20 Terminal equipment 100 control section 110 Input section 120 Model Setting Section 130 Retirement Benefit Forecasting Department 140 Payment Amount Diagnostics Department 150 Diagnostic result output unit 160 Output interface (I / F) section 200 Storage section 210 Prediction model storage unit 220 Prediction Data Storage Unit 230 Diagnostic related information storage unit

Claims

1. A computer system that diagnoses the appropriateness of the planned amount of retirement benefits to be paid to executives from the perspective of the risk of tax denial, The computer system includes: An input unit that receives basic information for diagnosing whether the planned amount of retirement benefits to be paid to the executive officer to be diagnosed is within a range of amounts appropriate for tax purposes; a storage unit that stores a prediction model derived in advance by a statistical method or machine learning, which is used to predict the appropriate retirement allowance amount for the executive, and the received basic information; a retirement allowance prediction unit that inputs the received basic information into the prediction model and predicts an appropriate retirement allowance for the executive; a payment amount diagnosis unit that calculates an amount equivalent to a threshold for the occurrence of excessive denial risk for tax purposes based on the retirement allowance amount appropriate to the executive predicted by the retirement allowance amount prediction unit and the prediction error of the prediction model, determines whether the planned retirement allowance payment amount for the executive included in the basic information is equal to or less than the amount equivalent to the calculated threshold for the occurrence of excessive denial risk, and diagnoses that the risk of excessive denial is low and that the planned retirement allowance payment amount is appropriate if the planned retirement allowance payment amount is equal to or less than the amount equivalent to the threshold for the occurrence of excessive denial risk, and diagnoses that the risk of excessive denial is high and that the planned retirement allowance payment amount is inappropriate if the planned retirement allowance payment amount exceeds the amount equivalent to the threshold for the occurrence of excessive denial risk; a diagnostic result output unit that outputs the diagnostic result to a display means in a predetermined display format; a control unit that controls the input unit, the storage unit, the retirement allowance amount prediction unit, the payment amount diagnosis unit, the diagnosis result output unit, and the entire system; The executive retirement pay diagnosis system is characterized in that the payment amount diagnosis unit sets the amount equivalent to the threshold for the occurrence of the excessive denial risk as an amount obtained by multiplying the prediction error of the prediction model by a predetermined coefficient and adding the result to the predicted retirement pay amount appropriate to the executive.

2. The executive retirement pay diagnosis system described in claim 1, characterized in that the basic information further includes, as data items, the position of the executive to be diagnosed, years of service as executive, monthly remuneration at time of retirement, and corporate information of the executive's place of employment, and the corporate information of the executive's place of employment includes, as data items, one or more of the type of corporation, number of employees, annual sales, and capital.

3. The executive retirement pay diagnosis system of claim 1, characterized in that the computer system further includes a model setting unit that sets and changes the type of the predictive model and the data items to be used based on user input.

4. The executive retirement benefit diagnosis system described in claim 2, characterized in that the predictive model derived by the statistical method is a regression model derived by multiple regression analysis using executive retirement benefit payment actual information of a statistically significant number of corporations collected in advance, with at least two or more data items contained in the basic information as explanatory variables.

5. The predictive model derived by the machine learning is a regression model derived by deep learning using actual executive retirement pay payment information of a statistically significant number of corporations collected in advance, and has at least two or more data items included in the basic information as an input layer, a predicted value of the retirement pay amount as an output layer, and includes one or more data items from the basic information that are not set in the input layer or the output layer as an intermediate layer.

6. 2. The executive retirement allowance diagnosis system according to claim 1, wherein the control unit stores the received basic information and the results of the diagnosis in the memory unit for each executive to be diagnosed.

Images