Information processing method, information processing device, and program

Abstract

An information processing method comprises: acquiring a first actual demand including an actual demand of an object in a first period (S30); inputting, to a demand prediction model that is trained to output a plurality of demand variation scenarios indicating time-series information of demand variation of the object, and a plurality of occurrence probabilities respectively corresponding to the plurality of demand variation scenarios when the actual demand of the object is received as an input, the first actual demand to acquire, as outputs, a plurality of demand variation scenarios in a second period after the first period and a plurality of occurrence probabilities respectively corresponding to the plurality of demand variation scenarios (S40); and outputting production plan information that indicates a production plan of the object and is based on the plurality of acquired demand variation scenarios and the plurality of acquired occurrence probabilities (S90).

Description

Information Processing Method, Information Processing Apparatus, and Program

[0001] The present disclosure relates to an information processing method, an information processing apparatus, and a program.

[0002] Patent Document 1 discloses a production plan formulation apparatus capable of formulating a production plan.

[0003] Japanese Patent No. 7410345

[0004] By the way, it is important to accurately predict the demand for an object when creating a production plan. However, since it is predicting the future, the demand prediction may not be accurate. Therefore, it is desirable that such uncertainty in demand prediction be considered in the production plan to be created.

[0005] Therefore, the present disclosure provides an information processing method and the like capable of outputting a production plan considering the uncertainty of demand prediction.

[0006] An information processing method according to an aspect of the present disclosure acquires first demand results including the demand performance of an object in a first period, inputs the demand performance of the object, and inputs the first demand results to a demand prediction model learned to output a plurality of demand fluctuation scenarios indicating time series information of the demand fluctuation of the object and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios. A plurality of demand fluctuation scenarios in a second period after the first period and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios output thereby are acquired, and production plan information indicating a production plan of the object is output based on the acquired plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

[0007] An information processing device according to one aspect of the present disclosure includes: a performance acquisition unit that acquires a first demand performance including the demand performance of an object during a first period; a demand acquisition unit that takes the demand performance of the object as input and acquires a plurality of demand fluctuation scenarios for a second period following the first period, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, which are output by inputting the first demand performance into a demand forecasting model that has been trained to output a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations of the object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios; and an output unit that outputs production plan information showing a production plan for the object based on the acquired plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

[0008] A program relating to one aspect of this disclosure is a program that causes a computer to execute the above-described information processing method.

[0009] According to one aspect of this disclosure, it is possible to realize an information processing method, etc., that can output a production plan that takes into account the uncertainty of demand forecasts.

[0010] Figure 1 is a block diagram showing the configuration of the information processing device according to the embodiment. Figure 2A is a diagram showing an example of past PSI performance data. Figure 2B is a diagram showing an example of information indicating various costs. Figure 3A is a diagram showing an example of the uncertainty of the demand forecast according to the embodiment being expressed probabilistically in a scenario tree. Figure 3B is a diagram showing an example of a production plan including production volume for each demand fluctuation scenario according to the embodiment. Figure 4A is a diagram for explaining the demand fluctuation scenario when there are three branching levels. Figure 4B is a diagram showing a list of demand fluctuation scenarios with an occurrence probability of a predetermined value or higher. Figure 5 is a flowchart showing the operation of the information processing device according to the embodiment. Figure 6 is a flowchart showing the detailed operation of step S70 shown in Figure 5.

[0011] (Background to this disclosure) Before explaining this disclosure, we will explain the background to this disclosure.

[0012] In formulating PSI plans (Production, Sales, and Inventory), the traditional MinMax method often requires setting a higher safety stock level because demand forecasts can be inaccurate. Supply chain management (SCM) employs this safety stock management method, which can result in excess inventory and a deterioration of cash flow. Conversely, holding too little inventory can lead to stockouts and lost sales opportunities. Cash flow can be calculated by subtracting the increase or decrease in costs and inventory from sales (wholesale price x actual sales volume). Costs are calculated as (cost of goods sold + production cost + transportation cost) x production volume + inventory storage cost x warehouse inventory volume. The increase or decrease in inventory is calculated as wholesale price x (ending warehouse inventory volume - beginning warehouse inventory volume). For example, a decrease in inventory is added to cash flow, and an increase in inventory is subtracted from cash flow. Note that production volume can be either the number of units produced or the weight of units produced.

[0013] With current technology, demand forecasting is uncertain, making it difficult to formulate an appropriate PSI plan (in this case, a production plan).

[0014] Therefore, the inventor of the present invention diligently studied information processing methods that can output production plans that take into account the uncertainty of demand forecasts, and devised the following information processing methods.

[0015] Furthermore, in order to account for the uncertainty of demand forecasts, it is important to comprehensively cover demand fluctuation scenarios (hereinafter simply referred to as "scenarios") that show time-series information on future demand fluctuations, and to calculate the probability of each occurring. By considering multiple demand fluctuation scenarios and their probabilities, it is possible to output a production plan that can respond to any actual demand fluctuation scenario and is aligned with the branching of the demand fluctuation scenario. This makes it possible to formulate a PSI plan that can respond to a wide range of demand fluctuation scenarios and branch flexibly.

[0016] Furthermore, it is necessary to optimize the PSI plan so that the expected value of the objective function, such as profit or cash flow, is maximized. By creating such a PSI plan, it becomes possible to achieve both improved cash flow through inventory reduction and prevention of stockouts.

[0017] However, if we were to flexibly branch the plan according to demand fluctuation scenarios and constantly try to maximize the expected value thereafter, the calculations would become extremely complex, making it difficult to resolve within a realistic timeframe.

[0018] Therefore, the inventors of this application are also diligently studying how to reduce the processing load of an information processing device that outputs a production plan that takes into account the uncertainty of demand forecasts. Specifically, they are reducing the processing load of the information processing device by dividing multiple demand fluctuation scenarios into subproblems using a scenario aggregation method that applies Lagrangian relaxation, and further by applying LP relaxation using a capacity scaling method to each subproblem.

[0019] The information processing method according to the first aspect of this disclosure acquires a first demand record including the actual demand for an object during a first period, and inputs the first demand record into a demand forecasting model that has been trained to take the actual demand for the object as input and output a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations for the object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, which are output by inputting the first demand record into the demand forecasting model that has been trained to output a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations for the object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, and outputs production plan information showing a production plan for the object based on the acquired plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

[0020] This allows for the output of production planning information that shows production plans based on multiple demand fluctuation scenarios and multiple probability occurrences. Compared to production plans created from only one demand fluctuation scenario, this method enables the output of production plans that take into account the uncertainty of demand forecasts.

[0021] Furthermore, for example, the information processing method according to the second embodiment is the information processing method according to the first embodiment, wherein the unpredictability constraint imposed on the time series forecast of production volume using two or more demand fluctuation scenarios from the plurality of demand fluctuation scenarios and two or more occurrence probabilities for each of the two or more demand fluctuation scenarios is relaxed to calculate the production volume, the extent to which the unpredictability constraint is violated is calculated, and the production volume is recalculated by adding a penalty term corresponding to the relaxation of the unpredictability constraint to the objective function for calculating the production plan of the object.

[0022] This allows for the calculation of production volume without having to calculate unpredictability constraints, thus reducing the processing load compared to when unpredictability constraints are calculated.

[0023] Furthermore, for example, the information processing method according to the third embodiment is the information processing method according to the second embodiment, wherein, with respect to a mathematical model that uses the unpredictability constraint in calculations, the unpredictability constraint is relaxed in a Lagrangian manner, divided into subproblems for each of the two or more demand fluctuation scenarios, and the production quantity is recalculated by adding the penalty term, which may include resolving the subproblems by adding the penalty term.

[0024] This allows for an effective reduction in processing load using Lagrangian relaxation.

[0025] Furthermore, for example, the information processing method according to the fourth embodiment is the information processing method according to the third embodiment, and the mathematical model divided into subproblems may be a mathematical model in which the penalty term is added to the objective function.

[0026] This allows us to reduce processing load by using a mathematical model in which a penalty term is added to the objective function.

[0027] Furthermore, for example, the information processing method relating to the fifth embodiment is an information processing method relating to any one of the second to fourth embodiments, wherein the penalty term may include the difference between the provisional solution of production volume calculated without using the unpredictability constraint in the calculation and the statistical value of the provisional solution corresponding to each of the two or more demand fluctuation scenarios, and a weight.

[0028] This allows us to obtain a solution that satisfies the unpredictability constraint without having to compute the unpredictability constraint itself, by adding penalty terms that include differences and weights.

[0029] Furthermore, for example, the information processing method according to the sixth embodiment is the information processing method according to the fifth embodiment, wherein the weights are calculated by updating the Lagrange multiplier using the subgradient method based on the difference, and the addition of the penalty term and recalculating the production quantity may include recalculating the production quantity using the updated weights.

[0030] This allows the production quantity to be recalculated using weights corresponding to the difference, potentially leading to a more feasible solution.

[0031] Furthermore, for example, the information processing method according to the seventh embodiment may be an information processing method according to the fifth or sixth embodiment, in which the penalty term is added and the production quantity is recalculated repeatedly until the difference becomes zero.

[0032] This ensures that a solution satisfying the unpredictability constraint can be obtained.

[0033] Furthermore, for example, the information processing method according to the eighth embodiment is an information processing method according to any one of the fifth to seventh embodiments, wherein in each of the two or more demand fluctuation scenarios, the integer constraints imposed on the time series forecast are relaxed by LP (Linear Programming) using a capacity scaling method, thereby calculating a provisional solution for the production quantity calculated without using the unpredictability constraint in the calculation.

[0034] This reduces the processing load required to calculate production quantities while considering integer constraints.

[0035] Furthermore, for example, the information processing method according to the ninth embodiment is an information processing method according to the eighth embodiment, wherein the integer constraint includes the predetermined coefficient for calculating the feasibility constraint regarding the execution of production taking an integer value of either a first value or a second value higher than the first value, and the LP relaxation may include relaxing the possible values ​​of the predetermined coefficient to be between the first value and the second value.

[0036] This allows for a reduction in processing load by relaxing the integer constraint on a given coefficient.

[0037] Furthermore, for example, the information processing method according to the tenth embodiment is an information processing method according to the ninth embodiment, wherein the maximum production quantity for producing the target product is updated using the production quantities corresponding to each of the two or more demand fluctuation scenarios obtained by the LP relaxation, a provisional solution is calculated using the updated maximum production quantity, the updating of the maximum production quantity and the calculation of the provisional solution are repeatedly performed until the predetermined coefficient becomes either the first value or the second value, and the provisional solution may be the provisional solution obtained when the predetermined coefficient becomes either the first value or the second value.

[0038] This makes it easier to calculate a provisional solution where a predetermined coefficient is 1, thereby reducing the amount of processing required to calculate a provisional solution that satisfies the integer constraint.

[0039] Furthermore, for example, the information processing method according to the 11th embodiment is an information processing method according to any one of the second to tenth embodiments, and among the plurality of demand fluctuation scenarios, demand fluctuation scenarios with an occurrence probability of a predetermined value or higher may be extracted as the two or more demand fluctuation scenarios.

[0040] This allows for the use of two or more demand fluctuation scenarios with a high probability of occurrence, making it possible to output an effective production plan that takes into account the uncertainty of demand forecasts.

[0041] Furthermore, for example, the information processing method according to the 12th embodiment is an information processing method according to any one of the second to 11th embodiments, and the production plan information may include information for displaying the two or more demand fluctuation scenarios and the two or more occurrence probabilities side by side on the same screen.

[0042] This allows the production plan to be displayed to the user, thereby supporting the user in making decisions about the production plan.

[0043] Furthermore, for example, the information processing method according to the 13th embodiment is an information processing method according to any one of the second to 12th embodiments, and the production plan corresponding to the two or more demand fluctuation scenarios may be displayed in a tree format.

[0044] This allows for an intuitive and easy-to-understand tree-like display, effectively supporting users in making production plan decisions.

[0045] Furthermore, for example, the information processing method according to the 14th embodiment is an information processing method according to any one embodiment of the 1st to 13th embodiments, wherein the demand forecasting model is a Markov model, and the Markov model may be constructed by obtaining the second demand results of the object in the third period prior to the first period, and setting at least one of the transition probability between nodes in the Markov model and the number of branches for the nodes based on the second demand results.

[0046] This makes it possible to automatically build a Markov model as a demand forecasting model.

[0047] Furthermore, the information processing device according to the 15th aspect of this disclosure includes: a performance acquisition unit that acquires a first demand performance including the demand performance of an object during a first period; a demand acquisition unit that takes the demand performance of the object as input and acquires a plurality of demand fluctuation scenarios for a second period following the first period, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, which are output by inputting the first demand performance into a demand forecasting model that has been trained to output a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations of the object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios; and an output unit that outputs production plan information showing a production plan for the object based on the acquired plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

[0048] This produces the same effect as the information processing method described above.

[0049] Furthermore, the program relating to the sixteenth aspect of this disclosure is a program that causes a computer to execute an information processing method relating to any one of the first to fourteenth aspects.

[0050] This produces the same effect as the information processing method described above.

[0051] These general or specific embodiments may be implemented using a system, method, integrated circuit, computer program, or a non-temporary recording medium such as a computer-readable CD-ROM, or any combination of a system, method, integrated circuit, computer program, or recording medium. The program may be pre-stored on the recording medium or supplied to the recording medium via a wide-area communication network, including the Internet.

[0052] The embodiments will be described in detail below with reference to the drawings.

[0053] The embodiments described below are all comprehensive or specific examples. The numerical values, shapes, components, arrangement and connection configurations of components, steps, and the order of steps shown in the following embodiments are examples only and are not intended to limit this disclosure. Furthermore, any components in the following embodiments that are not described in an independent claim will be described as optional components.

[0054] Furthermore, in this specification, terms indicating relationships between elements such as agreement, as well as numerical values ​​and numerical ranges, are not expressions that represent only strict meanings, but also expressions that include substantially equivalent ranges, for example, differences of a few percent (or about 10%).

[0055] Furthermore, in this specification, ordinal numbers such as "first," "second," etc., do not indicate the number or order of components unless otherwise specified, but are used to avoid confusion and to distinguish similar components.

[0056] (Embodiment) The information processing method and the like according to this embodiment will be described below with reference to Figures 1 to 6.

[0057] [1. Configuration of the Information Processing Device] First, the configuration of the information processing device according to this embodiment will be explained with reference to Figures 1 to 2B. Figure 1 is a block diagram showing the configuration of the information processing device 100 according to this embodiment. Note that Figure 1 shows an exemplary functional configuration of the information processing device 100, and the functional configuration of the information processing device 100 is not limited to Figure 1. The information processing device 100 outputs information regarding the PSI plan or cash flow for the second period, which is after the first period, based on the PSI performance data for the first period. In this embodiment, an example of outputting a production plan will be described as an example of a PSI plan. The first period is, for example, a period in the past from the present, and the second period is, for example, a period in the future from the present, but is not limited to these.

[0058] As shown in Figure 1, the information processing device 100 comprises an input unit 10, a processing unit 20, and an output unit 30 as its functional configuration. The information processing device 100 also comprises a processor and memory as its hardware configuration. The memory is ROM (Read Only Memory) and RAM (Random Access Memory), and can store programs executed by the processor. Each function of the information processing device 100 is realized by the processor and other components that execute the programs stored in memory. The information processing device 100 may be realized by a stationary PC (Personal Computer), a mobile terminal such as a smartphone or tablet, a server device, or a combination of two or more of these.

[0059] The input unit 10 acquires various information for outputting a PSI plan (in this case, a production plan). The input unit 10 may be configured to include a communication interface and acquire various information through communication, or it may have a reception unit such as a button, touch panel, or sound collection device and acquire various information by receiving operations from the user. The communication may be wireless or wired.

[0060] Here, an example of the various types of information acquired by the input unit 10 will be explained with reference to Figures 2A and 2B. Figure 2A is a diagram showing an example of past PSI performance data. Figure 2B is a diagram showing an example of information indicating various costs.

[0061] As shown in Figure 2A, the input unit 10 may acquire historical PSI performance data including module number, date, production volume, sales volume, and inventory volume. In the example in Figure 2A, PSI performance data for a total of three months, one month at a time, is shown. Three months is an example of the first period, but the first period is not limited to three months and can be any period appropriate to the product type. The first period may be, for example, one month or four months or more.

[0062] Figure 2A shows historical PSI performance data obtained when forecasting demand using a demand forecasting model. The input unit 10 also obtains historical PSI performance data used when constructing the demand forecasting model. This PSI performance data may include, for example, PSI performance data for a period longer than the first period.

[0063] As shown in Figure 2B, the input unit 10 may acquire information including various costs. In Figure 2B, transportation + cost, inventory storage cost, production cost, and selling price are shown as examples, but the input unit 10 only needs to acquire at least one of the transportation + cost, inventory storage cost, production cost, and selling price. This information can be used to calculate cash flow.

[0064] In this embodiment, the input unit 10 only needs to acquire past PSI performance data as shown in Figure 2A. The input unit 10 may also acquire various parameters used in processing by the processing unit 20. These parameters may include, for example, parameters for constructing a demand forecasting model and parameters for setting constraints. Parameters for constructing a demand forecasting model may include, for example, the number of levels at the branching point. Parameters for setting constraints may include, for example, minimum production volume and maximum production volume, but are not limited to these.

[0065] Referring again to Figure 1, the processing unit 20 performs various processes to predict the production volume included in the production plan. The processing unit 20 includes a demand forecasting model construction unit 21, a demand forecasting unit 22, and an optimization calculation unit 23.

[0066] The demand forecasting model construction unit 21 executes a process to construct a demand forecasting model capable of predicting the time-series information of the amount of demand and its probability of occurrence in the second period, based on the PSI actual data (in this case, at least actual demand) for the first period. The demand forecasting model construction unit 21 constructs a demand forecasting model based on, for example, the PSI actual data for a third period that is longer than the first period. The third period may be, for example, several months or several years.

[0067] In this embodiment, the demand forecasting model construction unit 21 will be described as using a Markov model (demand fluctuation Markov model) as the demand forecasting model, but other mathematical models may also be used.

[0068] The demand forecasting unit 22 uses the demand fluctuation Markov model and pathfinding algorithm constructed by the demand forecasting model construction unit 21 to predict multiple demand fluctuation scenarios for the second period and multiple occurrence probabilities corresponding to each of the multiple demand fluctuation scenarios, based on the PSI actual data (in this case, at least actual demand) for the first period. There is a one-to-one correspondence between each demand fluctuation scenario and one occurrence probability.

[0069] The optimization calculation unit 23 calculates the optimal solution in the PSI plan that satisfies one or more constraints for creating the PSI plan, and performs a process to output the PSI plan based on the optimal solution. In this embodiment, the optimization calculation unit 23 predicts the production volume for each demand fluctuation scenario that satisfies one or more constraints for predicting production volume as the optimal solution. The optimal solution means a solution that satisfies one or more constraints (in this case, at least the production volume).

[0070] Furthermore, since the optimization calculation unit 23 would require a large amount of processing power if it performed predictions while imposing one or more constraints, it first relaxes one or more constraints and calculates a provisional production quantity (a provisional solution) for each demand fluctuation scenario. It then adjusts (recalculates) the provisional production quantity for each demand fluctuation scenario so that it satisfies one or more constraints, thereby calculating the optimal solution. The processing of the optimization calculation unit 23 will be described later with reference to Figures 5 and 6.

[0071] The output unit 30 outputs production plan information based on multiple demand fluctuation scenarios and multiple occurrence probabilities predicted by the demand forecasting unit 22. The production plan information is related to the PSI plan, and may be, for example, a PSI plan or a cash flow. In this embodiment, the production plan information includes at least the production volume for each demand fluctuation scenario.

[0072] The output unit 30 may include a communication interface and transmit production plan information via communication, or it may have a display unit such as an LCD panel or sound output device and display the production plan information.

[0073] [2. Demand Fluctuation Scenarios and Production Plans] Next, the demand forecast using the demand forecasting model configured as described above by the demand forecasting unit 22 will be explained with reference to Figures 3A to 4B. Figure 3A is a diagram showing an example of the uncertainty of the demand forecast according to this embodiment being expressed probabilistically in a scenario tree. In Figure 3A, the horizontal axis represents time, and the vertical axis represents the relationship between the magnitudes of demand. Time t=0 represents the present time, and time t=1 to 3 represents future time. Note that time t may be a specific point in time, or it may be a time with a predetermined range (for example, one day, one month, etc.).

[0074] In Figure 3A, for illustrative purposes, an example is shown where the production volume at the current time is divided into two levels (large and small) for the production volume at the next time point. Large means that the production volume at the next time point will be higher than at the current time point, and small means that the production volume at the next time point will be lower than at the current time point. The number of levels in the branching is set by the user, but it may be a fixed value, or it may be automatically set by the demand forecasting model construction unit 21 based on past performance, etc. Also, Figure 3A shows the probability of occurrence of each demand fluctuation scenario (p1, p2, ...). Large and small are examples of time series information. Large and small also represent nodes in the demand forecasting model. In the example in Figure 3A, one node is divided into two nodes.

[0075] As shown in Figure 3A, at the present time (time t=0), the demand for the following month (time t=1) is at two levels, large and small, relative to the present time. Similarly, at the next month (time t=1), the demand for the month after that (time t=2) is also at two levels, large and small, relative to the month after that. The demand for three months later (time t=3) is also at two levels, large and small, relative to the month after that. For example, the branch point for high demand at time t=1 is at two levels: large and small.

[0076] In this case, there are eight possible demand fluctuation scenarios starting from time t=0. Furthermore, the probability that the demand at the next time point will be high (e.g., a transition probability using a Markov model) and the probability that the demand will be low (e.g., a transition probability using a Markov model) can be calculated from past demand data. For example, taking demand high at time t=1 as an example, the first probability that demand will increase at time t=2 and the second probability that demand will decrease at time t=2 can be calculated from past demand data. In the example in Figure 3A, the sum of the two probabilities is 1. For example, if there is a decreasing trend from high to low demand, the probability that it will remain low can be calculated from past demand data. Using these probabilities, the probability of each of the eight demand fluctuation scenarios occurring can be calculated.

[0077] In this way, the demand forecasting unit 22 uses a demand forecasting model to predict multiple pairs of possible demand fluctuation scenarios and their probabilities for future demand fluctuations that are difficult to predict. By creating a production plan using these multiple demand fluctuation scenarios and their probabilities, it is possible to cover a certain extent of future demand fluctuation scenarios, thus enabling the creation of a production plan that takes into account the uncertainty of demand forecasting. Since it becomes possible to create a production plan that covers a wide range of demand fluctuations, it is possible to maximize the expected value of profits.

[0078] Furthermore, the solution (production quantity) at time = 0 may be the solution that maximizes the expected value of profit at this point for all branches from time t = 1 onwards. Similarly, the solution (production quantity) at time = 1 may be the solution that maximizes the expected value of profit at this point for all branches from time t = 2 onwards.

[0079] Furthermore, the production plan may also change according to the branching of the demand fluctuation scenario. For example, a production plan that maximizes profit for subsequent demand fluctuation scenarios is always derived. If this month's demand is 80 units, for example, if next month's demand is low, instead of suddenly reducing production, it may be best to produce 70 units, considering the probability that demand will recover afterward, as this would yield the highest expected profit. In this way, the demand for the following month may be determined by considering demand in the month after next and beyond. The derivation of the production plan may be performed by the information processing device 100 or by another device.

[0080] Furthermore, the branching is not limited to two levels. Also, the demand fluctuation scenarios are not limited to eight; two or more scenarios are sufficient.

[0081] Furthermore, the tree-like diagram of the demand fluctuation scenario shown in Figure 3A, along with its probability of occurrence, may be output as a production plan and displayed to the user. For example, a production plan corresponding to two or more demand fluctuation scenarios may be displayed in tree format.

[0082] Figure 3B is a diagram showing an example of a production plan including production volume for each demand fluctuation scenario according to this embodiment. In Figure 3B, an example of a production plan output from the information processing device 100 is shown when there are two branching levels, large and small. Figure 3B shows the results of predicting monthly production volume for a three-month period.

[0083] As shown in Figure 3B, the production plan includes a scenario (demand fluctuation scenario), a month, and its probability of occurrence. Thus, the production plan includes production levels for each of the two levels for each month. The production plan may also be presented in tabular form.

[0084] As shown in Figures 3A and 3B, the production plan information may include information for displaying two or more demand fluctuation scenarios and two or more occurrence probabilities on the same screen. The production plan only needs to include production quantities for the two or more demand fluctuation scenarios. Furthermore, the production plan does not need to include occurrence probabilities.

[0085] Figure 4A is a diagram illustrating demand fluctuation scenarios when there are three branching levels. Figure 4B is a diagram showing a list of demand fluctuation scenarios where the probability of occurrence is above a predetermined value.

[0086] As shown in Figure 4A, the branching can be at three levels: large, medium, and small. Medium means that the production volume at the next time point is equal to that at the current time point. In this case, 81 different demand fluctuation scenarios are generated up to time t=4.

[0087] As shown in Figure 4B, a demand forecasting model and a pathfinding algorithm may extract two or more demand fluctuation scenarios from among multiple demand fluctuation scenarios, each with a probability of occurrence above a predetermined value (0.05 or higher in the example in Figure 4B), in order to create a PSI plan. Alternatively, a predetermined number of demand fluctuation scenarios with the highest probability of occurrence may be extracted from among multiple demand fluctuation scenarios, or the user may extract the demand fluctuation scenarios. In this way, a demand forecasting model and a pathfinding algorithm may enumerate demand fluctuation scenarios whose probability of occurrence becomes relatively high when they follow recent demand performance.

[0088] In addition, the "large," "medium," and "small" categories shown in Figure 4B may include specific production volumes. Furthermore, a table containing information on the production volumes (large, medium, small) corresponding to each demand fluctuation scenario shown in Figure 4B, along with their probability of occurrence, may be output as production planning information and displayed to the user.

[0089] Furthermore, if the optimization period is set to six months, with demand fixed at a predetermined level for the first two months and fluctuating at five levels for the remaining four months, the number of demand fluctuation scenarios becomes 625, which would result in an enormous amount of computation and may prevent obtaining an ideal solution.

[0090] [3. Challenges of the mathematical model of the PSI plan to maximize the expected value of profits] As described above, in this embodiment, multiple demand fluctuation scenarios are created. The probability that a demand fluctuation scenario s occurs (corresponding to the occurrence probability shown in Figure 3A, etc.) is p ｓ Let k be the unit price of the object, let variable 1 be the quantity demanded at time t in demand fluctuation scenario s, let variable 2 be the quantity in stock at time t in demand fluctuation scenario s, let variable 3 be the presence or absence of production at time t in demand fluctuation scenario s (an integer variable that is 1 if production is produced and 0 if no production is produced), and let variable 4 be the quantity out of stock at time t in demand fluctuation scenario s. Then the objective function that maximizes the expected value of profit for all demand fluctuation scenarios is expressed by equation 1 below.

[0091]

[0092]

[0093]

[0094]

[0095]

[0096] By solving Equation 1, we can obtain the value of the production quantity that maximizes profit (corresponding to variable 5 shown below) as the solution. Furthermore, the following part of Equation 1 (Equation 2) shows the profit obtained by subtracting various costs (here, inventory storage costs, production costs, and sales losses due to stockouts) from sales.

[0097]

[0098] Variable 2 is calculated using the following equation 3, where the production volume at time t in the demand fluctuation scenario s is defined as variable 5.

[0099]

[0100]

[0101] Note that l represents the production lead time, and s represents the number of demand fluctuation scenarios for which the production plan is created. Also, variable 2 is assumed to be greater than or equal to 0. By solving equation 1, it is possible to obtain the production quantity value that maximizes the expected profit for each demand fluctuation scenario.

[0102] In each demand fluctuation scenario, the following constraints exist.

[0103] If the minimum production quantity per cycle is q and the maximum production quantity is Q, then the feasibility constraint shown in Equation 4 below exists as a constraint on the production capacity of the production equipment. The feasibility constraint is a constraint on the maximum and minimum production quantities.

[0104]

[0105] Here, variable 3 satisfies the following equation 5, which is an integer constraint imposed on time series forecasting in the subproblem. The integer constraint is a constraint imposed on the production of the object, and includes the fact that variable 3, which is used to calculate the feasibility constraint regarding the execution of production, takes either an integer value of 0 (first value) or 1 (second value). Note that variable 3 is an example of a predetermined coefficient.

[0106]

[0107] In other words, in equation 4, the variable 3 (simply u ｔ When the variable 5 (also written as x) is zero, meaning that no production takes place, ｔ (Also written as) becomes zero. Also, in equation 4, the variable u ｔ When indicating that is 1, that is, that production takes place, the quantity produced x ｔ The constraint must be that the value is greater than or equal to the minimum output q and less than or equal to the maximum output Q. ｔ is a variable that can take either 0 or 1, and hereafter it is an integer variable u ｔIt is also described. Here, the time t = 1 indicates the most recent time (e.g., the next month) for creating the production plan, and the time t = T indicates the most future time (e.g., one month out of six months later) for creating the production plan, but is not limited thereto.

[0108] Equation 1 is a mixed integer programming (MIP) problem subject to integer constraints (where u ｔ takes an integer value of 0 or 1) such as such feasibility constraints, so it is NP-hard.

[0109] Also, at the stage of time t, since it is impossible to make a decision (unpredictable) in anticipation of the branching of the first demand fluctuation scenario s １ and the second demand fluctuation scenario s ２ into two different future demand fluctuation scenarios, the demand volume values at time t for the two demand fluctuation scenarios should be the same value. Also, the variable u ｔ should similarly have the same value at time t for the two demand fluctuation scenarios. Since it is the same for the production volume x ｔ , there is an unpredictability constraint shown in Equation 6 below. The unpredictability constraint is a constraint condition that in a plurality of demand fluctuation scenarios, at the same time t, the production volume x ｔ and the variable u ｔ take the same value, and is a constraint condition imposed on the time series prediction of the production volume using the demand fluctuation scenario and the occurrence probability. These unpredictability constraints are shown in Equation 6 below. Time series prediction means predicting two or more production volumes at regular intervals.

[0110]

[0111] Note that B(s, t) is a set of demand fluctuation scenarios whose history up to time t is equal to the demand fluctuation scenario s.

[0112] Equation 1 is mainly subject to two constraint conditions, such integer constraints and unpredictability constraints, so the computational complexity is huge and there is a risk that an ideal solution cannot be obtained.

[0113] Therefore, the information processing device 100 according to this embodiment performs the following information processing method to reduce the computational load while satisfying the two constraints of production quantity x ｔ Obtain it.

[0114] [4. Operation of the Information Processing Device] Next, the operation of the information processing device 100 configured as described above will be explained with reference to Figures 5 and 6. Figure 5 is a flowchart showing the operation (information processing method) of the information processing device 100 according to this embodiment. Steps S10 and S20 shown in Figure 5 represent the process of constructing a demand forecasting model (here, a demand fluctuation Markov model), steps S30 and S40 represent the process of estimating multiple demand fluctuation scenarios and multiple occurrence probabilities, and steps S50 to S90 represent the process of calculating an optimal solution that satisfies two constraints (for example, by simulation).

[0115] As shown in Figure 5, first, the demand forecasting model building unit 21 acquires PSI performance data for the past several years (S10). The demand forecasting model building unit 21 may acquire PSI performance data stored in the storage unit (not shown) of the information processing device 100, or it may acquire PSI performance data via the input unit 10. Note that in step S10, it is sufficient to acquire at least performance data related to demand (second demand performance).

[0116] Several years is just one example of the third period. For example, the third period may be longer than the first period. The period for the acquired PSI performance data is not limited to several years, but may be any period for which a demand forecasting model can be constructed. The storage unit is implemented by, for example, a non-volatile storage device (SSD (Solid State Drive) or HDD (Hard Disk Drive)).

[0117] Next, the demand forecasting model construction unit 21 constructs a demand fluctuation Markov model from the PSI actual data (S20). Constructing a demand fluctuation Markov model includes, for example, setting at least one of the transition probabilities between nodes in the Markov model and the number of branches for the nodes, based on the second demand actual data. The method for constructing the demand fluctuation Markov model from the PSI actual data is not particularly limited, and any known method may be used. A demand fluctuation Markov model is constructed for each type (model) of object. The demand fluctuation Markov model corresponding to the model is constructed from the PSI actual data of that model, but is not limited to this.

[0118] Next, the input unit 10 acquires the demand data for the most recent two months (first demand data) for the specified model to be optimized (S30). The input unit 10 acquires the demand data for the most recent two months from the storage unit of the information processing device 100, or from an external device via the input unit 10. The period for which the demand data is acquired is not limited to the most recent two months; it may be the most recent one month, a period of three months or more, or a period corresponding to the period for which the production plan is created. The input unit 10 functions as a data acquisition unit that acquires demand data. The most recent two months is an example of the first period. The first period is not limited to months; it may be hours, days, weeks, years, etc.

[0119] Next, the demand forecasting unit 22 uses a demand fluctuation Markov model and a pathfinding algorithm to estimate K (K: an integer greater than or equal to 2) demand fluctuation scenarios for the next N months (for example, N is an integer greater than or equal to 1) based on the demand performance of the most recent two months, in order of decreasing probability of occurrence (S40). A demand fluctuation Markov model is constructed for each machine model, and here, a demand fluctuation Markov model corresponding to the machine model to be optimized is used. Also, N months is an example of a second period. Note that the second period is not limited to months, but may be hours, days, weeks, years, etc.

[0120] Furthermore, the demand forecasting unit 22 creates multiple demand fluctuation scenarios from a demand fluctuation Markov model using a pathfinding algorithm. In the pathfinding algorithm, one of the levels is selected at each time point, and the time-series information of the selected level is extracted as a demand fluctuation scenario. The demand forecasting unit 22 also calculates the probability of occurrence of the demand fluctuation scenario from the transition probabilities between nodes that pass through the demand fluctuation scenario, and obtains, for example, K demand fluctuation scenarios with a high probability of occurrence. The demand forecasting unit 22 functions as a demand acquisition unit.

[0121] In step S40, an example was described in which K demand fluctuation scenarios with a high probability of occurrence are extracted from among multiple demand fluctuation scenarios, but this is not limited to this example, and the extraction process does not have to be performed. If the extraction process is not performed, all of the multiple demand fluctuation scenarios become the subject of steps S50 and beyond.

[0122] Next, the optimization calculation unit 23 obtains a lower bound value by annealing (S50). The lower bound value here is the solution (production quantity) for Equation 1, and since the search is performed while the two constraints are satisfied, a relatively good solution (lower bound value) can be obtained among the feasible solutions. A feasible solution is a solution in the feasible region of the search space in which the value of the objective function satisfies the constraints. Annealing is an example of an approximate search method and is also called simulated annealing. Note that the method for obtaining the lower bound value is not limited to annealing; for example, a poor solution or a genetic algorithm may also be used.

[0123] The lower bound is a fixed value and is obtained only once. Furthermore, the lower bound is independent of the demand fluctuation scenario estimated in step S40 and is used when calculating the magnitude of the Lagrange multiplier (weight) described later.

[0124] Next, the optimization calculation unit 23 divides the unpredictability constraint (the mathematical model including the unpredictability constraint) into subproblems for each scenario (demand fluctuation scenario) by Lagrangian relaxation (S60). The optimization calculation unit 23 divides the unpredictability constraint into subproblems for each demand fluctuation scenario by Lagrangian relaxation using the scenario aggregation method. In this specification, Lagrangian relaxation means relaxing (for example, ignoring) the unpredictability constraint.

[0125] The optimization calculation unit 23 uses the following equation 7 as a subproblem for the demand fluctuation scenario s.

[0126]

[0127] F ｓ (X) represents the profit for a demand fluctuation scenario s and is calculated, for example, using Equation 2. X is a feasible solution included in the set of feasible solutions C that satisfy the minimum production quantity, etc.

[0128] The portion of Equation 8 shown in Equation 7 is a term (penalty term) added to the objective function as a result of Lagrangian relaxation (relaxing a portion of the constraints), representing a violation of the constraints. A problem to which a penalty term has been added to the objective function is also called a Lagrangian relaxation problem. Equation 7 can also be described as a mathematical model in which the unpredictability constraint used in the calculation is Lagrangian relaxed, and the model is divided into subproblems for each of two or more demand fluctuation scenarios.

[0129]

[0130] lol ｓｔ This is the Lagrange multiplier. The Lagrange multiplier w ｓｔ This will be discussed later. Using the following equation 9, a solution that satisfies the unpredictability constraint at that point in the calculation process (variable 6 below) is calculated.

[0131]

[0132]

[0133] The solution that satisfies the unpredictability constraint shown in Equation 9 is the production quantity x corresponding to each demand fluctuation scenario.ｓｔ And the probability p of occurrence of the demand fluctuation scenario in question. ｓ It is expressed as the average of the values ​​obtained by multiplying by and, but is not limited to this. Note that A is a set of scenarios in which it is unpredictable for the demand fluctuation scenarios to branch into different scenarios at time t, and includes demand fluctuation scenario s.

[0134] Equation 8 is: Production volume x ｓｔ This formula calculates how far the provisional solution for production volume (calculated without using the unpredictability constraint) deviates from the mean value by taking the difference between (for example, a provisional solution) and the mean value shown in Equation 9, and then calculates the value to be added to the objective function by multiplying it by the Lagrange multiplier. Note that the mean value is just one example of a statistical value. Furthermore, the statistical value is not limited to the mean value; it may also be the median, mode, etc.

[0135] Note that production volume x ｔ If the unpredictability constraint is satisfied, then the integer variable u is necessarily ｔ It is also considered that the unpredictability constraint is satisfied.

[0136] Next, the optimization calculation unit 23 executes the process of solving the integer mixed programming (MIP) problem for each scenario (each demand fluctuation scenario) (S70).

[0137] Step S70 will be explained in detail with reference to Figure 6. Figure 6 is a flowchart showing the detailed operation (information processing method) of step S70 shown in Figure 5. Figure 6 shows the operation of LP mitigation using the capacity scaling method for subproblems (MIP and NP hard) for each demand fluctuation scenario.

[0138] As shown in Figure 6, the optimization calculation unit 23 solves the LP relaxation problem by relaxing the integer constraint (see, for example, equation 10 below) to the relaxation constraint (see, for example, equation 11 below), and the obtained solution is (x ＬＰ u ＬＰ ) (S71). In other words, the optimization calculation unit 23 calculates the integer variable u ｔ Normally, it can only take the values ​​of 0 or 1, but the integer variable u can take values ​​other than integers between 0 and 1 (inclusive). ｔ Relax the constraints on and solve equation 7 (or equation 1). When using equation 7, the Lagrange multiplier w ｓｔThe value may be set to any value (for example, 1), and the term in Equation 8 may be ignored during the calculation. The solution obtained in step S71 includes the production quantities corresponding to each of the two or more demand fluctuation scenarios obtained by LP relaxation.

[0139]

[0140] 0 ≤ u ｔ ≦1...(Formula 11)

[0141] Thus, the optimization calculation unit 23 imposes the relaxation constraint shown in Equation 11 and solves Equation 7 to obtain the solution (x ＬＰ u ＬＰ ) Let the variable u ＬＰ Since can take on decimal values ​​such as 0.2 or 0.8, the solution obtained here does not satisfy the integer constraint.

[0142] Next, the optimization calculation unit 23 calculates the integer variable u ｔ Whether all of them are integers or not, that is, the integer variable u ｔ Determine whether all of them are either 0 or 1 (S72). Integer variable u ｔ This is the variable u for all demand fluctuation scenarios at time t. ＬＰ This includes. Since relaxation constraints are imposed, it is assumed that the determination in step S72 will be No if it is performed for the first time.

[0143] The optimization calculation unit 23 calculates the integer variable u ｔ If it is determined that all of them are not integers (No in S72), then for t∈T that does not satisfy the integer constraint (see equation 12 below), the maximum production quantity Q ｔ The constraint is updated as shown in Equation 13 below (S73). However, Equation 13 satisfies Equation 14 shown below.

[0144]

[0145]

[0146]

[0147] The optimization calculation unit 23 calculates the maximum output Q at time t. ｔ The constraint is set by the following variable 7, which represents the output quantity (output quantity x at time t). ＬＰReplace with ). For example, the variable at time t shown in variable 8 below (the variable u at time t) ＬＰ If the ratio is 0.8 and the production volume is 8000 units, then the maximum production volume Q ｔ The constraint is set to 8000, regardless of the original value.

[0148]

[0149]

[0150] Production volume x ＬＰ This should be a relatively good solution, so the production quantity x of equation 13 that we will solve from now on ｔ The value can also be close to 8000. In that case, the integer variable u ｔ This inevitably becomes 1. By using equation 13, the integer variable u ｔ This makes it easier to obtain a solution of =1.

[0151] Then, the optimization calculation unit 23 calculates the solution (x ＬＰ u ＬＰ ) is obtained. The solution here is (x ＬＰ u ＬＰ ) is shown in equation 13 (x ｔ u ｔ ) The optimization calculation unit 23 calculates the solution (x ＬＰ u ＬＰ ) is the provisional solution (x) calculated by equation 13. ｔ u ｔ It can also be said that it updates to the updated solution (x). Then, the optimization calculation unit 23 calculates the updated solution (x ＬＰ u ＬＰ Using ), the maximum production volume Q ｔ Update the constraints again.

[0152] The optimization calculation unit 23 calculates the integer variable u ｔ Steps S72 and S73 are repeatedly executed until all values ​​become integers (either 0 or 1). The optimization calculation unit 23 also calculates the integer variable u ｔ If it is determined that all values ​​are integers (Yes in S72), proceed to step S74.

[0153] Next, the optimization calculation unit 23 calculates the integer variable u ｔThe original subproblem is solved by fixing the integer variable u to a value that satisfies the integer constraint (S74). The optimization calculation unit 23 calculates the integer variable u, which is an integer (i.e., can take the value of either 0 or 1). ｔ Using this, we solve equation 7. This calculates the production quantity xt for each demand fluctuation scenario that satisfies the integer constraint. The solution (provisional solution) calculated in step S74 is the integer variable u ｔ This is a provisional solution for when the value is either 0 or 1, and it does not satisfy the unpredictability constraint.

[0154] Referring again to Figure 5, the optimization calculation unit 23 then calculates the provisional solution (x) calculated in step S74 of Figure 6. ｔ u ｔ ) determines whether the unpredictability constraint is satisfied (S80). The optimization calculation unit 23 determines whether the same node in each demand fluctuation scenario (x ｔ u ｔ The optimization calculation unit 23 determines whether the production volume x ｔ You may also determine whether the unpredictability constraint is satisfied for that.

[0155] The optimization calculation unit 23 calculates a solution that satisfies the unpredictability constraint (i.e., the production volume at the same node is equal in the demand fluctuation scenario) for the provisional solution using equation 9 above.

[0156] The optimization calculation unit 23 uses equation 9 to determine the production quantity x corresponding to each demand fluctuation scenario. ｓｔ And the probability p of occurrence of the demand fluctuation scenario in question. ｓ The average of the values ​​obtained by multiplying by and is calculated as a solution that satisfies the unpredictability constraint. Production quantity x ｓｔ This is the production volume obtained as the solution to step S74 shown in Figure 6.

[0157] The optimization calculation unit 23 calculates the production volume x corresponding to each demand fluctuation scenario. ｓｔ The unpredictability constraint is satisfied if the solution that satisfies the unpredictability constraint calculated in Equation 9 matches all of the above (for each demand fluctuation scenario and for each time). Satisfying the unpredictability constraint means that the (x) of the same node in each demand fluctuation scenario ｔ u ｔ This means that they match.

[0158] Next, the optimization calculation unit 23 satisfies the unpredictability constraint, that is, the same node in each demand fluctuation scenario (x ｔ u ｔ If the two conditions match (Yes in S80), the provisional solution (x ｔ u ｔ The optimal solution is defined as ) and the optimal solution is output (S90). A determination of Yes in step S80 indicates that the problems (subproblems and LP relaxation problems) for all demand fluctuation scenarios have been solved, that is, a solution that satisfies the unpredictability constraint and integer constraint has been obtained. Step S90 corresponds to outputting production plan information.

[0159] Furthermore, the optimization calculation unit 23 determines that the unpredictability constraint is not satisfied, that is, at least one (x) of the same node in each demand fluctuation scenario. ｔ u ｔ If they do not match (No in S80), the Lagrange multiplier is updated using the subgradient method (S100), and the processing from step S70 onwards is executed.

[0160] In step S100, the optimization calculation unit 23 updates the Lagrange multiplier to be used next using the following subgradient method with the following equation 15, assuming that the variable 9 is the Lagrange multiplier to be used next. ＵＬＢ The improvement from the previous value is used to update the Lagrange multiplier. Note that the initial value of the Lagrange multiplier may be set beforehand.

[0161]

[0162]

[0163] Hereafter, the Lagrange multiplier calculated in Equation 15 is w ｓｔ It is also written as: Lagrange multiplier w ｓｔ It has a scaling function that converts units of production volume into units of money. Lagrange multiplier w ｓｔ This is updated each time step S100 is executed.

[0164] Here, the current Lagrange multiplier is the following variable 10, and λ represents the update coefficient of the Lagrange multiplier. Specifically, λ indicates how much better the current solution is compared to the lower bound value F ＵＬＢ and how much it affects (how much importance is attached to) the update of the Lagrange multiplier. λ is, for example, set by the user. F ＵＢ represents the upper bound value, and specifically, it represents the value of the objective function at the best solution at that time. Initially, a huge number is given, and the value is updated every time a good solution appears during the repeated calculations. F ＵＬＢ represents the lower bound value obtained in step S50, and r(w) is the subgradient vector shown in the following equation 16.

[0165]

[0166]

[0167] Note that the method for calculating the Lagrange multiplier is not limited to the subgradient method, and other known methods may be used.

[0168] Next, the optimization calculation unit 23 proceeds to step S70 and executes the process of step S70 again using the updated equation 8 for the Lagrange multiplier w ｓｔ The optimization calculation unit 23 repeatedly executes steps S70 and S100 until it is determined as Yes in step S80 (for example, until the difference becomes zero).

[0169] Thus, the optimization calculation unit 23 calculates the production volume while ignoring (relaxing) the predictability constraint, and then calculates how much it violates the predictability constraint (for example, refer to equation 8), adds it to the objective function as a penalty term, and repeats solving the problem (recalculating the production volume) again. The penalty term is calculated every time step S70 is executed.

[0170] As a result, the production volume x, which is the solution at the current time ｓｔThis method allows us to approach a solution that satisfies the unpredictability constraint, thus obtaining a solution that satisfies the unpredictability constraint even though the unpredictability constraint is not used in the calculation. Furthermore, since the calculation is performed using Lagrange relaxation and LP relaxation, meaning that the unpredictability constraint and integer constraint are not used in the calculation, the amount of computation can be reduced compared to when the unpredictability constraint and integer constraint are used in the calculation. In addition, since multiple demand fluctuation scenarios are considered (for example, the production quantity for one demand fluctuation scenario is affected by the production quantities for one or more other demand fluctuation scenarios), it is possible to calculate a production quantity that can withstand fluctuations in actual demand to some extent (suppressing the occurrence of large losses).

[0171] (Other Embodiments) Although information processing methods, etc., relating to one or more embodiments have been described above based on embodiments, this disclosure is not limited to these embodiments. Without departing from the spirit of this disclosure, various modifications that a person skilled in the art could conceive of may be applied to these embodiments, and forms constructed by combining components from different embodiments may also be included in this disclosure.

[0172] For example, in step S73 shown in Figure 6 above, production quantity x ｔ The maximum output per cycle Q decreases monotonically. ｔ To prevent the result from being less than or equal to the optimal solution, (i) the relaxation constraint shown in Equation 11 is set to 0 ≤ u ｔ Change to (ii) Equation 14 to the following Equation 17, (iii) x ｔ ≦Q （ｉｎｉｔｉａｌ） At least one of the following operations may be performed: adding the following to the constraints.

[0173]

[0174] Q （ｉｎｉｔｉａｌ） This indicates the original maximum production volume.

[0175] Furthermore, while the above embodiment describes an example in which the demand acquisition unit acquires multiple demand fluctuation scenarios and multiple occurrence probabilities for each of the multiple demand fluctuation scenarios by inputting actual demand data into a demand forecasting model, the invention is not limited to this example. Another device may calculate multiple demand fluctuation scenarios and multiple occurrence probabilities calculated by inputting actual demand data into a demand forecasting model, and acquire multiple demand fluctuation scenarios and multiple occurrence probabilities (or one or more demand fluctuation scenarios and one or more occurrence probabilities) through that other device.

[0176] Furthermore, in the above embodiment, each component may be implemented by being composed of dedicated hardware or by executing a software program suitable for each component. Each component may also be implemented by a program execution unit such as a CPU or processor reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory.

[0177] Furthermore, the order in which each step in the flowchart is performed is illustrative for the purpose of specifically illustrating this disclosure, and may be in a different order. Also, some of the above steps may be performed simultaneously (in parallel) with other steps, and some of the above steps may not be performed.

[0178] Furthermore, the division of functional blocks in the block diagram is just one example; multiple functional blocks can be implemented as a single functional block, a single functional block can be divided into multiple parts, or some functions can be moved to other functional blocks. In addition, the functions of multiple functional blocks with similar functions can be processed in parallel or time-sharing by a single piece of hardware or software.

[0179] Furthermore, the information processing device according to the above embodiments may be implemented as a single device or as a plurality of devices (for example, an information processing system). When the information processing device is implemented as a plurality of devices, the components of the information processing device may be distributed among the plurality of devices in any manner. When the information processing device is implemented as a plurality of devices, the method of communication between the plurality of devices is not particularly limited and may be wireless communication or wired communication. In addition, wireless communication and wired communication may be combined between the devices.

[0180] Furthermore, each component described in the above embodiment may be implemented as software, or typically as an integrated circuit (LSI). These may be individually integrated onto a single chip, or some or all of them may be integrated onto a single chip. Here, we refer to it as an LSI, but depending on the degree of integration, they may also be called ICs, system LSIs, super LSIs, or ultra LSIs. Moreover, the method of integrated circuit implementation is not limited to LSIs; it may also be implemented using dedicated circuits (general-purpose circuits that execute dedicated programs) or general-purpose processors. After LSI manufacturing, a programmable FPGA (Field Programmable Gate Array) or a reconfigurable processor that can reconfigure the connections or settings of circuit cells inside the LSI may be used. Furthermore, if an integrated circuit implementation technology that replaces LSIs emerges due to advances in semiconductor technology or other derived technologies, it is natural that the components may be integrated using that technology.

[0181] A system LSI is a highly functional LSI manufactured by integrating multiple processing units onto a single chip. Specifically, it is a computer system composed of a microprocessor, ROM, RAM, and other components. The ROM stores the computer program. The system LSI achieves its function by having the microprocessor operate according to the computer program.

[0182] Furthermore, one aspect of this disclosure may be a computer program that causes a computer to perform each characteristic step included in the information processing method shown in either Figure 5 or Figure 6.

[0183] Furthermore, for example, the program may be a program to be executed by a computer. Also, in one aspect of this disclosure, such a program may be recorded on a computer-readable non-temporary recording medium. For example, such a program may be recorded on a recording medium and distributed or made available. For example, by installing the distributed program on a device having another processor and having that processor execute the program, it becomes possible to have that device perform the above-mentioned processes.

[0184] This disclosure is useful for information processing devices, etc., related to the creation of PSI plans.

[0185] 10 Input unit (performance acquisition unit) 20 Processing unit 21 Demand forecasting model construction unit 22 Demand forecasting unit (demand acquisition unit) 23 Optimization calculation unit 30 Output unit 100 Information processing device

Claims

1. An information processing method that obtains a first demand record including the actual demand for the target object during the first period, inputs the first demand record into a demand forecasting model that takes the actual demand for the target object as input and outputs a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations for the target object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, outputs a plurality of demand fluctuation scenarios for the second period following the first period, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, and outputs production plan information showing the production plan for the target object based on the obtained plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

2. The information processing method according to claim 1, comprising: relaxing the unpredictability constraint imposed on a time series forecast of production volume using two or more demand fluctuation scenarios from among the plurality of demand fluctuation scenarios, and two or more occurrence probabilities for each of the two or more demand fluctuation scenarios, calculating the production volume; calculating the extent to which the unpredictability constraint is violated; and adding a penalty term corresponding to the relaxation of the unpredictability constraint to the objective function for calculating the production plan of the object, and recalculating the production volume.

3. The information processing method according to claim 2, wherein the unpredictability constraint is relaxed in a mathematical model used in calculations, the unpredictability constraint is divided into subproblems for each of the two or more demand fluctuation scenarios, and the production quantity is recalculated by adding the penalty term, which includes resolving the subproblems by adding the penalty term.

4. The information processing method according to claim 3, wherein the mathematical model divided into subproblems is a mathematical model in which the penalty term is added to the objective function.

5. The information processing method according to any one of claims 2 to 4, wherein the penalty term is a term that includes the difference between a provisional solution of production volume calculated without using the unpredictability constraint in the calculation and the statistical value of the provisional solution corresponding to each of the two or more demand fluctuation scenarios, and a weight.

6. The information processing method according to claim 5, wherein the weights are calculated by updating the Lagrange multipliers using a subgradient method based on the difference, and the addition of the penalty term and recalculating the production quantity again includes recalculating the production quantity again using the updated weights.

7. The information processing method according to claim 5, which involves repeatedly adding the penalty term and recalculating the production quantity until the difference becomes zero.

8. The information processing method according to claim 3 or 4, wherein, in each of the two or more demand fluctuation scenarios, the integer constraints imposed on the time series forecast are relaxed by Linear Programming (LP) using a capacity scaling method, thereby calculating a provisional solution for the production quantity calculated without using the unpredictability constraint in the calculation.

9. The information processing method according to claim 8, wherein the integer constraint includes that a predetermined coefficient for calculating feasibility constraints relating to the execution of production takes an integer value of either a first value or a second value higher than the first value, and the LP relaxation includes relaxing the possible values ​​of the predetermined coefficient to be between the first value and the second value.

10. The information processing method according to claim 9, wherein the maximum production quantity for producing the target product is updated using the production quantities corresponding to each of the two or more demand fluctuation scenarios obtained by the LP relaxation, a provisional solution is calculated using the updated maximum production quantity, the updating of the maximum production quantity and the calculation of the provisional solution are repeatedly performed until the predetermined coefficient becomes either the first value or the second value, and the provisional solution is the provisional solution obtained when the predetermined coefficient becomes either the first value or the second value.

11. The information processing method according to any one of claims 2 to 4, wherein demand fluctuation scenarios with an occurrence probability of a predetermined value or higher are extracted from the plurality of demand fluctuation scenarios as two or more demand fluctuation scenarios.

12. The information processing method according to any one of claims 2 to 4, wherein the production plan information includes information for displaying the two or more demand fluctuation scenarios and the two or more occurrence probabilities side by side on the same screen.

13. The information processing method according to any one of claims 2 to 4, wherein the production plan corresponding to the two or more demand fluctuation scenarios is displayed in a tree format.

14. The information processing method according to any one of claims 1 to 4, wherein the demand forecasting model is a Markov model, and the method involves obtaining a second demand record for the object in a third period prior to the first period, and constructing the Markov model by setting at least one of the transition probabilities between nodes in the Markov model and the number of branches for the nodes based on the second demand record.

15. Information processing device comprising: a performance acquisition unit that acquires a first demand performance including the demand performance of an object during a first period; a demand acquisition unit that takes the demand performance of the object as input and acquires a plurality of demand fluctuation scenarios for a second period following the first period, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios, which are output by inputting the first demand performance into a demand forecasting model that is trained to output a plurality of demand fluctuation scenarios showing time-series information of demand fluctuations of the object, and a plurality of occurrence probabilities for each of the plurality of demand fluctuation scenarios; and an output unit that outputs production plan information showing a production plan for the object based on the acquired plurality of demand fluctuation scenarios and the plurality of occurrence probabilities.

16. A program for causing a computer to execute the information processing method described in any one of claims 1 to 4.

Images