Usage fee calculation system

Abstract

We provide a fee calculation system that calculates the usage fee for a system that determines whether or not equipment and molds can be disposed of. [Solution] The usage fee calculation system calculates the amount that a user pays to the operator as a usage fee for a determination system that determines whether or not to discard equipment or molds for manufacturing replacement parts for automobiles. The determination system includes a means for calculating the probability of unexpected demand occurring from the part number assigned to the replacement part using a formula included in a trained model, determining that no unexpected demand will occur if the probability is smaller than a threshold probability value included in the trained model, and estimating the profit the user will receive if the determination is correct. The system calculates the range within which the usage fee is valid such that the operator's first profit / loss is not negative, and the second profit / loss based on the profit the user receives is not negative, by simulation using an inventory trend database.

Description

[Technical Field]

[0001] This invention relates to a system for calculating usage fees. [Background technology]

[0002] Patent Document 1 discloses a parts management system that determines the discontinuation of parts supply in a form specified for a particular sales unit in response to the arrival of the discontinuation period for sales. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2003-114712 [Overview of the project] [Problems that the invention aims to solve]

[0004] Conventional technologies do not disclose specific steps or functions for determining whether equipment or molds should be discarded, making it difficult to determine this using computer resources.

[0005] This disclosure is made to solve such problems and provides a fee calculation system for calculating the usage fee for a determination system that determines whether or not equipment or molds can be disposed of. [Means for solving the problem]

[0006] The fee calculation system relating to this disclosure is a fee calculation system that calculates the amount payable by a user to an operator as a fee for using a determination system that determines whether or not to dispose of equipment or molds, wherein the equipment or molds are used to manufacture replacement parts for automobiles, and the determination system includes means for calculating the probability of unexpected demand occurring from the part number assigned to the replacement parts using a formula included in a trained model, and means for comparing the probability with a threshold probability value included in the trained model, wherein if the probability is smaller than the threshold probability value, the system predicts that the unexpected demand will not occur and disposes of the equipment or molds. The system outputs information indicating that it is possible to do so, and if the probability is greater than or equal to the threshold probability value, it outputs information indicating that it is predicted that the unexpected demand will occur and that the equipment or model will be maintained. The usage fee calculation system includes means for estimating the profit the user will receive if the prediction result by the judgment system is correct, and an inventory trend database showing the changes in the inventory of each of the multiple replacement parts. The system calculates the range within which the usage fee is established such that the first profit or loss for the operator is not negative, and the second profit or loss based on the profit the user receives is not negative, by simulation using the inventory trend database. [Effects of the Invention]

[0007] According to this disclosure, a fee calculation system can be provided that calculates the usage fee for a system that determines whether or not equipment or molds can be disposed of. [Brief explanation of the drawing]

[0008] [Figure 1] This is a diagram illustrating the usage fee calculation system according to Embodiment 1. [Figure 2] This is a diagram illustrating the determination system according to Embodiment 1. [Figure 3] This is a diagram to explain the data on replacement parts. [Figure 4] This is a diagram illustrating how to set threshold probability values. [Figure 5] This diagram illustrates the set of input values. [Figure 6]This is a diagram for explaining the setting range of the usage fee.

Embodiment for Implementing the Invention

[0009] <Usage Fee Calculation System> Referring to FIG. 1, the usage fee calculation system takes the loss compensation amount \L and the user profit \P as input values (1), and outputs the amount that the user pays to the operator as the usage fee \C of the determination system (1) for determining the availability of equipment or disposal (2). The function of the usage calculation system may be realized by the processor executing a program read into the memory. \L is determined based on the cost assumed when reproducing production equipment or molds due to the prediction of the determination system (10) being off. \P is determined based on the profit due to the prediction of the determination system (10) being correct and the storage cost of the discarded production equipment or molds being suppressed. In the operator profit and loss calculation unit (3), the operator profit and loss \U , i , for each individual user A i is calculated by the following formula (A). \U i = \C × N i - \L × O i - \C × N i × β … Formula (A) Here, N i : Number of determination item numbers, O i : Number of item numbers with incorrect predictions among the item numbers determined to be disposable, β: Ratio of distribution and management costs. The subscript i corresponds to the subscript of user A i . \C, \L, \P, etc. without subscripts represent uniform values.

[0010] In the user profit and loss calculation unit (4), the user profit and loss \R i for each individual user A i is calculated by the following formula (B). \R i = \L × O i - (Unexpected demand response cost) + \P × H i - \C × N i ≈ \P × H i - \C × N <00​​​​: Represents the number of product numbers that were deemed desalable and for which the prediction was correct. The cost of responding to unexpected demand is generally the same as the amount of loss compensation. The provisional setting value for the usage fee C is determined by the usage fee calculation system (5). The provisional setting value may be read from a pre-prepared initial value array or entered manually.

[0011] As part of the supply inventory trend database D(6), simulation dataset Df i (7) is randomly selected n times and subjected to the profit and loss simulation (8) each time. The number of repetitions n is the total number of users in one set (e.g., 10) for which the total profit and loss will be calculated later. n can be any number and may be set appropriately depending on the desired accuracy of the results.

[0012] Df i (7) N is labeled with the result of whether or not unexpected demand occurred. i It consists of individual part number information (9). The determination system (10) determines whether the production system can be "abandoned" or "maintained". i By comparing the individual judgment results (11) with the results of whether or not unexpected demand occurred, the number of prediction errors was O i and the number of times the prediction was correct H i This is required. Using the loss compensation amount \L, user profit \P, and provisional usage fee value \C, \U i This is calculated using the above formula (A), \R i This is calculated using the above formula (B). The calculation result is \U i and \R i User A i It is linked to and sent to the profit and loss calculation unit (13) (internal memory) for each user each time. i and \R i The calculation is repeated n times, and n \U i \R i The array is generated in the per-user profit and loss calculation unit (13). From this, the operator's average profit and loss is calculated. ave The average profit / loss of users is calculated using the following formula (C), and R ave This is calculated using the following formula (D). U ave =Σ\Ui / n (i=1~n)…Formula (C) \R ave =Σ\R i / n (i=1~n)…Formula (D)

[0013] If the conditions for setting the usage fee \C defined by the usage fee \C determination algorithm (15) are met by the provisional setting value of usage fee \C, the provisional setting value is determined to be valid and stored as the valid usage fee \C (16). Alternatively, the comments "valid" or "invalid" may be output. Until it is determined that the calculation of all valid usage fees \C has been completed (17), calculations for other provisional setting values ​​of usage fee \C are performed and the range of valid usage fees \C from the determination system (10) is output (2).

[0014] Next, we will explain the outline of the five usage fee determination algorithms (15). <How to determine \C, Part 1> Let C (Charge) be the provisional setting value for the usage fee. Let P (Profit) be the amount expected to be the profit the user will receive from the disposal of the mold. Let L (Loss) be the amount the operator will compensate the user for losses as an unexpected demand response cost for product numbers whose predictions were incorrect. Let β be the setting value for the ratio of selling, general and administrative expenses to sales (e.g., 1, 20%). These are set as one set, and ave ≥0 and \R ave If the condition ≥ 0 is met, "Valid" is output; otherwise, "Invalid" is output.

[0015] The operator can determine whether a provisional usage fee of C set for a given P, L, and β is "effective" or "ineffective" if both their own operating profit and loss and the average profit and loss of all users are positive, and can use this information as a reference when setting C.

[0016] <How to determine \C, part 2> Multiple patterns of \L and \P, that is, \L j and \P k The following will be set. Individual usage fees for each pattern. jk The average profit / loss for operators is set. ave ≥0 and average user profit / loss\R ave ≥ 0jk The range is output. Furthermore, \U ave ≥0 and \R ave Usage fee \C for which ≥ 0 holds true. jk For patterns where \U does not exist, ave <0, or \R ave Information indicating <0 is output. Furthermore, \C jk If there is a range in which the following holds true, ave and \R ave The value of \C at which equilibrium occurs is output for each pattern. Furthermore, various \L j and \P k From, \C jk The coefficients and intercept values ​​for multiple regression analysis, which can be used for prediction, are output. Preferably, the above output is done for various settings of the selling, general and administrative expenses-to-sales ratio β (e.g., 1-20%).

[0017] In this case, the operator can flexibly set various usage fees \C to accommodate various user needs (various \L and \P). That is, the amount of \P: profits obtained from mold disposal and the amount expected to be obtained differ depending on the user, and the amount of compensation the operator will provide to the user as unexpected demand handling costs for part numbers where the operator's prediction was wrong also differs. In the case of molds, this is because the cost of remanufacturing a discarded mold differs depending on its size and complexity. Also, if storage costs and mold management costs are high, the profits obtained from mold disposal will also be high. If the range in which \C is valid for such various cases is output, a win-win operation can be achieved depending on the combination of \L and \P. Furthermore, if there is no range in which \C is valid, the operator can avoid setting it. In addition, the operator can easily set \C by referring to the outputted equilibrium point value. Moreover, \C can be easily derived from the equation consisting of the coefficients and intercept of the multiple regression analysis, along with \L and \P.

[0018] <How to determine \C, part 3> \L j and \P k This is set. When the same \C is set for all patterns, \L j and \P k For all combinations \U ave The sum of these is the combined average profit / loss of the operators Σ\U aveAs, ave The sum of the user average profit / loss is Σ\R ave Let's assume that Σ\U ave ≥0 and Σ\R ave Output the range of \C such that ≥0, Σ(j=1~s), (k=1~t).

[0019] Several patterns of \L and \P are set depending on the user, and the same usage fee \C is charged to all users in all patterns. This ensures that the operator can secure a profit from the combined value of all patterns, and all users also receive a positive profit from the overall combined value. Because the usage fee \C is the same for all users, it is easy for users to understand the usage fee \C.

[0020] <How to determine \C, part 4> Similar to method 2, j and \P k This is set. Mimicking the concept of property insurance, j proportional to \C j This is set for all combinations. ave The sum of Σ\U ave ≥0 and \R ave The sum of Σ\R ave A usage fee array C such that ≥ 0 jk The range in which the condition is met will be output.

[0021] Since the amount of compensation (¥C) is set according to the degree of risk for which losses will be compensated, users feel that paying ¥C is acceptable. The operator can meet the needs of all users without incurring overall losses.

[0022] <How to determine \C, part 5> Similar to method 2, j and \P k This is set. \C for each pattern jk This is set. Method 2 is \C jk For patterns where the range of validity does not exist, \C jk Set Σ\U ave ≥0 and Σ\R ave A usage fee array C such that ≥ 0 jkOutput one or more. C jk is the combination array (\L j , \P k ).

[0023] For the pattern where the usage fee could not be set in the second determination method, \C jk is set, and it becomes a win-win situation for the user and the operator regarding the combined value.

[0024] <Determination system> Referring to FIG. 2, the part number attached to the replenishment part to be predicted is input to the computer (21). The characteristics of the part associated with the part number are generated and set as a plurality of explanatory variables (22). Referring to FIG. 3, the explanatory variables include, for example, vehicle type (classified into three categories: A, B, and C), part category (classified into seven categories) determined from the upper digit of the part number, two classifications of whether it is an Assy or Parts, two classifications of whether the overseas ratio is high or not, two classifications of whether the attenuation rate (the ratio of the inventory quantity to the previous year) is high or not, two classifications of whether it is a design part or a functional part, and two classifications of whether the part is manufactured using a mold or not. Note that for explanatory variables with three or more classifications, one level can be excluded from the explanatory variables as a reference. For example, when removing the engine of vehicle type A and part category, the weighting of the levels of other explanatory variables is interpreted as relative to the reference. For example, the explanatory variables of part number 12345-70030 indicate information such as vehicle type B, engine, Assy, parts with a low overseas ratio, parts with a low attenuation rate, functional parts, and parts other than those manufactured using a mold.

[0025] Referring back to FIG. 2, from the equation for calculating the probability that an "unexpected demand" included in the learned model occurs, the linear predictor y and the probability p that an unexpected demand occurs i are calculated by the following equation (E) (23). p i = Pr(Y i = 1|X) = 1 / (1 + e -y ) = 1 / (1 + e<able>)… Equation (E) y is called a linear predictor. For example, in the case of part number 12345-70030 in FIG. 3, y = -0.371 + 0 + 0.534 - 2.233 = -1.699 is calculated, and pi = 1 / (1 + e -(-1.699) ) = 1 / (1 + 2.7813 1.699 ) is calculated to be 0.155.

[0026] Referring to FIG. 2 again, the threshold probability value (P th = 0.093) in the learned model is read (24). The calculated value of the probability of occurrence of unexpected demand is compared with the threshold probability value (25). If the calculated value is smaller than the threshold probability value, a comment indicating that the production system can be discarded is output (26). If the calculated value is larger than the threshold probability value, a comment indicating that the production system is to be maintained is output (27). For example, when p i = 0.155 > p th = 0.093, a comment indicating that the production system is to be maintained is output.

[0027] Next, a training method for the learned model will be described. First, a training dataset is prepared. The training dataset is composed of data of 437 replenishment parts that generally cover all part categories for vehicle types (A, B, C). These parts are produced collectively under a mass production system and supplied from inventory. In this dataset, the quantity produced collectively and the time-series data of the inventory number change after being supplied from inventory (inventory balance per year, overseas order ratio) are linked to the part numbers of the replenishment parts. Also, as explanatory variables, two classifications are set: whether the decay rate (ratio of the inventory number to the previous year) is high or not, and whether the overseas ratio is high or not.

[0028] The quantity produced collectively is determined by the following method. First, the shipment number change data of the corresponding part number in the past is obtained from the database. Then, from the parameter table obtained from past empirical rules, the predicted decay rate β based on the shipment number change data and the group to which the corresponding part number belongs is obtained. Then, mass production is carried out according to the predicted lifetime demand quantity using β.

[0029] Using the method already described, multiple explanatory variables are set from the product code. Next, a dependent variable is associated with each product code. The dependent variable corresponds to the training answer and is set to 1 if unexpected demand occurs, and 0 otherwise. Unexpected demand means that the actual demand for each product code exceeds the total production volume within a certain period. For example, if the latest inventory is zero, if extrapolation of past inventory trends predicts that the inventory will be zero within the next year, or if the annual inventory decrease rate falls within the top 5% of all product codes, the dependent variable is set to 1; otherwise, the dependent variable is set to 0. Although not shown in the diagram, the dependent variable for product codes 11123-70030, 28113-70020, and 42450-14030 is 0, the dependent variable for product code 74231-14030 is 1 because the inventory is predicted to be zero within the next year, and the dependent variable for product code 84560-20020 is 1 because the latest inventory is zero.

[0030] The trained model is obtained using logistic regression with maximum likelihood estimation, a type of machine learning. Logistic regression can be performed on a computer, for example, by selecting a generalized linear model from R Commander using the statistical software R. The input formula is, for example, glm(formula=X1 is unexpected demand ~ X1 vehicle type B + X1 vehicle type C + X2. fuel + X3. transmission + X4. axle + X567. body + X89. lamps radio + X89. electrical + X0 function.1 design + X1 is molded. not + X0 Parts.1 Assy + X1 overseas many + X1 attenuation rate, family=binomial(logit), data=Dataset). A binomial (binomial) link function family is selected, and logit (natural logarithm of odds of probability p) is selected as the link function. The R calculation shows the weighting coefficients (X0~89) and the intercept corresponding to each explanatory variable. From each explanatory variable, its corresponding weighting coefficient (X0~89), and the intercept, we get the linear predictor y = α + β1*x1 + … + β k *x k And p i = 1 / (1+e -y The probability of unexpected demand occurring due to ) is pi This is required.

[0031] In Figure 4, the horizontal axis represents the value of the linear predictor, and the vertical axis represents the probability of unexpected demand occurring. If the threshold probability value is set to 0.10, the corresponding threshold for the linear predictor will be approximately -2.2. In this case, it is determined that the production system for product numbers where the linear predictor value lies to the left of the vertical line in Figure 4 can be abandoned. If unexpected demand occurs for some product numbers among those where the linear predictor is less than -2.2, there is a risk of significant losses.

[0032] Let's consider plotting points on Figure 4 indicating whether or not unexpected demand occurs for each spare part. The value on the horizontal axis of each point will be the value of the linear predictor for that spare part. The value on the vertical axis of each point will be 1 if unexpected demand occurs for that spare part, and 0 otherwise. In this case, each spare part will be associated with one of the regions A11, A12, A21, and A22. Unexpected demand occurs for spare parts associated with regions A21 and A22. For spare parts associated with regions A11 and A21, information will be output indicating that the production system can be abandoned. For spare parts associated with regions A12 and A22, information will be output indicating that the production system should be maintained.

[0033] A result corresponding to regions A11 and A12 is called negative, and a result corresponding to regions A21 and A22 is called positive. A result corresponding to region A11 is called a true negative, a result corresponding to region A12 is called a false positive, a result corresponding to region A21 is called a false negative, and a result corresponding to region A22 is called a true positive.

[0034] The loss function provides a profit for each replacement part that is correctly judged and a loss for each replacement part that is incorrectly judged. The loss function consists of unexpected demand costs, mold storage costs, mold disposal profits, and profits from anticipating unexpected demand. Unexpected demand costs are expressed as (-A)*(false negatives). If the production system is a mold, -A represents the cost of reproducing the mold. Mold storage costs are expressed as (-B)*(true positives + false positives). -B represents the cost required to store the mold. Mold disposal profits are expressed as (+C)*(true negatives + false negatives). +C represents the profit obtained from selling scrap when the mold is disposed of. Profits from anticipating unexpected demand are expressed as (+D)*(true positives). +D represents the profit obtained from selling replacement parts in response to unexpected demand. The judgment system (10) calculates the total loss amount for all linear predictor values. The judgment system (10) then sets the probability of the linear predictor with the smallest total loss to a threshold probability value.

[0035] <Verification Results> The inventors calculated the total loss for each linear predictor value for both the training data and the validation data. For the training data, the total loss was smallest when the predictor value was (-2.27). For the validation data, the total loss when the predictor value was (-2.27) was 16 million yen, which is less than the total loss of 19 million yen if no action was taken, thus demonstrating that the total loss can be suppressed. It was also found that the loss could be further suppressed by setting the linear predictor to (-1.8).

[0036] In the training data, when the predictor was (-2.27), it was determined that production could be discontinued for 63 product numbers. Of these, the prediction was wrong for 3 product numbers, and unexpected demand actually occurred. On the other hand, the number of product numbers that were judged to maintain production was 323. This indicates that the benefit of discarding the molds for 60 (=63-3) product numbers outweighs the loss from the 3 data points where the prediction was wrong. In contrast, in the validation data, it was determined that production could be discontinued for 8 product numbers, and production was maintained for 43 product numbers. Unexpected demand occurred for 0 product numbers. Since a benefit was generated from discarding the molds for 8 product numbers, it can be seen that the method is effective.

[0037] Next, we will explain the results of verifying the effectiveness of the trained model. The inventors verified the effectiveness of the trained model using k-fold (k=10) cross-validation with random sampling. A dataset consisting of 437 product codes was divided into training data and validation data in approximately a 9:1 ratio. Here, random numbers between 1 and 100 were assigned to the 437 product codes, and the product codes assigned between 1 and 10 were used as the validation data. The weighting coefficients and threshold probability values ​​of the trained model were calculated using the training data, and the effect of using this trained model to suppress the total loss amount was verified. According to the verification results, the profit and loss improved in 9 out of 10 trials. Furthermore, the average loss improvement amount over the 10 trials was ¥1,175,600, confirming the effectiveness of the trained model.

[0038] <Examples> The following explains the examples of methods 1 to 5 for determining \C mentioned above.

[0039] <How to determine \C, Part 1: Example> The results of the above 10 verifications, Val1 to 10, are shown in Figure 1, Df. i (7) Let's assume that Val1 to 10 are different users A01 to 10. iLet's assume the result is that a determination of the individual part numbers was requested. The profit P that the user can obtain by discarding the molds is assumed to be 450,000 yen, which is the sum of the profit of not having to store the molds in the future (400,000 yen) and the profit of selling the molds (50,000 yen). The amount assumed to be the cost of responding to unexpected demand (L) is assumed to be 8,000,000 yen (per part number). The usage fee (C) is assumed to be 50,000 yen. The calculation result of the profit and loss improvement from the user's perspective, without considering C, was 1,175,600 yen. In Val08, a huge loss of -9,844,000 yen occurred, but this is because two unexpected demands occurred for product numbers, and the cost of responding to them exceeded the cost of discarding the molds.

[0040] Next, we consider the operator who collects C and calculate the profit and loss for both the operator and the user. For example, if the number of verification data for Val01 is 39, then the usage fee that user A01 pays to the operator is C*N. i = 50,000 * 39 = 1,950,000. On the other hand, the benefit received by user A01 is the amount that is assumed to be the benefit obtained from the disposal of the mold, and from the above formula (B) ... i ≒\P*H i -\C*N i =(¥400,000+¥50,000)*2-¥1,950,000=-¥1,050,000. The user's profit / loss ratio (=profit / loss / total usage fee) is -¥1,050,000 / ¥1,950,000=-54%, meaning they incurred a loss. However, looking at users A01~A010, the ratio of users with negative profit / loss ratios to those with positive profit / loss ratios is 5:5, and the average user profit / loss is R ave The average profit / loss per user is +16%. ave is, \R ave =Σ(\Ri)≒Σ(\P*Hi-\C*Ni)=\27,900,000-\24,000,000=\3,900,000, and the average user profit / loss is \3,900,000 / \24,000,000=16%. And the operator's profit / loss is \U ave =Σ(\C*Ni-\C*Ni*β-\L*Oi), and if the selling and administrative expenses sales ratio β=20%, then \U ave=\24,000,000 - \24,000,000 * 20% - 16,000,000 = \3,200,000. The average profit / loss ratio for operators was \3,200,000 / \24,000,000 = 15%. Therefore, when \C=50,000 and β=20%, both users and operators are in a win-win situation, and the output indicates that the provisional setting value of \C is valid.

[0041] <How to determine \C, Part 2: Examples> Twenty combinations of \P and \L were set corresponding to B11-B14, B21-B24, B32-B33, B41-B44, B52-B54, and B62-B64 in Figure 5. For each combination, several provisional usage fee values ​​\C were set in the range of \30,000 to \100,000, and a profit and loss simulation was performed. The total number of users was set to 10 for each \C, and the simulation dataset Df i (i=1~10) were randomly and repeatedly extracted from the database 10 times, and each time a profit and loss simulation was performed. For example, when \P=450,000, \L=\8,000,000, and β=1%, the average user profit / loss ratio is +16%, similar to method 1. Similarly, when β=20%, the average user profit / loss ratio is +15%. The simulation result is the average user profit / loss R ave , average user profit / loss ratio (R ave (Total usage fees), average profit / loss for operators ave , and the average profit / loss ratio of operators (\U ave (Total usage fee)

[0042] Referring to Figure 6, we will explain how to determine the setting range of \C. The horizontal axis represents the usage fee \C, and the vertical axis represents the user profit / loss ratio or the operator profit / loss ratio. Curve C1 represents the user profit / loss ratio, curve C2 represents the operator profit / loss ratio when β=20%, and curve C3 represents the operator profit / loss ratio when β=1%. The intersection point (90) of curves C1 and C2 is also called the equilibrium point. The range in which the range of \C where curve C1 is above 0% and \C is 58,100 or less overlaps with the range of curve C2 above 0% and \C above 42,000 is the setting range of \C when β=20%. Similarly, the setting range of \C when β=1% can also be determined. The algorithm output was \42,000 to \58,100 when selling and administrative expenses were 20%, and \34,000 to \58,100 when selling and administrative expenses were 1%. Although the range of usage fee \C where both the user profit / loss ratio and the operator profit / loss ratio are positive was explained as being derived from the diagram, it can also be determined by iterative calculation using a computer (e.g., Excel's Goal Seek function, generally known as Newton's method). A similar diagram can be considered for each of the 20 combinations of \P and \L. Then, the range in which usage fee \C is valid is calculated for each of the 20 combinations. For combinations in which no range in which usage fee is valid is indicated. For example, for B13-B14, B23-B24, B41, B44, B54, and B62 in Figure 5, it is indicated that no range in which usage fee is valid is valid.

[0043] When examining the usage fee at which equilibrium occurs, it was found that the larger _L is, the smaller the equilibrium value _C is, and the larger _P is, the larger the equilibrium value _C is. Therefore, the equilibrium value _C can be predicted using a multiple regression equation with _P and _L as explanatory variables. A computer-based multiple regression analysis showed that the correlation coefficient R² = 0.9977 is extremely close to 1, indicating that the equilibrium value _C can be predicted with high accuracy.

[0044] In Figure 5, for B13-B14, B23-B24, B44, and B54, the user's profit / loss is always negative. In these cases, it was found that the number of discarded molds was relatively small and there were no unexpected demands. On the other hand, in B41 and B62, where the operator's profit / loss is always negative, it was found that the number of discarded molds was relatively large and there were many unexpected demands. From the above, it can be concluded that when \P is set small, the profit from discarding molds is small, so the prediction by the judgment system (10) is biased towards not discarding molds, and the payment of the usage fee \C for the user outweighs the benefit, resulting in failure. When \P is set large, the prediction by the judgment system (10) is biased towards discarding molds, and the payment of the cost of dealing with unexpected demand for the operator outweighs the expected usage fee, resulting in failure. Therefore, it can be said that the usage fee \C is valid as long as the number of discarded molds and the number of unexpected demands are within a certain range. For example, it was found that if (number of unexpected demands) / (number of models discarded) is between 3.2% and 5.2%, then the usage fee ¥C is valid.

[0045] <How to determine \C, Part 3: Examples> As shown in Figure 5, 20 combinations of \P and \L are set. j and \P k For the set, similar to method 2, the total number of users will be set to 10, and the simulation dataset Df i (i=1~10) was randomly and repeatedly extracted from database D 10 times, and each time it was subjected to a profit and loss simulation. Calculations were performed for both 20% and 1% selling, general, and administrative expenses.

[0046] All j and \P k When the same \C is set for a set of \L j and \P k For all combinations, the average profit / loss for operators is \U ave The sum of these is the combined average profit / loss of the operators Σ\U ave Assuming the average user profit / loss is R ave The sum of the user combined profit and loss ΣR ave Let's assume that Σ\U ave The range of \C such that ≥ 0 and Σ\Rave ≥ 0 will be output.

[0047] For example, when 10 simulations were performed with P=850,000, L=12,000,000, and usage fee C=50,000, the average user profit / loss ratio was +198%, and the average operator profit / loss ratio was -41%. The calculation method for the average user profit / loss ratio and the average operator profit / loss ratio is the same as in method 2. When the average user profit / loss ratio and the average operator profit / loss ratio were calculated for all 20 cases, there were patterns where the average user profit / loss was positive and the average operator profit / loss was negative, and the opposite pattern. Also, there was only one pattern where both the average user profit / loss and the average operator profit / loss were positive. In all 20 cases, when the operation was conducted with a usage fee C=50,000, the overall average operator profit / loss ratio was -24%, indicating that the operation would not be viable. On the other hand, the average user profit / loss ratio was +81%. From this, it was found that C needs to be set a little higher.

[0048] Profit and loss simulations were performed for four usage fee C patterns: ¥50,000, ¥70,000, ¥85,000, and ¥100,000. The range within which the usage fee C can be established can be determined by using the average user profit / loss ratio in Figure 6 as the combined average user profit / loss ratio and the average operator profit / loss ratio as the combined average operator profit / loss ratio. The algorithm output was ¥73,000 to ¥93,500 when selling and administrative expenses were 20%, and ¥57,500 to ¥93,500 when selling and administrative expenses were 1%. Similar to method 2, the usage fee C can be calculated by iterative calculation using a computer.

[0049] <How to determine \C, Part 4: Examples> Figure 5 shows 20 different ways of \P k and \L j This is set. \P k and \L j For the set, similar to method 2, the total number of users will be set to 10, and the simulation dataset Df i (I-10) were randomly and repeatedly extracted from database D 10 times, and each time a profit and loss simulation was performed. Selling, general and administrative expenses were set to 20% or 1%.

[0050] Mimicking the concept of property insurance, the amount of compensation for losses is ¥L j Usage fee proportional to C j The following settings were applied. For example, the usage fee for B11-B14 was set to ¥120,000, the usage fee for B21-B24 was set to ¥100,000, the usage fee for B32-B33 was set to ¥90,000, the usage fee for B41-B44 was set to ¥80,000, the usage fee for B52-B54 was set to ¥70,000, and the usage fee for B62-B64 was set to ¥50,000. The average operator profit / loss for all patterns was set to ¥U ave The sum of these is the combined average profit / loss of the operators Σ\U ave Assuming the average user profit / loss is R ave The sum of the user average profit / loss is Σ\R ave Let's assume that Σ\U ave ≥0 and Σ\R ave A usage fee array C such that ≥ 0 jk The range in which the condition is met will be output.

[0051] For example, in 10 simulations with P=850,000, L=12,000,000, C=120,000, and selling and administrative expenses of 20%, the average profit / loss ratio for users was +24%, and the average profit / loss ratio for operators was +21%. For 20 different patterns, the combined average profit / loss ratio for users was +4%, and the combined average profit / loss ratio for operators was +19%, both of which are greater than or equal to zero, indicating that the above settings can be adopted.

[0052] We have explained the case where the usage fee for B11 is ¥120,000, but the usage fees for B11 to B14 are ¥C 1k Ten different amounts were set within the range of 140,000 to 80,000. Then, C was set to be proportional to L. 2k ~C 4k This was also set. The combined average profit / loss ratio for users and the combined average profit / loss ratio for operators were calculated for selling, general and administrative expenses of 1% and 20%, respectively. Then, as in Figure 6, C was set so that the combined average profit / loss ratio for users and the combined average profit / loss ratio for operators would be above 0%. 1kThe range in which the condition is met was calculated. The algorithm's output was ¥96,000 to ¥125,000 when selling, general and administrative expenses were 20%, and ¥94,000 to ¥125,000 when selling, general and administrative expenses were 1%.

[0053] <How to determine \C, Part 5: Examples> As shown in Figure 5 \P k and \L j 20 possible combinations were set. j and \P k Similar to method 2 for determining the set, the total number of users is set to 10, and the simulation dataset Df i (i=1~10) were randomly and repeatedly extracted from database D 10 times, and each time a profit and loss simulation was performed. Selling, general and administrative expenses were set to 20% or 1%. And, as with method 2, the usage fee was \C jk This is set. For example, User benefit amount \P k For \L j Regardless, the usage fee is the same. k The following is set. Then, the operator's average profit / loss for all combinations is... ave The sum of these is the combined average profit / loss of the operators Σ\U ave Assuming the average user profit / loss is R ave The sum of the user average profit / loss Σ\ ave As, Σ\U ave ≥0 and Σ\R ave A usage fee array C such that ≥ 0 j The range in which the condition is met will be output. [Explanation of Symbols]

[0054] (3) Operator Profit and Loss Calculation Department, (4) User Profit and Loss Calculation Department, (6) Replacement Parts Inventory Trend Database, (8) Profit and Loss Simulation, (10) Judgment System, (13) Per-User Profit and Loss Calculation Department

Claims

1. A fee calculation system for determining whether or not equipment or molds can be disposed of, which calculates the amount that a user pays to the operator as a fee for using the system, The aforementioned equipment or molds are used to manufacture replacement parts for automobiles. The aforementioned determination system is A means for calculating the probability of unexpected demand occurring from the part number assigned to the replacement part, using a formula included in a trained model, A means for comparing the aforementioned probability with a threshold probability value included in the trained model, Includes, If the probability is less than the threshold probability value, information is output indicating that the equipment or mold can be discarded as it is predicted that the unexpected demand will not occur; if the probability is equal to or greater than the threshold probability value, information is output indicating that the equipment or mold should be maintained as it is predicted that the unexpected demand will occur. The aforementioned fee calculation system is: A means for estimating the benefits the user will receive if the prediction result by the judgment system is correct, A database showing the inventory trends of each of multiple replacement parts, and Includes, The range of usage fees that ensures the first profit / loss for the operator is not negative, and the second profit / loss based on the profits obtained by the user is not negative, is calculated by simulation using the inventory trend database. A system for calculating usage fees.

2. A set of input values ​​is defined that includes the profit P obtained by the user and the unexpected demand response cost L that the operator pays to the user when the probability is smaller than the threshold probability value and the unexpected demand occurs. The scope of the applicable usage fee is calculated for the aforementioned input value set. The usage fee calculation system according to claim 1.

3. Multiple sets of input values ​​are defined, The scope of the aforementioned usage fee is calculated for each set of input values. The usage fee calculation system according to claim 2.

4. Multiple sets of input values ​​are defined, The range of usage fees is calculated such that the sum of the first profit and loss across multiple input value sets is not negative, and the sum of the second profit and loss is not negative. The usage fee calculation system according to claim 2.

5. Multiple sets of input values ​​are defined, For each set of input values, set a usage fee C that is proportional to the unexpected demand response cost L. The range of usage fees is calculated such that the sum of the first profit and loss across multiple input value sets is not negative, and the sum of the second profit and loss is not negative. The usage fee calculation system according to claim 2.

Images