Forecasting system, demand forecasting device, demand forecasting method and program

Abstract

A prediction system (100) that predicts demand using Bayes statistics comprises: a prior probability distribution generation device (1) that generates a prior probability distribution to be used in demand prediction; and a demand prediction device (2) that generates a demand prediction probability distribution, which is a subsequent probability distribution, on the basis of the prior probability distribution. The prior probability distribution device (1) includes a prior probability distribution generation unit that generates a prior probability distribution using supply plan data indicating a supply plan for an article, and generates prior probability distribution data indicating the prior probability distribution. The demand prediction device (2) includes a demand prediction data generation unit that: generates a demand prediction probability distribution, which is a subsequent probability distribution, by using a demand model representing a demand trend for the article, actual demand data indicating the actual demand for the article, and the prior probability distribution data; and generates demand prediction data indicating a demand prediction probability distribution.

Description

[Technical Field]

[0001] The present disclosure relates to a forecasting system, a demand forecasting device, a demand forecasting method, and a program. [Background technology]

[0002] In business activities, it is important to control the supply and demand balance to avoid holding excess inventory while shortening demand lead times and responding to demand fluctuations. Demand forecasting is one of the methods used to achieve this. Conventional demand forecasting methods involve performing various calculations using commonly used estimation statistics (hereafter referred to as conventional statistics). Conventional statistics calculate statistical quantities (such as the mean and standard deviation) for a certain set of parameters as unique values ​​based on data extracted from a sample.

[0003] Patent Document 1 discloses a demand forecasting device that probabilistically forecasts future demand for each of multiple businesses that make up a group of businesses with a high correlation between demand and supply. The technology described in Patent Document 1 models performance trend fluctuations from past performance data using regression analysis based on conventional statistics. Furthermore, business fluctuations are calculated by subtracting the trend fluctuations from the total fluctuations, and are modeled using a stochastic differential equation. These models are used to calculate the demand forecast probability distribution. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-326346 Summary of the Invention [Problem to be solved by the invention]

[0005] The technology described in Patent Document 1 requires calculations based on data accumulated over a sufficiently long period of time, so it may not be applicable to business fields where changes are rapid. Also, with the technology described in Patent Document 1, if there is insufficient sample past performance data, the accuracy of demand forecasting decreases, making it difficult to put into practical use.

[0006] The present disclosure has been made to solve the above-mentioned problems, and aims to realize a demand forecast that can be applied to fast-changing business fields and is practically usable. [Means for solving the problem]

[0007] In order to achieve the above object, a prediction system according to the present disclosure is a prediction system that performs demand prediction using Bayesian statistics. The prediction system includes a prior distribution generation device and a demand prediction device. The prior distribution generation device generates a prior distribution to be used for demand prediction. The demand prediction device generates a demand forecast probability distribution, which is a posterior distribution, based on the prior distribution. The prior distribution generation device has a prior distribution generation unit. The prior distribution generation unit generates a prior distribution using supply plan data that indicates a supply plan for an item, and generates prior distribution data that indicates the prior distribution. The demand prediction device has a demand forecast data generation unit. The demand forecast data generation unit generates a demand forecast probability distribution, which is a posterior distribution, using a demand model that indicates the demand trend of the item, demand record data that indicates the actual demand for the item, and prior distribution data, and generates demand forecast data that indicates the demand forecast probability distribution. [Effects of the Invention]

[0008] According to the present disclosure, by using Bayesian statistics as a method for forecasting demand for goods, statistical quantities can be estimated probabilistically even if sample data is insufficient, making it possible to realize demand forecasting that is applicable to rapidly changing business fields and is easy to put into practical use. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing a configuration example of a prediction system according to a first embodiment. [Figure 2] FIG. 1 is a diagram illustrating an example of a functional configuration of a prior distribution generating device according to a first embodiment. [Figure 3] FIG. 1 is a diagram showing an example of a method for generating a prior distribution according to the first embodiment; [Figure 4] FIG. 1 is a diagram illustrating an example of a functional configuration of a demand prediction device according to a first embodiment. [Figure 5] FIG. 10 is a diagram showing an example of a method for generating a demand forecast probability distribution according to the first embodiment. [Figure 6] FIG. 1 is a diagram showing an example of demand forecast data according to the first embodiment. [Figure 7] A diagram showing an example of generating a demand forecast probability distribution using the previously generated demand forecast probability distribution as a prior distribution. [Figure 8] 1 is a flowchart showing an example of the operation of a demand forecasting process according to the first embodiment. [Figure 9] FIG. 10 is a diagram showing a configuration example of a prediction system according to a second embodiment. [Figure 10] FIG. 10 is a diagram showing an example of a functional configuration of a KPI prediction device according to a second embodiment; [Figure 11] FIG. 10 is a diagram showing an example of a method for calculating a KPI value according to the second embodiment. [Figure 12] FIG. 10 is a diagram showing an example of KPI data according to the second embodiment. [Figure 13] FIG. 10 is a diagram showing an example of an alert condition according to the second embodiment. [Figure 14] 10 is a flowchart showing an example of the operation of a KPI prediction process according to the second embodiment. [Figure 15] FIG. 10 is a diagram showing a configuration example of a prediction system according to a third embodiment. [Figure 16] FIG. 10 is a diagram showing an example of a functional configuration of a management index prediction device according to a third embodiment. [Figure 17] FIG. 10 is a diagram showing an example of a method for calculating a management index value according to the third embodiment. [Figure 18] 10 is a flowchart showing an example of the operation of a management index prediction process according to the third embodiment. [Figure 19] FIG. 10 is a diagram showing an example of a sampling method of a prior distribution according to a modified example. [Figure 20] FIG. 10 is a diagram showing an example of a display screen of a sampling result of a prior distribution according to a modified example. [Figure 21] FIG. 1 is a diagram showing an example of the hardware configuration of a prior distribution generation device, a demand forecasting device, a KPI forecasting device, and a management index forecasting device according to first to third embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0010] A forecasting system, a demand forecasting device, a demand forecasting method, and a program according to the present embodiment will be described in detail below with reference to the drawings. Note that identical or corresponding parts in the drawings are given the same reference numerals. In this embodiment, an example of forecasting product demand will be described. In this embodiment, Bayesian statistics is used as the demand forecasting method. In this embodiment, demand forecasting is performed on a daily basis.

[0011] (Embodiment 1) The configuration of a prediction system 100 according to the first embodiment will be described with reference to Fig. 1. The prediction system 100 includes a prior distribution generation device 1 that generates a prior distribution to be used for demand prediction, and a demand prediction device 2 that generates a demand prediction probability distribution, which is a posterior distribution, based on the prior distribution.

[0012] The functional configuration of the prior distribution generation device 1 will be described with reference to Fig. 2. The prior distribution generation device 1 includes a shipment plan data storage unit 11 that stores shipment plan data indicating a product shipment plan, an error data storage unit 12 that stores error data indicating an error between the product shipment plan and actual shipment results, a statistical model storage unit 13 that stores a statistical model that indicates the relationship between the shipment plan data and the error data, a prior distribution generation unit 14 that generates a prior distribution using the shipment plan data, the error data, and the statistical model, and a prior distribution output unit 15 that outputs prior distribution data indicating the prior distribution.

[0013] It is advisable to use product shipping plans drawn up by experts as shipping plan data. This makes it possible to utilize the know-how of experts that has not yet been digitized in demand forecasting, which is expected to improve the accuracy of demand forecasting. Shipping plans and shipping results are examples of supply plans and supply results, respectively. Shipping plan data is an example of supply plan data. Shipping plan data and error data are generated by product model name, which is SKU (Stock Keeping Unit).

[0014] A method for generating a prior distribution executed by the prior distribution generation unit 14 will be described with reference to FIG. 3. In the example of FIG. 3, the daily planned quantities of the shipment plan for products with model names a, b, and c are 18 units, 105 units, and 50 units, respectively. It is predetermined that the statistical model representing the relationship between the shipment plan data and error data for products with model names a and b is statistical model SM1, and the statistical model representing the relationship between the shipment plan data and error data for product with model name c is statistical model SM2. Statistical model SM1 is a normal distribution, with parameters being the mean and standard deviation. Statistical model SM2 is a Poisson distribution, with a parameter being the mean.

[0015] The prior distribution generation unit 14 generates, as a prior distribution, a normal distribution with a mean of 18 (units) and a standard deviation of 20 for products with model name a, based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 generates, as a prior distribution, a normal distribution with a mean of 105 (units) and a standard deviation of 50 for products with model name b, based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 generates, as a prior distribution, a Poisson distribution with a mean of 50 (units) for products with model name c, based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 generates prior distribution data indicating the prior distribution.

[0016] Returning to FIG. 2, the prior distribution output unit 15 outputs the prior distribution data generated by the prior distribution generation unit 14 to the demand prediction device 2.

[0017] Next, the functional configuration of the demand prediction device 2 will be described with reference to Fig. 4. The demand prediction device 2 includes a prior distribution acquisition unit 21 that acquires prior distribution data from the prior distribution generation device 1, a demand model storage unit 22 that stores a demand model that represents a demand trend of a product, a demand record data storage unit 23 that stores demand record data that indicates the demand record of the product, a demand forecast data generation unit 24 that uses the prior distribution data, the demand model, and the demand record data to generate a demand forecast probability distribution that is a posterior distribution, and generates demand forecast data that indicates the demand forecast probability distribution, and a demand forecast data output unit 25 that outputs the demand forecast data.

[0018] When the prior distribution acquisition unit 21 acquires the prior distribution data, the demand forecast data generation unit 24 calculates a likelihood based on the demand model and actual demand data up to the previous day. The demand forecast data generation unit 24 multiplies the likelihood by the prior distribution indicated by the prior distribution data, performs MCMC (Markov chain Monte Carlo methods) sampling, and generates a demand forecast probability distribution, which is a posterior distribution.

[0019] The method of generating a demand forecast probability distribution executed by the demand forecast data generation unit 24 will be described with reference to FIG. 5. In the example of FIG. 5, the prior distribution is a probability distribution of an average λ of the number of units shipped per day, with an error of ±20% centered around 18 units. The average number of units shipped from the first to fifth days indicated by the actual demand data is 13 units / day. The demand forecast data generation unit 24 calculates a likelihood based on this actual demand data and a demand model. The demand model is a Poisson distribution, and the likelihood is a probability distribution of an average λ centered around 13 units. The demand forecast data generation unit 24 multiplies the prior distribution and the likelihood to generate a demand forecast probability distribution, which is a posterior distribution. At this time, MCMC sampling is performed to facilitate numerical calculations. The demand forecast data generation unit 24 generates demand forecast data indicating the demand forecast probability distribution.

[0020] The demand forecast data will now be explained using Figure 6. The demand forecast data shown in Figure 6 shows the demand forecast probability distribution of products with model names a, b, and c. For each of the products with model names a, b, and c, the statistics λ(a), λ(b), and λ(c) were sampled 1,000 times using MCMC sampling. Each of the 1,000 samples was assigned a sampling number (denoted as NO. in the figure) ranging from 1 to 1,000. For each sampling number, the values ​​of the demand forecast probability distribution Ypro_(a,1) to Ypro_(a,End) for products with model name a, the demand forecast probability distribution Ypro_(b,1) to Ypro_(b,End) for products with model name b, and the demand forecast probability distribution Ypro_(b,1) to Ypro_(b,End) for products with model name c were calculated.

[0021] Returning to Fig. 4, the demand forecast data output unit 25 outputs the demand forecast data generated by the demand forecast data generation unit 24. As shown in Fig. 5, the demand forecast data is output as a graph that represents, for example, the demand forecast probability distribution of sampling Nos. 1 to 1000 for each product model name as a distribution group. This allows the user to visually grasp the demand forecast in the form of a probability distribution.

[0022] The demand forecast data generation unit 24 performs the process of generating a demand forecast probability distribution, for example, once a day. If the cycle of the process of generating a demand forecast probability distribution is short, the process of generating a demand forecast probability distribution may be performed using the previously generated demand forecast probability distribution (posterior distribution) as the prior distribution. FIG. 7 shows an example of generating a demand forecast probability distribution using the demand forecast probability distribution (posterior distribution) generated in the example of FIG. 5 as the prior distribution. The number of units shipped on the sixth day indicated by the demand actual data is 10 units / day. The demand forecast data generation unit 24 calculates a likelihood using a demand model (Poisson distribution) based on this demand actual data. The likelihood is a probability distribution with a mean λ centered on 10 units. The demand forecast data generation unit 24 multiplies the prior distribution and the likelihood, performs MCMC sampling, and generates a demand forecast probability distribution, which is a posterior distribution.

[0023] By generating a demand forecast probability distribution using the previously generated demand forecast probability distribution (posterior distribution) as a prior distribution, it is possible to expect improved processing speed. In addition, the demand forecast probability distribution (posterior distribution) can be updated based on newly acquired actual demand data. As shown in Figures 5 and 7, by repeating the process of generating the demand forecast probability distribution, the probability distribution of the statistic λ of the demand forecast probability distribution becomes sharper, and the variance in the graph showing the demand forecast probability distributions of sampling numbers 1 to 1000 as a distribution group becomes smaller.

[0024] Next, the flow of the demand forecasting process executed by the forecasting system 100 will be described with reference to Fig. 8. The demand forecasting process shown in Fig. 8 starts, for example, when a demand forecasting instruction is input to the prior distribution generation device 1. When the demand forecasting instruction is input, the prior distribution generation unit 14 of the prior distribution generation device 1 generates a prior distribution for each product model name based on the shipment plan data stored in the shipment plan data storage unit 11, the error data stored in the error data storage unit 12, and the statistical model stored in the statistical model storage unit 13 (step S11).

[0025] In the example of Figure 3, the prior distribution generation unit 14 generates, as prior distributions, a normal distribution with a mean of 18 (units) and a standard deviation of 20 for the product with model name a, a normal distribution with a mean of 105 (units) and a standard deviation of 50 for the shipping plan data for the product with model name b, and a Poisson distribution with a mean of 50 (units) for the product with model name c.

[0026] Returning to FIG. 8, the prior distribution generation unit 14 generates prior distribution data indicating the prior distribution (step S12), and the prior distribution output unit 15 outputs the prior distribution data generated by the prior distribution generation unit 14 to the demand forecasting device 2 (step S13).

[0027] When the prior distribution acquisition unit 21 of the demand forecasting device 2 acquires prior distribution data (step S14), the demand forecast data generation unit 24 calculates likelihood based on the demand model and actual demand data up to the previous day (step S15). The demand forecast data generation unit 24 multiplies the likelihood by the prior distribution indicated by the prior distribution data, performs MCMC sampling, and generates a demand forecast probability distribution, which is a posterior distribution (step S16).

[0028] In the example of FIG. 5, the prior distribution is a probability distribution of the average number of units shipped per day λ, with a center of 18 units and an error of ±20%. The actual demand data indicates that the average number of units shipped from the first to fifth days is 13 units / day. The demand forecast data generation unit 24 calculates the likelihood using a demand model based on this actual demand data. The demand model is a Poisson distribution, and the likelihood is a probability distribution of the average λ centered on 13 units. The demand forecast data generation unit 24 multiplies the prior distribution and the likelihood and performs MCMC sampling to generate a demand forecast probability distribution, which is a posterior distribution.

[0029] 8, the demand forecast data generation unit 24 generates demand forecast data indicating a demand forecast probability distribution (step S17). The demand forecast data output unit 25 outputs the demand forecast data generated by the demand forecast data generation unit 24 (step S18), and the process ends.

[0030] In the example of Fig. 5, the demand forecast data is output as a graph that represents the demand forecast probability distribution of sampling numbers 1 to 1000 as a distribution group. This allows the user to visually grasp the demand forecast in the form of a probability distribution.

[0031] The process of generating a demand forecast probability distribution in steps S14 to 18 is performed, for example, once a day. If the cycle of the process of generating a demand forecast probability distribution is short, in step S14, the demand forecast probability distribution generated in the previous step S16 may be acquired as a prior distribution, and steps S14 to 18 may be performed.

[0032] According to the forecasting system 100 of embodiment 1, by using Bayesian statistics as a method for forecasting product demand, it is possible to estimate statistical quantities probabilistically even if there is insufficient sample data, making it applicable to business fields that undergo rapid change and enabling demand forecasting that is easy to put into practical use.

[0033] In Bayesian statistics, statistics are estimated using a probability distribution, assuming that parameters are subject to change. Bayesian statistics has the advantage that statistics are output as a probability distribution (posterior distribution), and that posterior distributions can be updated based on newly acquired data. Compared to traditional statistics, these statistics have the advantage of being able to estimate statistics probabilistically even when actual data is insufficient. These advantages make it applicable to business fields where change is rapid. Furthermore, when there is instability in the data due to insufficient data, missing data, or outliers, the posterior probability distribution outputs a forecast result with a wide tail, allowing users to visually grasp the instability of the data. By setting an interval such as a 5% to 95% interval, users can easily grasp the lower and upper limits of the demand forecast. Users can consider the reliability of the demand forecast when making their decisions.

[0034] Another example of a situation where demand forecasts can be utilized is production planning, but high accuracy in production planning is maintained through the undigitized demand forecasting know-how of experts, and forecast results based on mathematical demand forecasting logic have not always been satisfactory.By generating a prior distribution based on shipping plans by experts, the undigitized know-how of experts can be utilized in demand forecasting, and improved demand forecasting accuracy can be expected.

[0035] (Embodiment 2) The configuration of a prediction system 200 according to the second embodiment will be described with reference to Fig. 9. In addition to the prior distribution generation device 1 and demand prediction device 2 of the prediction system 100, the prediction system 200 according to the second embodiment includes a KPI prediction device 3 that converts statistics of a demand prediction probability distribution into KPI (Key Performance Indicator) values ​​and generates KPI data including a KPI probability distribution, which is a probability distribution of the KPI values. In the second embodiment, the demand prediction device 2 outputs demand prediction data to the KPI prediction device 3.

[0036] The functional configuration of the KPI prediction device 3 will be described with reference to Fig. 10. The KPI prediction device 3 includes a demand forecast data acquisition unit 31 that acquires demand forecast data from the demand forecast device 2, a confirmed production data storage unit 32 that stores confirmed production data indicating the quantity of products whose production has been confirmed but not included in the actual demand data, a product master storage unit 33 that stores a product master that is data related to products such as product lots, product unit prices, and production costs per lot, a previous inventory data storage unit 34 that stores previous inventory data indicating the actual inventory of products, a KPI data generation unit 35 that converts statistics of the demand forecast probability distribution into practical KPI values ​​and other numerical values ​​based on the confirmed production data, the product master, and the previous inventory data to generate a KPI probability distribution and generate KPI data including the KPI probability distribution, and a KPI data output unit 36 ​​that outputs the generated KPI data.

[0037] The KPI prediction device 3 also includes a KPI alert condition storage unit 37 that stores predetermined KPI alert conditions, which are conditions for issuing an alert for the KPI probability distribution included in the KPI data, and a KPI alert determination unit 38 that issues an alert when the KPI probability distribution included in the generated KPI data satisfies the KPI alert conditions.

[0038] Practical KPI values ​​are values ​​related to production (P), sales (S), and inventory (I). The KPI data generation unit 35 calculates KPI values ​​and other numerical values ​​from parameters of the demand forecast probability distribution (posterior distribution) for each sample number to generate a KPI probability distribution, and generates KPI data including the KPI probability distribution. The KPI data generation unit 35 defines an interval for the KPI probability distribution, for example, from the 5% point to the 95% point.

[0039] Here, a method for calculating KPI values ​​will be explained using Fig. 11. In the example of Fig. 11, the KPI data generation unit 35 executes a statistical distribution simulation on the demand forecast probability distribution (posterior distribution) at the current time (the fifth day) to predict the daily sales volume for the sixth to twentieth days. The statistical distribution simulation is, for example, random sampling. At this time, the KPI data generation unit 35 calculates the value of the 5th percentile and the value of the 95th percentile of the demand forecast probability distribution. Hereinafter, the value at the nth percentile in the probability distribution will be referred to as the nth percentile value.

[0040] The KPI data generation unit 35 calculates the cumulative number of units sold (graph G1 in the figure), which is the cumulative sum of the daily number of units sold. At this time, the KPI data generation unit 35 also calculates the cumulative number of units sold when the daily number of units sold for the 6th to 20th days is the 5th percentile of the demand forecast probability distribution (graph G2 in the figure), and the cumulative number of units sold when the daily number of units sold for the 6th to 20th days is the 95th percentile of the demand forecast probability distribution (graph G3 in the figure). Note that the cumulative number of units sold for the 6th to 20th days in graph G1 is not limited to the cumulative number of units sold obtained by adding up the daily number of units sold based on a statistical distribution simulation, and may be, for example, the cumulative number of units sold when the daily number of units sold is the 50th percentile of the demand forecast probability distribution.

[0041] The KPI data generation unit 35 calculates a KPI value (graph G4 in the figure), which is the cumulative sales amount, from the cumulative sales volume, and generates KPI data including a KPI probability distribution. Specifically, the KPI data generation unit 35 references the product master and calculates the cumulative sales amount, which is the cumulative sales volume multiplied by the unit price of the product, as the KPI value. At this time, the KPI data generation unit 35 also calculates a KPI value (graph G5 in the figure) when the daily sales volume for the sixth to twentieth days is the 5th percentile of the demand forecast probability distribution, and a KPI value (graph G6 in the figure) when the daily sales volume for the sixth to twentieth days is the 95th percentile of the demand forecast probability distribution. Note that the cumulative sales amount for the sixth to twentieth days in graph G4 is not limited to the cumulative sales amount calculated from the cumulative sales volume obtained by adding up the daily sales volume obtained by a statistical distribution simulation for each day, and may be, for example, the cumulative sales amount calculated from the cumulative sales volume when the daily sales volume is the 50th percentile of the demand forecast probability distribution.

[0042] The KPI data generating unit 35 calculates other values, such as the cost incurred over the entire period and the profit for the period, and generates KPI data for each sample number.

[0043] KPI data will now be described with reference to Fig. 12. The example of KPI data in Fig. 12 has the following items: "NO." which is the sample number; "Average" which indicates the average of the demand forecast probability distribution; "5th percentile value" which indicates the 5th percentile value of the demand forecast probability distribution; "95th percentile value" which indicates the 95th percentile value of the demand forecast probability distribution; "S6" which indicates the daily sales volume on the 6th day through "S20" which indicates the daily sales volume from the 20th day; "Ssum1" which indicates the cumulative sales volume on the 1st day through "Ssum20" which indicates the cumulative sales volume from the 20th day; "Samt1" which indicates the cumulative sales amount on the 1st day through "Samt20" which indicates the cumulative sales amount from the 20th day; "C" which indicates the cost incurred over the entire period; and "Pro" which indicates the profit for the period.

[0044] The KPI data generation unit 35 refers to the product master and calculates, for example, the cost required for the entire period by dividing the production cost per lot by the number of lots and multiplying the result by the cumulative number of units sold on the 20th day.The KPI data generation unit 35 calculates the profit for the period by subtracting the cost required for the entire period from the cumulative sales amount on the 20th day.

[0045] In the example of FIG. 12, the KPI value is the sales amount, and the other values ​​are costs and profits, but the present invention is not limited to this. The KPI value and other values ​​may be any values ​​related to production (P), sales (S), and inventory (I) that can be calculated based on parameters of the demand forecast probability distribution. For example, the inventory quantity of a product may be calculated as the KPI value based on production confirmation data, inventory performance data, and the number of units sold. Furthermore, if other values ​​are not necessary, the KPI data generation unit 35 may calculate only the KPI value.

[0046] Returning to Figure 10, the KPI prediction device 3 is equipped with a production confirmation data storage unit 32, a product master storage unit 33, and an inventory performance data storage unit 34, but is not limited to these, and the KPI prediction device 3 only needs to store PSI data regarding the production, shipment, and inventory of products required to generate KPI data.

[0047] The KPI data output unit 36 ​​outputs the demand forecast data generated by the KPI data generation unit 35. The KPI alert determination unit 38 outputs an alert when the KPI probability distribution included in the KPI data generated by the KPI data generation unit 35 satisfies the KPI alert condition stored in the KPI alert condition storage unit 37.

[0048] The KPI alert condition will be explained using Fig. 13. Fig. 13 is a graph for a product with model name a, with the KPI value on the vertical axis y and the day on the horizontal axis x. In the example of Fig. 13, the KPI data generation unit 35 calculates a KPI value based on the demand forecast probability distribution (posterior distribution) at the present time (Today) and a KPI value based on the prior distribution.

[0049] Ppro is the KPI probability distribution at the end of the period calculated based on the demand forecast probability distribution. Ppri is the KPI probability distribution at the end of the period calculated based on the prior distribution. Graphs Ypro_5%(x), Ypro_50%(x), and Ypro_95%(x) are graphs obtained when the KPI value is calculated using the 5th, 50th, and 95th percentile values ​​of the demand forecast probability distribution, respectively. The slope of graph Ypro_50%(x) is the average λ of the demand forecast probability distribution. Graphs Ypri_5%(x), Ypri_50%(x), and Ypri_95%(x) are graphs obtained when the KPI value is calculated using the 5th, 50th, and 95th percentile values ​​of the prior distribution, respectively. The slope of graph Ypri_50%(x) is the average of the prior distribution.

[0050] In the example of FIG. 13 , the KPI alert condition is whether Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End). The KPI alert determination unit 38 determines whether Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End) based on the KPI data, and outputs an alert when Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End) is no longer true. This allows the user to know that the KPI value predicted based on the prior distribution is different from the KPI value predicted based on the demand forecast probability distribution, that is, that the KPI value is deviating from the shipping plan (e.g., a shipping plan made by an expert) on which the prior distribution is based. Note that the KPI alert condition is not limited to this. For example, the KPI alert determination unit 38 may output an alert when Ypro_50%(x=End) is below a threshold.

[0051] The KPI alert determination unit 38 outputs an alert to a spreadsheet application, a BI (Business Intelligence) tool, a new user interface system, or the like on the terminal used by the user. The alert may be output, for example, by screen display or audio output. Furthermore, if the alert is output together with a graph of KPI values ​​as shown in FIG. 13, the user can visually determine the alert conditions.

[0052] Next, the flow of the KPI prediction process executed by the prediction system 200 will be described with reference to Fig. 14. Steps S21 to S27 of the KPI prediction process shown in Fig. 14 are similar to steps S11 to S17 of the demand prediction process shown in Fig. 8, and therefore description thereof will be omitted.

[0053] The demand forecast data output unit 25 of the demand forecasting device 2 outputs the demand forecast data generated by the demand forecast data generation unit 24 to the KPI prediction device 3 (step S28). When the demand forecast data acquisition unit 31 of the KPI prediction device 3 acquires the demand forecast data (step S29), the KPI data generation unit 35 calculates the KPI value and other numerical values ​​from the parameters of the demand forecast probability distribution (posterior distribution) for each sample number (step S30), and generates KPI data including the KPI probability distribution (step S31).

[0054] In the example of Figure 12, the KPI data has the following items: "NO." which is the sample number; "Average" which indicates the average of the demand forecast probability distribution; "5th percentile value" which indicates the 5th percentile value of the demand forecast probability distribution; "95th percentile value" which indicates the 95th percentile value of the demand forecast probability distribution; "S6" which indicates the daily sales volume on the 6th day through "S20" which indicates the daily sales volume from the 20th day; "Ssum1" which indicates the cumulative sales volume on the 1st day through "Ssum20" which indicates the cumulative sales volume from the 20th day; "Samt1" which indicates the cumulative sales amount on the 1st day through "Samt20" which indicates the cumulative sales amount from the 20th day; "C" which indicates the cost incurred over the entire period; and "Pro" which indicates the profit for the period.

[0055] The KPI data generation unit 35 refers to the product master and calculates, for example, the cost required for the entire period by dividing the production cost per lot by the number of lots and multiplying the result by the cumulative number of units sold on the 20th day.The KPI data generation unit 35 calculates the profit for the period by subtracting the cost required for the entire period from the cumulative sales amount on the 20th day.

[0056] 14, the KPI data output unit 36 ​​outputs the KPI data generated by the KPI data generation unit 35 (step S32). The KPI alert determination unit 38 determines whether the KPI probability distribution included in the KPI data generated by the KPI data generation unit 35 satisfies the KPI alert condition stored in the KPI alert condition storage unit 37 (step S33). If the KPI alert condition is satisfied (step S33; YES), an alert is output (step S34), and the processing ends. If the KPI alert condition is not satisfied (step S33; NO), the processing ends.

[0057] 13, the KPI alert condition is whether or not Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End). The KPI alert determination unit 38 determines whether or not Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End) based on the KPI data, and outputs an alert when Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End) is no longer true.

[0058] According to the forecasting system 200 of the second embodiment, by using Bayesian statistics as a method for forecasting product demand, statistics can be estimated probabilistically even when sample data is insufficient, making it possible to realize demand forecasting that is applicable to business fields that undergo rapid change and that is easy to put into practical use. Furthermore, the forecasting system 200 of the second embodiment automatically converts the statistics of the demand forecast probability distribution into KPI values ​​and other values, thereby preventing the inclusion of worker intentions due to artificial calculations, the complicating of work, and a decline in the accuracy of KPI forecasts due to the worker's level of proficiency.

[0059] (Embodiment 3) The configuration of a prediction system 300 according to the third embodiment will be described with reference to Fig. 15. The prediction system 300 according to the third embodiment includes a prior distribution generation device 1, a demand prediction device 2, a KPI prediction device 3, and in addition, a management index prediction device 4 that converts KPI values ​​of KPI data into management indexes and generates a management index probability distribution.

[0060] The functional configuration of the management index prediction device 4 will be described with reference to FIG. 16. The management index prediction device 4 includes a KPI data acquisition unit 41 that acquires KPI data from the KPI prediction device 3, an accounting conversion master storage unit 42 that stores accounting conversion master data indicating the correspondence between KPI values, which are practical numerical values, and accounting numerical values, a unit conversion master storage unit 43 that stores unit conversion master data indicating the correspondence between practical units, such as product model names and model names, and business management units, such as product group names and business division names, a management index data generation unit 44 that converts KPI values ​​and other values ​​included in the KPI data into accounting numerical values ​​and then into business management units based on the accounting conversion master data and the unit conversion master data to generate management index data, and a management index data output unit 45 that outputs the generated management index data. Hereinafter, accounting numerical values ​​will be referred to as accounting values, and values ​​obtained by converting KPI values ​​and other values ​​into accounting values ​​and then into business management units will be referred to as management index values.

[0061] The management indicator prediction device 4 also includes a management indicator alert condition storage unit 46 that stores management indicator alert conditions, which are predetermined conditions for issuing an alert for the management indicator probability distribution included in the management indicator data, and a management indicator alert determination unit 47 that issues an alert when the management indicator probability distribution included in the generated management indicator data satisfies the management indicator alert conditions.

[0062] Management indicator values ​​include, for example, sales, manufacturing costs, and shelf turnover rates for each product group or business division. Based on the accounting conversion master and the unit conversion master, the management indicator data generation unit 44 converts the KPI values ​​and other values ​​included in the KPI data for each sample number into management indicator values ​​to generate a management indicator probability distribution, and generates management indicator data including the management indicator probability distribution. The management indicator data generation unit 44 defines an interval for the management indicator probability distribution, for example, from the 5% point to the 95% point.

[0063] Here, a method for calculating management indicator values ​​will be explained using FIG. 17. In the example of FIG. 17, the same models with model names a and b are grouped together into model A, and the same models with model names c and d are grouped together into model B, and the KPI values ​​are converted into accounting values. The accounting values ​​are, for example, numerical values ​​for account items. Furthermore, models A and B, which are in the same business division, are grouped together into business division α and converted into management indicator values. The management indicator values ​​are, for example, sales for each business division. Ppro, the management indicator value for business division α, is the management indicator probability distribution at the end of the period (End in the figure) calculated based on the demand forecast probability distribution. Ppri is the management indicator probability distribution at the end of the period (End in the figure) calculated based on the prior distribution.

[0064] Returning to FIG. 16 , the management indicator data output unit 45 outputs the management indicator data generated by the management indicator data generation unit 44. The management indicator alert determination unit 47 outputs an alert when the management indicator probability distribution included in the management indicator data generated by the management indicator data generation unit 44 satisfies the management indicator alert condition stored in the management indicator alert condition storage unit 46. The management indicator alert condition may be determined based on the management indicator data, for example, as with the KPI alert condition, to determine whether Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End), and output an alert when Ypro_5%(x=End)≦Ypri_50%(x=End)≦Ypro_95%(x=End) is no longer true. This allows the user to know that the management indicator value predicted based on the prior distribution and the management indicator value predicted based on the demand forecast probability distribution are deviating, i.e., that the management indicator value is deviating from the shipping plan (e.g., a shipping plan made by an expert) on which the prior distribution is based. The management index alert condition is not limited to this, and the management index alert determination unit 47 may output an alert when Ypro_50% (x=End) falls below a threshold value.

[0065] The management index alert determination unit 47 outputs an alert to a spreadsheet application, a BI tool, a new user interface system, or the like on the terminal used by the user. The alert may be output, for example, by displaying on a screen or by audio output. Furthermore, if the alert is output together with a graph of management index values ​​as shown in FIG. 17, the user can visually determine the alert conditions.

[0066] Next, the flow of the management index prediction process executed by the prediction system 300 will be described with reference to Fig. 18. Steps S41 to S51 of the management index prediction process shown in Fig. 18 are similar to steps S21 to S31 of the KPI prediction process shown in Fig. 14, and therefore description thereof will be omitted.

[0067] The KPI data output unit 36 ​​of the KPI prediction device 3 outputs the KPI data generated by the KPI data generation unit 35 to the management indicator prediction device 4 (step S52). When the KPI data acquisition unit 41 of the management indicator prediction device 4 acquires the KPI data (step S53), the management indicator data generation unit 44 converts the KPI value and other values ​​included in the KPI data for each sample number into management indicator values ​​based on the accounting conversion master and the unit conversion master (step S54), and generates management indicator data including a management indicator probability distribution (step S55).

[0068] In the example of FIG. 17, the same models with model names a and b are grouped together into model A, and the same models with model names c and d are grouped together into model B, and the KPI values ​​are converted into accounting values. The accounting values ​​are, for example, numerical values ​​for account items. Furthermore, models A and B, which are in the same business division, are grouped together into business division α and converted into management index values. The management index values ​​are, for example, sales for each business division.

[0069] 18, the management indicator data output unit 45 outputs the management indicator data generated by the management indicator data generation unit 44 (step S56). The management indicator alert determination unit 47 determines whether the management indicator probability distribution included in the management indicator data generated by the management indicator data generation unit 44 satisfies the management indicator alert condition stored in the management indicator alert condition storage unit 46 (step S57). If the management indicator alert condition is satisfied (step S57; YES), an alert is output (step S58), and the processing ends. If the management indicator alert condition is not satisfied (step S57; NO), the processing ends.

[0070] According to the forecasting system 300 of the third embodiment, by using Bayesian statistics as a method for forecasting product demand, statistics can be estimated probabilistically even when sample data is insufficient. This makes it possible to realize demand forecasting that is applicable to rapidly changing business fields and easy to implement in practice. Furthermore, the forecasting system 300 of the third embodiment automatically converts KPI values ​​and other values ​​into management index values, thereby preventing the influence of workers' will due to artificial calculations, the complication of work, and the reduction in the accuracy of management index value forecasts due to the worker's level of proficiency. Furthermore, automatic forecasting of management index values ​​allows for rapid strategy changes in response to changes in the external environment.

[0071] In the above embodiment, the prior distribution generation unit 14 of the prior distribution generation device 1 generates a prior distribution using shipment plan data, error data, and a statistical model, but this is not limited to this. The prior distribution generation device 1 may also generate a prior distribution based on shipment plan data that indicates the shipment plan that has been formulated including the error distribution, without using the error from past performance.

[0072] In the above embodiment, the prior distribution output unit 15 of the prior distribution generating device 1 outputs the prior distribution data generated by the prior distribution generating unit 14 to the demand forecasting device 2, but this is not limiting. For example, the prior distribution output unit 15 may display the results of sampling the prior distribution generated by the prior distribution generating unit 14 on a screen.

[0073] This modification will be described with reference to FIGS. 19 and 20. As shown in FIG. 19, the prior distribution generation unit 14 generates a normal distribution with a mean of 18 (units) and a standard deviation of 20 as a prior distribution for a product with type name a based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 generates a normal distribution with a mean of 105 (units) and a standard deviation of 50 as a prior distribution for a product with type name b based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 generates a Poisson distribution with a mean of 50 (units) as a prior distribution for a product with type name c based on the shipment plan data, error data, and a statistical model. The prior distribution generation unit 14 performs MCMC sampling on these prior distributions. In the example of FIG. 19, the products with type names a, b, and c are each sampled 1,000 times by MCMC sampling. Each of the 1000 samples is assigned a sampling number (denoted as "NO." in the figure) ranging from 1 to 1000. A parameter value (average) is calculated for each sampling number. The prior distribution generating unit 14 may organize these sampling results by model name into units larger than the model name, such as by model type. In the example of FIG. 19, the models of products with model names a and b are model A, and the model of product with model name c is model B.

[0074] The prior distribution output unit 15 displays on a screen the results of sampling the prior distribution generated by the prior distribution generation unit 14. FIG. 20 is an example of a display screen of the sampling results. The sampling results of the prior distribution for model names a, b, and c, and the sampling results for models A and B, which are a summary of the sampling results by model name, are displayed. This allows the planner to visually grasp the prior information distribution based on the shipping plan created by the planner on an SKU basis or in larger units, thereby improving the accuracy of the shipping plan.

[0075] In the above embodiment, demand forecasting is performed on a daily basis, but the unit of demand forecasting is not limited to this, and it may be performed on a half-day, two-day, one-week, one-month or other unit.

[0076] In the above embodiment, the shipping plan data and the error data are generated for each product model name, which is a SKU (Stock Keeping Unit), but this is not limitative and the data may be generated for a unit larger than the SKU, for example, for each product model name.

[0077] In the above embodiment, the KPI data includes the KPI value and other numerical values, but is not limited to this, and the KPI data may include at least the KPI value.

[0078] In the above embodiment, the KPI prediction device 3 includes the KPI alert condition storage unit 37 and the KPI alert determination unit 38, but if it is not necessary to output an alert, the KPI prediction device 3 does not have to include the KPI alert condition storage unit 37 and the KPI alert determination unit 38. Similarly, the management indicator prediction device 4 includes the management indicator alert condition storage unit 46 and the management indicator alert determination unit 47, but if it is not necessary to output an alert, the management indicator prediction device 4 does not have to include the management indicator alert condition storage unit 46 and the management indicator alert determination unit 47.

[0079] In the above embodiment, the demand for a product is predicted, but the demand is not limited to this, and the item for which the demand is predicted may be any item that is supplied to the market.

[0080] The hardware configurations of the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4 will be described with reference to Fig. 21. As shown in Fig. 21, the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4 each include a temporary storage unit 101, a storage unit 102, a calculation unit 103, an input unit 104, a transmission / reception unit 105, and a display unit 106. The temporary storage unit 101, the storage unit 102, the input unit 104, the transmission / reception unit 105, and the display unit 106 are all connected to the calculation unit 103 via a BUS.

[0081] The calculation unit 103 is, for example, a CPU (Central Processing Unit). The calculation unit 103 executes the processes of the prior distribution generation unit 14 of the prior distribution generation device 1, the demand forecast data generation unit 24 of the demand forecasting device 2, the KPI data generation unit 35 and KPI alert determination unit 38 of the KPI prediction device 3, and the management index data generation unit 44 and management index alert determination unit 47 of the management index prediction device 4 in accordance with a control program stored in the storage unit 102.

[0082] The temporary storage unit 101 is, for example, a RAM (Random-Access Memory). The temporary storage unit 101 loads the control program stored in the storage unit 102 and is used as a work area for the calculation unit 103.

[0083] The storage unit 102 is a non-volatile memory such as a flash memory, a hard disk, a DVD-RAM (Digital Versatile Disc - Random Access Memory), a DVD-RW (Digital Versatile Disc - Rewritable), etc. The storage unit 102 pre-stores programs for causing the calculation unit 103 to perform the processing of the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4, and also supplies data stored by these programs to the calculation unit 103 in accordance with instructions from the calculation unit 103, and stores the data supplied from the calculation unit 103. The shipping plan data storage unit 11, error data storage unit 12, and statistical model storage unit 13 of the prior distribution generation device 1, the demand model storage unit 22 and actual demand data storage unit 23 of the demand forecasting device 2, the production confirmation data storage unit 32, product master storage unit 33, actual inventory data storage unit 34, and KPI alert condition storage unit 37 of the KPI prediction device 3, and the accounting conversion master storage unit 42, unit conversion master storage unit 43, and management index alert condition storage unit 46 of the management index prediction device 4 are configured in the storage unit 102.

[0084] The shipment plan data storage unit 11, error data storage unit 12, and statistical model storage unit 13 of the prior distribution generation device 1, the demand model storage unit 22 and actual demand data storage unit 23 of the demand forecasting device 2, the production confirmation data storage unit 32, product master storage unit 33, actual inventory data storage unit 34, and KPI alert condition storage unit 37 of the KPI prediction device 3, and the accounting conversion master storage unit 42, unit conversion master storage unit 43, and management index alert condition storage unit 46 of the management index prediction device 4 may be provided by an external device or system.

[0085] The input unit 104 is an interface device that connects input devices such as a keyboard, a pointing device, and a voice input device to the BUS. Information input by the user is supplied to the calculation unit 103 via the input unit 104.

[0086] The transmitter / receiver 105 is a network termination device or a wireless communication device connected to the network, and a serial interface or a LAN (Local Area Network) interface connected thereto. In a configuration in which the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4 transmit and receive data, the transmitter / receiver 105 functions as the prior distribution output unit 15 of the prior distribution generation device 1, the prior distribution acquisition unit 21 and the demand forecast data output unit 25 of the demand forecasting device 2, the demand forecast data acquisition unit 31 and the KPI data output unit 36 ​​of the KPI prediction device 3, and the KPI data acquisition unit 41 and the management index data output unit 45 of the management index prediction device 4. In a configuration in which the KPI prediction device 3 and the management index prediction device 4 output alerts to a terminal used by a user, the transmitter / receiver 105 functions as the KPI alert determination unit 38 of the KPI prediction device 3 and the management index alert determination unit 47 of the management index prediction device 4.

[0087] The display unit 106 is a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electroluminescence) display. In a configuration in which the demand prediction device 2, the KPI prediction device 3, and the management index prediction device 4 respectively display demand prediction data, KPI data, and management index data on a screen, the display unit 106 functions as a demand prediction data output unit 25 of the demand prediction device 2, a KPI data output unit 36 ​​of the KPI prediction device 3, and a management index data output unit 45 of the management index prediction device 4. In addition, in a configuration in which the KPI prediction device 3 and the management index prediction device 4 output alerts by screen display, the display unit 106 functions as a KPI alert determination unit 38 of the KPI prediction device 3 and a management index alert determination unit 47 of the management index prediction device 4.

[0088] The shipment plan data storage unit 11, error data storage unit 12, statistical model storage unit 13, prior distribution generation unit 14, and prior distribution output unit 15 of the prior distribution generation device 1 shown in FIG. 2; the prior distribution acquisition unit 21, demand model storage unit 22, demand result data storage unit 23, demand forecast data generation unit 24, and demand forecast data output unit 25 of the demand forecasting device 2 shown in FIG. 4; the demand forecast data acquisition unit 31, production confirmation data storage unit 32, product master storage unit 33, inventory result data storage unit 34, KPI data generation unit 35, and KPI data output unit 36 ​​of the KPI prediction device 3 shown in FIG. 6. The processing of the KPI alert condition memory unit 37 and the KPI alert determination unit 38, as well as the KPI data acquisition unit 41, accounting conversion master memory unit 42, unit conversion master memory unit 43, management indicator data generation unit 44, management indicator data output unit 45, management indicator alert condition memory unit 46 and management indicator alert determination unit 47 of the management indicator prediction device 4 shown in Figure 16, is executed by the control program using the temporary memory unit 101, calculation unit 103, memory unit 102, input unit 104, transmission / reception unit 105, display unit 106, etc. as resources.

[0089] Furthermore, the above hardware configuration and flowchart are merely examples and can be changed and modified as desired.

[0090] The core components of the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4, such as the calculation unit 103, the temporary storage unit 101, the storage unit 102, the input unit 104, the transmission / reception unit 105, and the display unit 106, can be realized using an ordinary computer system rather than a dedicated system. For example, a computer program for executing the above operations may be stored and distributed on a computer-readable recording medium such as a flexible disk, a CD-ROM (Compact Disc - Read Only Memory), or a DVD-ROM (Digital Versatile Disc - Read Only Memory), and the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4 that execute the above processes may be configured by installing the computer program on a computer. Alternatively, the prior distribution generation device 1, the demand forecasting device 2, the KPI prediction device 3, and the management index prediction device 4 may be configured by storing the computer program in a storage device of a server device on a communication network, such as the Internet, and downloading it into an ordinary computer system.

[0091] In addition, when the functions of the prior distribution generation device 1, the demand forecasting device 2, the KPI forecasting device 3, and the management index forecasting device 4 are realized by sharing the functions between an OS (Operating System) and an application program, or by collaboration between an OS and an application program, only the application program portion may be stored in a recording medium or storage device.

[0092] It is also possible to superimpose a computer program on a carrier wave and provide it via a communication network. For example, the computer program may be posted on a bulletin board system (BBS) on the communication network and provided via the communication network. The computer program may then be started and executed under the control of an OS in the same way as other application programs, thereby enabling the above-mentioned processing to be performed.

[0093] It should be noted that the present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to explain the disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and within the meaning of the disclosure equivalent thereto are considered to be within the scope of the present disclosure.

[0094] This application is based on Japanese Patent Application No. 2022-36692, filed on March 9, 2022. The entire specification, claims, and drawings of Japanese Patent Application No. 2022-36692 are incorporated herein by reference. Various aspects of the present disclosure are summarized below as appendices. (Appendix 1) A forecasting system that performs demand forecasting using Bayesian statistics, a prior distribution generation device that generates a prior distribution used in demand forecasting; a demand forecasting device that generates a demand forecast probability distribution, which is a posterior distribution, based on the prior distribution; Equipped with The prior distribution generating device a prior distribution generating unit that generates the prior distribution using supply plan data that indicates a supply plan for an item, and generates prior distribution data that indicates the prior distribution; The demand prediction device a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using a demand model that indicates a demand trend of the product, demand record data that indicates actual demand records of the product, and the prior distribution data, and generates demand forecast data that indicates the demand forecast probability distribution; Prediction system. (Appendix 2) The supply plan data is data indicating a supply plan for an item drawn up by an expert. 10. The prediction system of claim 1. (Appendix 3) the prior distribution generation unit generates a prior distribution using the supply plan data, error data indicating an error between the supply plan and actual supply of the item, and a statistical model indicating a relationship between the supply plan data and the error data, and generates prior distribution data indicating the prior distribution; 3. The prediction system of claim 1 or 2. (Appendix 4) The system further includes a KPI prediction device that converts statistics of the demand forecast probability distribution into KPI (Key Performance Indicator) values ​​to generate KPI data including the KPI probability distribution; The KPI prediction device includes: a KPI data generation unit that generates KPI data including a KPI probability distribution obtained by converting statistics of the demand forecast probability distribution into at least a KPI value based on PSI (Production Sales Inventory) data related to the production, shipment, and inventory of goods; 4. The prediction system of any one of appendixes 1 to 3. (Appendix 5) The KPI prediction device includes: The system further includes a KPI alert determination unit that outputs an alert when the KPI probability distribution satisfies a predetermined KPI alert condition. 10. The prediction system of claim 4. (Appendix 6) a management index prediction device that converts the KPI values ​​of the KPI data into management indexes and generates a management index probability distribution; The management index prediction device has a management index data generation unit that converts KPI values ​​included in the KPI data into accounting numerical values ​​and further converts them into management units based on an accounting conversion master that indicates the correspondence between KPI values ​​and accounting numerical values ​​and a unit conversion master that indicates the correspondence between practical units of goods and management units, generates a management index probability distribution, and generates management index data including the management index probability distribution. 6. The prediction system of claim 4 or 5. (Appendix 7) The management index prediction device The management index probability distribution further includes a management index alert determination unit that outputs an alert when the management index probability distribution satisfies a predetermined management index alert condition. 6. The prediction system of claim 5. (Appendix 8) A demand forecasting device that performs demand forecasting using Bayesian statistics, a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using prior distribution data generated using a demand model that indicates a demand trend of the product, demand record data that indicates actual demand for the product, and supply plan data that indicates a supply plan for the product, and generates demand forecast data that indicates the demand forecast probability distribution; Demand forecasting device. (Appendix 9) A demand forecasting method using Bayesian statistics, generating a prior distribution using supply planning data indicating a supply plan for the item; generating a demand forecast probability distribution, which is a posterior distribution, using a demand model representing a demand trend of the product, demand record data representing actual demand for the product, and prior distribution data representing the prior distribution, and generating demand forecast data representing the demand forecast probability distribution; Demand forecasting methods. (Appendix 10) A computer that uses Bayesian statistics to forecast demand, a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using prior distribution data generated using a demand model that represents a demand trend for the item, demand record data that represents actual demand for the item, and supply plan data that represents a supply plan for the item, and generates demand forecast data that represents the demand forecast probability distribution; A program that functions as a [Explanation of symbols]

[0095] 1 Prior distribution generation device, 2 Demand forecasting device, 3 KPI prediction device, 4 Management index prediction device, 11 Shipping plan data storage unit, 12 Error data storage unit, 13 Statistical model storage unit, 14 Prior distribution generation unit, 15 Prior distribution output unit, 21 Prior distribution acquisition unit, 22 Demand model storage unit, 23 Demand actual data storage unit, 24 Demand forecast data generation unit, 25 Demand forecast data output unit, 31 Demand forecast data acquisition unit, 32 Production confirmation data storage unit, 33 Product master storage unit, 34 Inventory actual data storage unit, 35 KPI data generation unit, 36 KPI data output unit, 37 KPI alert condition storage unit, 38 KPI alert determination unit, 41 KPI data acquisition unit, 42 Accounting conversion master storage unit, 43 Unit conversion master storage unit, 44 Management index data generation unit, 45 Management index data output unit, 46 Management index alert condition storage unit, 47 Management index alert determination unit, 100, 200, 300 prediction system, 101 temporary storage unit, 102 storage unit, 103 calculation unit, 104 input unit, 105 transmission / reception unit, 106 display unit, G1 to G6 graphs.

Claims

1. A forecasting system that performs demand forecasting using Bayesian statistics, a prior distribution generation device that generates a prior distribution used in demand forecasting; a demand forecasting device that generates a demand forecast probability distribution, which is a posterior distribution, based on the prior distribution; Equipped with The prior distribution generating device a prior distribution generating unit that generates the prior distribution using supply plan data that indicates a supply plan for an item, and generates prior distribution data that indicates the prior distribution; The demand prediction device a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using a demand model that indicates a demand trend of the product, demand record data that indicates actual demand records of the product, and the prior distribution data, and generates demand forecast data that indicates the demand forecast probability distribution; Prediction system.

2. The supply plan data is data indicating a supply plan for an item drawn up by an expert. The prediction system of claim 1 .

3. the prior distribution generation unit generates a prior distribution using the supply plan data, error data indicating an error between the supply plan and actual supply of the item, and a statistical model indicating a relationship between the supply plan data and the error data, and generates prior distribution data indicating the prior distribution; The prediction system according to claim 1 or 2.

4. a KPI prediction device that converts statistics of the demand forecast probability distribution into KPI (Key Performance Indicator) values ​​to generate KPI data including the KPI probability distribution; The KPI prediction device a KPI data generation unit that generates KPI data including a KPI probability distribution obtained by converting statistics of the demand forecast probability distribution into at least a KPI value based on PSI (Production Sales Inventory) data related to the production, shipment, and inventory of goods; The prediction system according to claim 1 or 2.

5. The KPI prediction device The system further includes a KPI alert determination unit that outputs an alert when the KPI probability distribution satisfies a predetermined KPI alert condition. The prediction system of claim 4 .

6. a management index prediction device that converts the KPI values ​​of the KPI data into management indexes and generates a management index probability distribution; The management indicator prediction device has a management indicator data generation unit that converts KPI values ​​included in the KPI data into accounting numerical values ​​and further converts them into management units based on an accounting conversion master that indicates the correspondence between KPI values ​​and accounting numerical values ​​and a unit conversion master that indicates the correspondence between practical units of goods and management units, generates a management indicator probability distribution, and generates management indicator data including the management indicator probability distribution. The prediction system of claim 4 .

7. The management index prediction device The management index probability distribution further includes a management index alert determination unit that outputs an alert when the management index probability distribution satisfies a predetermined management index alert condition. The prediction system of claim 6 .

8. A demand forecasting device that performs demand forecasting using Bayesian statistics, a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using prior distribution data generated using a demand model that indicates a demand trend of the product, demand record data that indicates actual demand for the product, and supply plan data that indicates a supply plan for the product, and generates demand forecast data that indicates the demand forecast probability distribution; Demand forecasting device.

9. A computer-implemented demand forecasting method using Bayesian statistics, comprising: generating a prior distribution using supply planning data indicating a supply plan for the item; generating a demand forecast probability distribution, which is a posterior distribution, using a demand model representing a demand trend of the product, demand record data representing actual demand for the product, and prior distribution data representing the prior distribution, and generating demand forecast data representing the demand forecast probability distribution; Demand forecasting methods.

10. A computer that uses Bayesian statistics to forecast demand, a demand forecast data generation unit that generates a demand forecast probability distribution, which is a posterior distribution, using prior distribution data generated using a demand model that represents a demand trend for the item, demand record data that represents actual demand for the item, and supply plan data that represents a supply plan for the item, and generates demand forecast data that represents the demand forecast probability distribution; A program that functions as a

Images