Shipment volume forecasting system, shipment volume forecasting method, and program

Abstract

This shipment amount prediction system includes a prediction unit and an output unit. The prediction unit predicts the future shipment amount of a product on the basis of one or more shipment amount prediction models generated from time-series data which corresponds to the sales period of the product and which includes an explanatory variable related to the distribution of the product and a target variable which is the shipment amount of the product. The output unit outputs the predicted future shipment amount.

Description

Shipment volume forecasting system, shipment volume forecasting method, and recording medium

[0001] The present disclosure relates to a shipping volume forecasting system and the like.

[0002] For example, a wholesaler's ordering personnel may predict the shipment volume of each product based on tacit knowledge. There are also technologies for predicting product shipment volumes. For example, Patent Literature 1 describes a technology for predicting the shipment volume of bottled beverages using past shipment volume records and weather conditions. Furthermore, for example, Patent Literature 2 describes a technology for learning multiple models that each predict shipment volume from each feature quantity based on the priorities of multiple feature quantities calculated according to information indicating the distribution of each of the feature quantities that indicate the characteristics of each period of a demand prediction unit and the shipment volume records, and outputting the sum of the shipment volumes predicted using the multiple models as a demand prediction value.

[0003] JP 2005-078277 A JP 2022-078560 A

[0004] Product shipment volumes fluctuate due to various factors. It can be difficult for wholesalers to predict product shipment volumes.

[0005] An example of an objective of the present disclosure is to provide a shipment volume forecasting system or the like that improves the accuracy of forecasting shipment volumes.

[0006] A shipment volume forecasting system according to one aspect of the present disclosure includes: a forecasting means for forecasting future shipment volumes of a product based on one or more shipment volume forecasting models generated from time-series data including explanatory variables that are information regarding the distribution of the product according to the sales period of the product, and a target variable that is the shipment volume of the product; and an output means for outputting the predicted future shipment volumes.

[0007] A shipment volume forecasting method according to one aspect of the present disclosure forecasts future shipment volumes of a product based on one or more shipment volume forecasting models generated from time series data including explanatory variables that are information about the distribution of the product according to the sales period of the product and a target variable that is the shipment volume of the product, and outputs the forecasted future shipment volumes.

[0008] A program in one aspect of the present disclosure causes a computer to execute a process of predicting future shipment volumes of a product based on one or more shipment volume prediction models generated from time series data including explanatory variables that are information about the distribution of the product according to the sales period of the product and a target variable that is the shipment volume of the product, and outputting the predicted future shipment volumes.

[0009] Each program may be stored in a non-transitory computer-readable recording medium.

[0010] According to the present disclosure, it is possible to improve the accuracy of predicting shipping volume.

[0011] 1 is a block diagram showing an example of the configuration of a shipment volume forecasting system according to embodiment 1. FIG. 2 is an explanatory diagram showing an example of classification of sales periods and types of shipment volume forecasting models. FIG. 3 is a flowchart showing an example of operation of the shipment volume forecasting system according to embodiment 1. FIG. 4 is an explanatory diagram showing an example of connection between a shipment volume forecasting system according to embodiment 2 and other devices. FIG. 5 is a block diagram showing an example of the configuration of a shipment volume forecasting system according to embodiment 2. FIG. 6 is an explanatory diagram showing an example of exclusion of a specific period of shipment volume at the start of release. FIG. 7 is an explanatory diagram showing an example of exclusion of a discontinuous period of product shipment volume. FIG. 8 is an explanatory diagram showing an example of abnormal values. FIG. 9 is an explanatory diagram showing an example of generation of a shipment volume forecasting model according to a sales period. FIG. 10 is an explanatory diagram showing a specific example of explanatory variables. FIG. 11 is an explanatory diagram showing an example of learning period and evaluation period for time series data. FIG. 12 is an explanatory diagram showing an example of level fluctuation of a non-seasonal product and an example of level fluctuation of a seasonal product. FIG. 13 is a flowchart showing example operation 1 of the shipment volume forecasting system according to embodiment 2. FIG. 14 is a flowchart showing example operation 2 of the shipment volume forecasting system according to embodiment 2. FIG. 15 is an explanatory diagram showing an example of the hardware configuration of a computer.

[0012] Hereinafter, with reference to the drawings, embodiments of a shipment volume forecasting system, a shipment volume forecasting method, a program, and a non-transitory recording medium for recording the program according to the present disclosure will be described in detail. The present embodiments do not limit the disclosed technology.

[0013] Here, for example, wholesalers sell products to stores such as convenience stores, drug stores, supermarkets, etc. For example, wholesalers may handle tens of thousands to hundreds of thousands of types of products.

[0014] (Embodiment 1) First, in embodiment 1, basic functions of a shipment volume prediction system will be described. Fig. 1 is a block diagram showing an example of the configuration of a shipment volume prediction system 10 according to embodiment 1. The shipment volume prediction system 10 includes a prediction unit 101 and an output unit 102.

[0015] The prediction unit 101 predicts the future shipment volume of a product based on one or more shipment volume prediction models generated from time-series data including explanatory variables corresponding to the sales period of the product and a target variable representing the shipment volume of the product. Here, the explanatory variables are information related to the distribution of the product. For example, the sales period is the period elapsed from the release date of the product or the period elapsed from the start date of shipment of the product. The sales period may be broadly divided. The release date is, for example, the date on which sales of the product first began, and the shipping start date is, for example, the date on which shipment of the product first began.

[0016] 2 is an explanatory diagram showing examples of broad classifications of sales periods and types of shipment volume forecasting models. In FIG. 2, sales periods are broadly classified into short term, medium term, medium to long term, and long term.

[0017] For example, a sales period of less than six months is called short-term. Products with short-term sales periods are called short-term products. Products with sales periods of six months or more but less than one year and three months are called medium-term. Products with medium-term sales periods are called medium-term products. Products with sales periods of one year and three months or more but less than two years and three months are called medium- to long-term. Products with medium- to long-term sales periods are called medium- to long-term products. Products with sales periods of two years and three months or more are called long-term. Products with long-term sales periods are called long-term products. Here, the ranges of each period, such as short-term, medium-term, medium-to-long-term, and long-term, are merely examples, are not particularly limited, and may differ depending on the product.

[0018] The broad classification of sales periods shown in FIG. 2 is an example, and the broad classification may be short-term, long-term, or the like, and is not particularly limited.

[0019] Specific examples of explanatory variables will be described in detail in embodiment 2. For example, the longer the sales period, the more types of explanatory variables there are, and the shorter the sales period, the fewer types of explanatory variables there are.

[0020] In Figure 2, for example, if the sales period of the product is short, a short-term shipment volume forecast model can be prepared as the shipment volume forecast model. For example, if the sales period of the product is medium, a short-term shipment volume forecast model and a medium-term shipment volume forecast model can be prepared as the shipment volume forecast model. For example, if the sales period of the product is medium to long, a short-term shipment volume forecast model, a medium-term shipment volume forecast model, and a medium- to long-term shipment volume forecast model can be prepared as the shipment volume forecast model. For example, if the sales period of the product is long, a short-term shipment volume forecast model, a medium-term shipment volume forecast model, a medium- to long-term shipment volume forecast model, and a long-term shipment volume forecast model can be prepared as the shipment volume forecast model.

[0021] Then, for example, the prediction unit 101 predicts the future shipment volume of the product based on one or more shipment volume prediction models. For example, the prediction unit 101 determines the future shipment volume to be the shipment volume predicted by one of the one or more shipment volume prediction models selected based on the error between the predicted shipment volume for the evaluation period and the actual shipment volume for the evaluation period for each of the one or more shipment volume prediction models. Alternatively, for example, the prediction unit 101 determines the future shipment volume to be the shipment volume predicted by a model selected based on the ratio of the level of actual shipment volume for the evaluation period to the level of actual shipment volume for the learning period, which can be set according to the sales period. The level of actual shipment volume for the learning period is the level when the statistic of actual shipment volume for a certain period of the learning period is set to 1. The certain period of the learning period may be the learning period itself or a partial period. Furthermore, the level of actual shipment volume for the evaluation period is the level when the statistic of actual shipment volume for a certain period of the evaluation period is set to 1. The certain period of the evaluation period may be the evaluation period itself or a partial period. The statistic is an average or median. An example of model selection will be described in detail using the second embodiment.

[0022] The output unit 102 outputs the predicted future shipment volume. Here, as an output method for outputting the shipment volume, for example, the output unit 102 outputs the shipment volume to an output device. The output device is a display device or an audio output device. For example, the output device may be provided in a terminal device. Hereinafter, an example in which the output unit 102 displays the shipment volume on a terminal device will be described.

[0023] 3 is a flowchart showing an example of the operation of the shipment volume prediction system 10 according to the first embodiment. The prediction unit 101 predicts the future shipment volume of a product based on one or more shipment volume prediction models generated from time-series data including explanatory variables corresponding to the sales period of the product and a target variable representing the shipment volume of the product (step S101). The output unit 102 then outputs the future shipment volume (step S102), and the shipment volume prediction system 10 terminates the process.

[0024] As described above, product shipment volume fluctuates due to various factors, making it difficult to predict product shipment volume. It is particularly difficult to predict the shipment volume of products sold over a long period of time. As described above, in the first embodiment, the shipment volume prediction system 10 predicts future product shipment volume based on one or more shipment volume prediction models generated from time-series data including explanatory variables related to product distribution according to the product's sales period and a target variable representing the product's shipment volume, and outputs the predicted future shipment volume. When the sales period is long, various data is accumulated, resulting in a greater number of available explanatory variables. For example, information on seasonal indices, such as those described in the second embodiment, can only be obtained through the seasons and is only available when the sales period is long, such as one year to several years. For example, the number of explanatory variables available for a product with a one-month sales period is greater than the number of explanatory variables available for a product with a one-year sales period. Furthermore, the explanatory variables available for a product with a one-year sales period include explanatory variables available for a product with a one-month sales period. Therefore, when the sales period is long, multiple models using explanatory variables can be prepared. Therefore, it is possible to generate models tailored to the product's life stage. Therefore, the shipment volume forecasting system 10 can improve the accuracy of forecasting shipment volumes.

[0025] (Embodiment 2) Next, embodiment 2 will be described in detail with reference to the drawings. In embodiment 2, an example of generating a shipment volume forecasting model and an example of selecting one of one or more shipment volume forecasting models will be described in detail. Below, explanations of content that overlaps with the above explanation will be omitted to the extent that the explanation of embodiment 2 is not unclear.

[0026] 4 is an explanatory diagram showing an example of a connection between the shipment volume prediction system according to the second embodiment and other devices. For example, the shipment volume prediction system is connected to a terminal device 21 via a communication network NT. In FIG. 4, the type of the terminal device 21 is not particularly limited, and may be a personal computer (PC), a smartphone, a tablet device, or the like. Note that the terminal device 21 may be pre-installed with an application program that can transmit information to the shipment volume prediction system or output information from the shipment volume prediction system.

[0027] 5 is a block diagram showing an example of the configuration of a shipment volume prediction system according to embodiment 2. The shipment volume prediction system includes a prediction unit 201, an output unit 202, an acquisition unit 203, a generation unit 204, an exclusion unit 205, and a selection unit 206. The shipment volume prediction system further includes the acquisition unit 203, the generation unit 204, the exclusion unit 205, and the selection unit 206 in addition to the shipment volume prediction system according to embodiment 1. The prediction unit 201 has the function of the prediction unit 101 according to embodiment 1 as a basic function. The output unit 202 has the function of the output unit 102 according to embodiment 1 as a basic function.

[0028] (Learning Phase) First, the learning phase will be described. In the learning phase, one or more shipment volume forecasting models are generated based on time-series data other than data from a sales period in which shipment volume fluctuates due to factors other than normal.

[0029] First, the acquiring unit 203 acquires time-series data to be used for learning. Specifically, for example, the acquiring unit 203 acquires time-series data including explanatory variables related to product distribution according to the sales period of the product and a target variable that is the shipment volume of the product.

[0030] <Exclusion of Abnormal Data> Next, the exclusion unit 205 excludes, from the time-series data, data for a period during the sales period in which the shipment volume fluctuates due to factors that are unusual. Specifically, as an exclusion process, for example, the exclusion unit 205 excludes, from the time-series data, data for a period during the sales period in which the shipment volume fluctuates due to factors that are unusual, from the data to be learned. As another exclusion process, for example, the exclusion unit 205 may replace, from the time-series data, data for a period during the sales period in which the shipment volume fluctuates due to factors that are unusual, with statistical data.

[0031] <Exclusion of Initial Release Period> For example, the fluctuating period to be excluded may be the initial release period of a product. FIG. 6 is an explanatory diagram showing an example of excluding an unusual period of initial release shipment volume. In FIG. 6, each graph shows the date on the horizontal axis and the shipment volume on the vertical axis. As shown in the "Initial Upward Variation Example" of FIG. 6, the shipment volume may increase at the initial release due to a large-scale campaign or the like. As shown in the "Initial Downward Variation Example" of FIG. 6, the shipment volume may decrease at the initial release due to low brand recognition. As shown in the "Initial Pinpoint Example" of FIG. 6, the shipment volume may increase in a pinpointed manner at the initial release and become stable. In this way, the initial release may result in a shipment volume that is different from normal.

[0032] Here, if the sales period is short or medium, the period to be learned is short. Therefore, if the sales period is long or medium to long term, the exclusion unit 205 may, for example, exclude data from the period when the product was first released from the time-series data used for learning. For example, the period when the product was first released may be a period from 0 to 3 months from the release date or shipping date. Note that 0 to 3 months is just an example and is not particularly limited.

[0033] Furthermore, the period of unusual shipment volume at the time of release may be specified in more detail. For example, the exclusion unit 205 may identify a period in which the statistical value of shipment volume for the period from the release date or the shipping date is abnormal, and exclude data for the specified period. As described above, the period from the release date or the shipping date is a period from 0 to 3 months after the release date or the shipping date. An example of using an average value as the statistical value is given. In more detail, for example, the exclusion unit 205 excludes data for the period from the release date or the shipping date if the average shipment volume for the period from the release date or the shipping date is more than 2 times or less than 0.5 times the average shipment volume for the entire learning period, and more than 3 times or less than 0.33 times the average shipment volume for the 3 to 6 months period. Each numerical value is an example and is not limited thereto.

[0034] <Exclusion of discontinuous periods> For example, it may be desirable to exclude data from periods in which the shipment volume is zero due to factors unrelated to demand from the data to be learned.

[0035] FIG. 7 is an explanatory diagram showing an example of excluding discontinuous periods of product shipment volume. In FIG. 7, the horizontal axis of each graph represents the date, and the vertical axis represents the shipment volume. As shown in the "Resurrected Product Example" in FIG. 7, some products may be temporarily discontinued and then resume sales. Alternatively, as shown in the "Discontinuous Example" in FIG. 7, an increase in demand due to an epidemic or other cause an increase in shipment volume may result in a shortage of inventory and zero shipment volume. For example, an increase in demand due to an infectious disease epidemic may temporarily cause medicated hand soap or masks to run out of stock, resulting in zero shipment volume despite demand.

[0036] In the "confusing example" in Figure 7, there are some products that have periods when the shipment volume is zero due to a temporary lack of sales, and periods when the shipment volume is low due to a simple lack of demand. In such cases, we want to exclude the periods when there are temporarily no sales from the learning period, but we do not want to exclude the periods when the shipment volume is low due to a lack of demand from the learning period.

[0037] In this way, we want to exclude from the learning period any period in which shipment volume is zero due to factors unrelated to demand, but we do not want to exclude from the learning period any period in which shipment volume is zero due to demand.

[0038] Therefore, the fluctuating period to be excluded may be a predetermined period in which the shipping volume is equal to or less than a predetermined amount. A predetermined period in which the shipping volume is equal to or less than a predetermined amount is, for example, a discontinuous period in which the shipping volume of a product temporarily disappears. The predetermined amount is, for example, 0 or an amount equal to or less than 1% of the average shipping volume during the learning period. The predetermined period is, for example, one to three months. For example, if there is a period in which the shipping volume is equal to or less than 0 during a period of three months or more during the sales period, the exclusion unit 205 excludes the data for that period from the time-series data for learning.

[0039] Furthermore, as in the "Example of a Seasonal Product" in Figure 7, there may be cases where demand is high from spring to autumn, resulting in a large shipment volume, but there is no shipment volume in winter, resulting from a lack of demand. Because demand varies and shipment volume fluctuates depending on the season, it is not desirable to exclude from the learning period periods periods in which shipment volume is zero due to demand. That is, the exclusion unit 205 does not exclude data for periods in which shipment volume decreases at the same time in different years. The fluctuating periods to be excluded may be predetermined periods in which shipment volume is below a predetermined volume, and may be periods other than periods in which shipment volume is below the predetermined volume at the same time in different years. The different years may be the previous year or two years prior. Taking the previous year as an example of a different year, a period in which the period in the previous year in which shipment volume is below the predetermined volume overlaps by 50% or more with the period in the target year in which shipment volume is below the predetermined volume is considered to be a period in which shipment volume is decreasing due to seasonal demand. Therefore, the fluctuating periods to be excluded are predetermined periods in which shipment volume is below a predetermined volume, and may be periods other than periods in which shipment volume is decreasing due to seasonal demand.

[0040] In this way, for example, the exclusion unit 205 excludes from the time-series data data a predetermined period in which the shipment volume is equal to or less than a predetermined volume and in which the period in the previous year in which the shipment volume is equal to or less than the predetermined volume and the period in the target year in which the shipment volume is equal to or less than the predetermined volume do not overlap by 50% or more. As shown in the "confusing example" in Figure 7, data from a period in which the shipment volume is temporarily zero for a short period of about two weeks is not excluded, but data from a period in which the shipment volume is zero for about one to three months is excluded.

[0041] Here, as in the example of excluding the initial release period, if the sales period is short or medium, the period to be learned is short, so if the sales period is long or medium to long term, for example, the exclusion unit 205 may exclude discontinuous periods.

[0042] <Exclusion of Abnormal Values> For example, the exclusion unit 205 may wish to exclude data for a period in which there is an abnormal shipment volume from the time-series data for learning.

[0043] FIG. 8 is an explanatory diagram showing an example of an outlier. For example, if learning is performed using time series data that includes sudden peaks, a shipment volume prediction model that responds to the outliers will be generated. Examples of sudden increases in shipment volume include large-volume transactions and special demand caused by disasters. Such demand is not subject to general shipment volume prediction, so it may be better to treat it as noise. Therefore, if learning is performed using time series data from which outlier data has been excluded, a shipment volume prediction model with stable shipment volume will be generated.

[0044] For example, the aforementioned fluctuating period is a period during which the difference between the shipment volume and the shipment volume statistic is equal to or greater than a predetermined value. Specifically, for example, the exclusion unit 205 excludes, from the training time-series data, shipment volume data whose difference from the shipment volume statistic included in the training time-series data is equal to or greater than a predetermined value. The statistic is the average or median of the shipment volume during the training period.

[0045] As shown in FIG. 8 , if the shipment volume is periodic, it is not treated as an abnormal value. In the wholesale industry, shipment volume may be significantly affected by the day of the week, such as when shipment volume increases once every two weeks on Tuesdays. In this way, shipment volume may periodically increase. Therefore, for example, if the difference between the shipment volume and the average value for the past year is three times or more for each day of the week, the exclusion unit 205 identifies the data for that day as an abnormal value candidate. Furthermore, if an abnormal value candidate occurs with a predetermined frequency on a specific day of the week and for a predetermined period of time, the exclusion unit 205 does not identify the abnormal value candidate for that specific day of the week as an abnormal value. The predetermined frequency is, for example, twice a month. The predetermined period is, for example, two months. On the other hand, if an abnormal value candidate does not occur with a predetermined frequency on a specific day of the week and has not occurred for a predetermined period of time on a specific day of the week, the exclusion unit 205 identifies the abnormal value candidate as an abnormal value and excludes the abnormal value data from the time-series data.

[0046] The sudden peak lasts for a short period of time, such as one day, etc. Therefore, for example, the exclusion unit 205 may replace the shipping volume of the data to be excluded with a statistical amount.

[0047] Here, as in the example of excluding the initial release period, if the sales period is short or medium, the period to be learned is short, so if the sales period is long or medium to long term, for example, the exclusion unit 205 may exclude abnormal values.

[0048] <Learning> Next, the generating unit 204 generates one or more shipment volume prediction models based on time-series data including explanatory variables corresponding to sales periods and a target variable that is the shipment volume. When the excluding unit 205 excludes data for a certain period, the generating unit 204 generates one or more shipment volume prediction models based on the time-series data after the exclusion.

[0049] FIG. 9 is an explanatory diagram showing an example of generating a shipment volume prediction model according to a sales period. For example, the longer the sales period, the more explanatory variables that can be obtained. Furthermore, the longer the sales period, the longer the period that can be used as a learning period. As described in the first embodiment, when the sales period is short, the generating unit 204 generates a shipment volume prediction model using short-term time series data including short-term explanatory variables. Furthermore, when the sales period is long, the generating unit 204 generates a shipment volume prediction model using short-term time series data including short-term explanatory variables and a shipment volume prediction model using long-term time series data including long-term explanatory variables.

[0050] Here, the explanatory variables will be explained. Fig. 10 is an explanatory diagram showing specific examples of explanatory variables. For example, the explanatory variables are information related to product distribution. Specifically, the explanatory variables are at least one of calendar information, information related to product sales, information related to past product shipment volumes, information related to the environment, and information related to the impact of the environment on product shipment volumes.

[0051] The calendar information may include information indicating, for example, a date, a weekday, a Saturday or Sunday, a public holiday, a day of the week, a holiday for the wholesaler, a holiday for the wholesaler's customer, etc. For example, the calendar information indicates what day of the week, whether it is a holiday, etc.

[0052] Information related to product sales includes, for example, sale information and unit price information as shown in Fig. 10. Sale information is information indicating whether a sale has been held and, if so, the sale price. Unit price information is information indicating the unit price of the product.

[0053] The information relating to past shipment volume performance may be, for example, at least one of information indicating trends in past shipment volume performance and statistical information obtained from past shipment volume performance.

[0054] Information representing trends in past shipment volume performance may be, for example, short-term trend information and long-term trend information, as shown in FIG. 10 . The short-term trend information represents, for example, the trend in short-term shipment volume. The trend in short-term shipment volume may be represented, for example, by whether the shipment volume is rising, falling, declining, rising, or stable over a short period of time, and the method of representation is not particularly limited. The long-term trend information represents, for example, the trend in long-term shipment volume. The trend in long-term shipment volume may be represented, for example, by whether the shipment volume is rising, falling, declining, rising, or stable over a long period of time, and the method of representation is not particularly limited.

[0055] Statistical information obtained from past shipment volume performance may be, for example, one-week moving average information by day of the week, four-week moving average information by day of the week, twelve-week moving average information by day of the week, and weekday / holiday moving average information, as shown in FIG. 10 . One-week moving average information by day of the week represents the moving average of one-week shipment volume by day of the week. Four-week moving average information by day of the week represents the moving average of four-week shipment volume by day of the week. Twelve-week moving average information by day of the week represents the moving average of 12-week shipment volume by day of the week. Weekday / holiday moving average information represents the moving average of shipment volume by weekday / holiday.

[0056] The environmental information may be, for example, information indicating temperature as shown in Fig. 10. The temperature may be an average temperature, etc. The environmental information may also be information indicating weather, wind strength, and humidity.

[0057] Information relating to the environmental impact on product shipment volume includes, for example, product category-specific seasonal index information and product-specific seasonal index information, as shown in Fig. 10. The product category-specific seasonal index indicates, for example, the sales volume of a product category relative to the average annual shipment volume. The product-specific seasonal index indicates the sales volume of a product relative to the average shipment volume.

[0058] 10, the explanatory variables that can be used vary depending on the sales period. For example, calendar information, product sales information, one-week moving average information by day of the week, and four-week moving average information by day of the week are explanatory variables that can be used in any sales period.

[0059] For example, the short-term trend, the 12-week moving average by day of the week, and the moving average by weekday and holiday are explanatory variables that can be used when the sales period is medium-term, medium-long term, or long term.

[0060] For example, average temperature information is an explanatory variable that can be used when the sales period is medium to long term or long term.

[0061] Long-term trends, product category-specific seasonal index information, and product-specific seasonal index information are explanatory variables that can be used when the sales period is long. For example, product category-specific seasonal index information and product-specific seasonal index information are useful for forecasting shipment volume if data covering one to two years is available, but are not useful for forecasting shipment volume if the period is short.

[0062] As shown in FIG. 10, the longer the sales period, the greater the variety of explanatory variables that can be used.

[0063] The generation unit 204 generates one or more shipment volume prediction models with a learning period that can be set according to the sales period, based on time-series data including explanatory variables corresponding to the sales period and a target variable that is the shipment volume. The learning period is the period over which the time-series data is learned. Furthermore, the evaluation period is the period over which the time-series data is evaluated. As described above, a longer sales period allows for the accumulation of various data, resulting in a larger number of available explanatory variables. Furthermore, a longer sales period allows for various learning periods to be set. For example, for a product that was recently launched and has a sales period of two months, the learning period is approximately one month. On the other hand, for a product that has been on sale for two years, the learning period may be one month, six months, one year, or one year and ten months.

[0064] The learning period and evaluation period will now be described with reference to Fig. 11. Fig. 11 is an explanatory diagram showing an example of the learning period and evaluation period for time-series data. Here, the learning period and evaluation period when training each shipment volume forecasting model in the case where the sales period is long will be described as an example.

[0065] For example, data for a part of the time-series data may be used as the evaluation period data, and data for the remaining part of the time-series data may be used as the learning period data. For example, the evaluation period may be the most recent month.

[0066] Here, for example, when the sales period is long, the generation unit 204 can generate multiple shipment volume prediction models. For example, when the sales period is long, explanatory variables that can be used when the sales period is shorter than the long term can also be used. Therefore, the generation unit 204 can generate a shipment volume prediction model using explanatory variables that can be used when the sales period is long, a shipment volume prediction model using explanatory variables that can be used when the sales period is medium to long term, a shipment volume prediction model using explanatory variables that can be used when the sales period is medium term, and a shipment volume prediction model using explanatory variables that can be used when the sales period is short term. Here, explanatory variables that can be used when the sales period is long are called explanatory variables for the long term. Explanatory variables that can be used when the sales period is medium to long term are called explanatory variables for the medium to long term. Explanatory variables that can be used when the sales period is medium term are called explanatory variables for the medium term. Explanatory variables that can be used when the sales period is short term are called explanatory variables for the short term.

[0067] For example, the generation unit 204 sets the learning period for learning a shipment volume prediction model using long-term explanatory variables to a long term. Specifically, for example, the generation unit 204 generates a long-term shipment volume prediction model based on long-term time-series data including long-term explanatory variables and shipment volumes.

[0068] For example, the generation unit 204 sets the learning period for learning a shipment volume forecasting model using medium- to long-term explanatory variables to the medium- to long-term. Specifically, for example, the generation unit 204 generates a medium- to long-term shipment volume forecasting model based on medium- to long-term time-series data including medium- to long-term explanatory variables and shipment volumes.

[0069] For example, the generating unit 204 defines the learning period for learning a shipment volume forecasting model using explanatory variables for the medium term as the medium term. Specifically, for example, the generating unit 204 generates a shipment volume forecasting model for the medium term based on medium-term time-series data including explanatory variables and shipment volumes for the medium term.

[0070] For example, the generating unit 204 sets the learning period for learning a shipment volume prediction model using short-term explanatory variables to a short period. Specifically, for example, the generating unit 204 generates a short-term shipment volume prediction model based on short-term time-series data including short-term explanatory variables and shipment volumes.

[0071] Here, we briefly explain why the learning period is divided into those using long-term explanatory variables, those using medium- to long-term explanatory variables, those using medium-term explanatory variables, and those using short-term explanatory variables. Empirically, the more explanatory variables there are and the longer the learning period, the higher the accuracy of shipment volume prediction models. However, for example, shipment volume trends may change recently. For products that experience such changes, the predictive accuracy of a learning model generated using short-term explanatory variables and data from the most recent short-term learning period may be higher than a shipping volume prediction model generated using older shipment volumes. Thus, depending on the product, the shipment volume trend may change suddenly. For products that experience such changes, the predictive accuracy of a learning model may be higher if the learning period is after the change occurs than if the learning period is longer. For this reason, the learning periods are divided.

[0072] (Inference Phase) Next, the inference phase will be described. Here, for example, the prediction unit 201 predicts future shipments using a shipment prediction model selected from one or more shipment prediction models.

[0073] <Selection of Shipment Volume Forecasting Model Based on Error in Evaluation Period> First, an example will be described in which a shipment volume forecasting model with a small error is selected from one or more shipment volume forecasting models.

[0074] The prediction unit 201 predicts the shipment volume for the evaluation period for each of the one or more shipment volume prediction models generated by the generation unit 204. Taking the example of a case where the sales period is long, a long-term shipment volume prediction model, a medium- to long-term shipment volume prediction model, a medium-term shipment volume prediction model, and a short-term shipment volume prediction model are generated. Specifically, for example, the prediction unit 201 provides explanatory variables for the evaluation period that are long-term in the time-series data to the long-term shipment volume prediction model to predict the shipment volume for the evaluation period. For example, the prediction unit 201 provides explanatory variables for the evaluation period that are medium- to long-term in the time-series data to the medium- to long-term shipment volume prediction model to predict the shipment volume for the evaluation period. For example, the prediction unit 201 provides explanatory variables for the evaluation period that are medium-term in the time-series data to the medium-term shipment volume prediction model to predict the shipment volume for the evaluation period. For example, the prediction unit 201 provides explanatory variables for the short term, which are explanatory variables for the evaluation period in the time series data, to a short-term shipment volume prediction model to predict the shipment volume for the evaluation period.

[0075] The selection unit 206 selects one of the one or more shipment volume forecasting models based on the error between the predicted shipment volume for the evaluation period for each of the one or more generated shipment volume forecasting models and the actual shipment volume for the evaluation period.

[0076] Specifically, for example, the selection unit 206 first calculates the error between the shipment volume predicted for the evaluation period and the actual shipment volume of the time-series data for the evaluation period for each of the generated shipment volume prediction models. Then, for example, the selection unit 206 determines the shipment volume prediction model to be used for prediction from the generated shipment volume prediction models based on the error calculated for each of the generated shipment volume prediction models. As an example of selection based on the error, the selection unit 206 selects the shipment volume prediction model with the smallest error from the generated shipment volume prediction models. Note that an error rate may be used instead of the error. The error rate may be the average weekly error rate during the evaluation period. For example, the selection unit 206 selects the shipment volume prediction model with the smallest error rate.

[0077] More specifically, a case where the sales period is long will be taken as an example. The selection unit 206 calculates the error between the shipment volume predicted using the long-term shipment volume prediction model and the actual shipment volume for the evaluation period. The selection unit 206 calculates the error between the shipment volume predicted using the medium- to long-term shipment volume prediction model and the actual shipment volume for the evaluation period. The selection unit 206 calculates the error between the shipment volume predicted using the medium-term shipment volume prediction model and the actual shipment volume for the evaluation period. The selection unit 206 calculates the error between the shipment volume predicted using the short-term shipment volume prediction model and the actual shipment volume for the evaluation period. For example, the selection unit 206 selects the shipment volume prediction model with the smallest calculated error from the long-term shipment volume prediction model, the medium- to long-term shipment volume prediction model, the medium-term shipment volume prediction model, and the short-term shipment volume prediction model.

[0078] <Selection of a Shipment Quantity Forecasting Model Based on Characteristics of Past Actual Shipment Quantities of Products> The selection unit 206 selects one of one or more shipment quantity forecasting models based on characteristics of past actual shipment quantities of products. Specifically, for example, the selection unit 206 selects a model from the shipment quantity forecasting models in which the ratio of the performance level during the evaluation period to the performance level during the learning period is within a predetermined range. As described in the first embodiment, the performance level during the learning period is the level obtained when the statistical value of the actual shipment quantity for a certain period within the learning period is set to 1. Furthermore, the performance level during the evaluation period is the level obtained when the statistical value of the actual shipment quantity for a certain period within the evaluation period is set to 1. The statistical value is the average or median. The ratio of the performance level during the evaluation period to the performance level during the learning period is the value obtained by dividing the performance level during the learning period by the performance level during the evaluation period. As described above, empirically, a shipment quantity forecasting model with a larger number of explanatory variables and a longer learning period often has higher shipment quantity forecast accuracy. Therefore, for example, the selection unit 206 may prioritize shipment volume forecasting models with longer learning periods and select models in which the ratio of the performance level in the evaluation period to the performance level in the learning period is within a predetermined range.

[0079] More specifically, for example, when the sales period is long, the selection unit 206 determines whether the ratio of the performance level of the evaluation period to the performance level of the learning period of the long-term shipment volume forecasting model is within a predetermined range. If the ratio of the performance level of the evaluation period to the performance level of the learning period of the long-term shipment volume forecasting model is within the predetermined range, the selection unit 206 selects the long-term shipment volume forecasting model. If the ratio of the performance level of the evaluation period to the performance level of the learning period of the long-term shipment volume forecasting model is not within the predetermined range, the selection unit 206 determines whether the ratio of the performance level of the evaluation period to the performance level of the learning period of the medium- to long-term shipment volume forecasting model is within the predetermined range. If the ratio of the performance level of the evaluation period to the performance level of the learning period of the medium- to long-term shipment volume forecasting model is within the predetermined range, the selection unit 206 selects the medium- to long-term shipment volume forecasting model. If the ratio of the performance level of the evaluation period to the performance level of the learning period of the medium- to long-term shipment volume forecasting model is not within the predetermined range, the selection unit 206 determines whether the ratio of the performance level of the evaluation period to the performance level of the learning period of the medium- to long-term shipment volume forecasting model is within the predetermined range. The selection unit 206 selects the shipment volume forecast model for the medium term if the ratio of the performance level of the evaluation period to the performance level of the learning period of the shipment volume forecast model for the medium term is within a predetermined range.The selection unit 206 selects the shipment volume forecast model for the short term if the ratio of the performance level of the evaluation period to the performance level of the learning period of the shipment volume forecast model for the medium term is not within a predetermined range.

[0080] Here, for seasonal products, the ratio of the performance level of the evaluation period to the performance level of the learning period may be smoothed based on the level difference due to the season of the evaluation period. FIG. 12 is an explanatory diagram showing an example of level fluctuations for non-seasonal products and an example of level fluctuations for seasonal products. It is expected that periodicity can be observed in the level fluctuations of seasonal products. In contrast, it is expected that periodicity cannot be observed in the level fluctuations of non-seasonal products. As shown in FIG. 12, whether or not there is seasonality in the shipment volume of a product can be estimated from the level fluctuations. The selection unit 206 corrects the ratio of the performance level of the evaluation period to the performance level of the learning period using an annual level ratio. The annual level ratio is the ratio of the level of the month including the three months of the evaluation period and the same season to the level including a one-year period.

[0081] <Prediction of Future Shipment Volume> Next, the acquisition unit 203 acquires explanatory variables related to product distribution according to the sales period. For example, the acquisition unit 203 acquires explanatory variables according to the selected shipment volume prediction model. Then, the prediction unit 201 predicts the future shipment volume based on the acquired explanatory variables using the selected shipment volume prediction model.

[0082] Here, future calendar information, information on product sales, etc. may be prepared in advance by a person in charge. If the current date is March 2023 and the shipping volume for each day in April 2023 is predicted, calendar information for April 2023, sale information for April 2023, and unit price information for April 2023 are prepared.

[0083] The information on actual shipment volume may be, for example, information on actual shipment volume for the same month of the previous year or information on actual shipment volume for the previous month. For example, when the shipment volume for each day in April 2023 is predicted, information on actual shipment volume for April 2022 or information on actual shipment volume for March 2023 may be used as information on actual shipment volume for April 2023.

[0084] The environmental information may include weather forecast information and past environmental information. For example, when predicting the shipping volume for each day in April 2023, the temperature information for April 2023 may include information about the temperature in the weather forecast for April 2023 and information about the temperature in April 2022.

[0085] For the information on the environmental impact on product shipment volume, information on the environmental impact on product shipment volume for the same month of the previous year may be used. For example, when the shipment volume for each day in April 2023 is predicted, information on the environmental impact on product shipment volume for April 2022 may be used as the information on the environmental impact on product shipment volume for April 2023.

[0086] For example, when a long-term shipment volume forecasting model is selected by the selection unit 206, the acquisition unit 203 acquires future calendar information, sale information, unit price information, short-term trend information, long-term trend information, one-week moving average information by day of the week, four-week moving average information by day of the week, 12-week moving average information by day of the week, weekday / holiday moving average information, average temperature information, seasonal index information by product category, and seasonal index information by product, as shown in Fig. 10. Then, for example, the prediction unit 201 inputs the information acquired by the acquisition unit 203 into the long-term shipment volume forecasting model, and acquires future shipment volumes from the long-term shipment volume forecasting model.

[0087] Furthermore, for example, when a medium-term shipment volume forecasting model is selected by the selection unit 206, the acquisition unit 203 acquires future calendar information, sale information, unit price information, short-term trend information, long-term trend information, one-week moving average information by day of the week, four-week moving average information by day of the week, 12-week moving average information by day of the week, and weekday / holiday moving average information, as shown in Fig. 10. Then, the prediction unit 201 inputs each piece of information acquired by the acquisition unit 203 into the medium-term shipment volume forecasting model, and acquires future shipment volumes from the medium-term shipment volume forecasting model.

[0088] In addition, examples of future shipment volume predictions using a short-term shipment volume prediction model and a medium- to long-term shipment volume prediction model are similar to the examples of predictions using a long-term shipment volume prediction model and a medium-term shipment volume prediction model, so detailed explanations will be omitted.

[0089] The output unit 202 then outputs the predicted future shipment volume. As described in the first embodiment, the method for outputting the shipment volume is not particularly limited. For example, the output unit 202 displays the shipment volume on a terminal device. The display format may be a graph in chronological order, and is not particularly limited.

[0090] 13 is a flowchart showing an operation example 1 of the shipment volume prediction system according to embodiment 2. In operation example 1, an example will be described in which a future shipment volume is predicted using a shipment volume prediction model with a low error in the evaluation period.

[0091] First, the acquiring unit 203 acquires time-series data including explanatory variables available for a sales period and a target variable, which is a shipment volume (step S201). The excluding unit 205 excludes data from the time-series data for a period in which the shipment volume fluctuates due to factors other than normal (step S202).

[0092] Next, the generating unit 204 generates one or more shipment volume prediction models based on the time series data of the learning period corresponding to the sales period from among the time series data after the exclusion (step S203).

[0093] The prediction unit 201 predicts the shipment volume for the evaluation period for each of the generated shipment volume prediction models based on the explanatory variables for the evaluation period in the acquired time-series data (step S204).Then, the selection unit 206 selects a shipment volume prediction model based on the difference between the predicted shipment volume for each of the generated shipment volume prediction models and the actual shipment volume for the evaluation period in the acquired time-series data (step S205).

[0094] The acquiring unit 203 acquires input data according to the selected shipment volume forecasting model (step S206). The forecasting unit 201 forecasts the shipment volume based on the input data using the selected shipment volume forecasting model (step S207).

[0095] The output unit 202 outputs the predicted shipment volume (step S208), and the shipment volume prediction system ends the process.

[0096] 14 is a flowchart showing an operation example 2 of the shipment volume prediction system according to embodiment 2. In operation example 2, an example is described in which a future shipment volume is predicted using a shipment volume prediction model selected based on the characteristics of the past shipment volume of a product.

[0097] First, the acquiring unit 203 acquires time-series data including explanatory variables available for a sales period and a target variable that is a shipment volume (step S211). The excluding unit 205 excludes data from the time-series data for a period in which the shipment volume fluctuates due to factors other than normal (step S212).

[0098] Next, the generating unit 204 generates one or more shipment volume prediction models based on the time series data of the learning period corresponding to the sales period from among the time series data after the exclusion (step S213).

[0099] The prediction unit 201 selects a shipment volume prediction model (step S214). In step S214, the selection unit 206 selects a shipment volume prediction model in order of the longest learning period, in which the ratio of the performance level in the evaluation period to the performance level in the learning period is within a predetermined range.

[0100] The acquiring unit 203 acquires input data according to the selected shipment volume prediction model (step S215). The predicting unit 201 predicts the shipment volume based on the input data using the selected shipment volume prediction model (step S216).

[0101] The output unit 202 outputs the predicted shipment volume (step S217), and the shipment volume prediction system ends the process.

[0102] As described above, in the second embodiment, the shipment volume prediction system generates one or more shipment volume prediction models based on time-series data, and predicts the future shipment volume of a product based on the generated one or more shipment volume prediction models. In this way, a shipment volume prediction model according to the life stage of the product is generated. Therefore, it is possible to improve the accuracy of shipment volume prediction.

[0103] The shipment volume forecasting system selects a shipment volume forecasting model based on the error between the shipment volume for an evaluation period predicted for each of one or more shipment volume forecasting models and the actual shipment volume for the evaluation period, and forecasts future shipment volumes based on the selected shipment volume forecasting model. This makes it possible to forecast future shipment volumes using an estimated shipment volume forecasting model with a small error, thereby improving the accuracy of shipment volume forecasts.

[0104] Furthermore, the shipment volume forecasting system selects a shipment volume forecasting model based on the ratio of the actual shipment volume level for the evaluation period to the actual shipment volume level for the learning period, which can be set according to the sales period, and predicts future shipment volume based on the selected shipment volume forecasting model. This makes it possible to predict future shipment volume using a learning period shipment volume forecasting model with low variance in shipment volume between the learning period and the most recent evaluation period. This therefore improves the accuracy of shipment volume forecasting.

[0105] The shipment volume forecasting system also excludes data from the time-series data for a period during the sales period when shipment volume fluctuates due to factors other than those that are normal, and generates one or more shipment volume forecasting models based on the excluded time-series data. The fluctuating period may be one day. This makes it possible to exclude abnormal data during learning, thereby preventing the generation of a model that makes abnormal predictions.

[0106] The period of fluctuation may be the initial period when the product is released, thereby eliminating abnormal shipping periods at the beginning of the release.

[0107] Furthermore, the fluctuating period may be a predetermined period in which the shipping volume is equal to or less than a predetermined volume. This allows discontinuous periods to be excluded. However, for seasonal products in which the shipping volume increases in some seasons and decreases in other seasons, this is a characteristic of the product's shipping volume, so it is not desirable to exclude them. Therefore, the fluctuating period is a predetermined period in which the shipping volume is equal to or less than a predetermined volume, and is a period other than a period in which the shipping volume is equal to or less than the predetermined volume at the same time in different years.

[0108] Furthermore, a period of fluctuation is a period in which the difference between the shipping volume and a statistical quantity such as the average or median of the shipping volume is equal to or greater than a predetermined value, thereby making it possible to exclude sudden noise.

[0109] This concludes the description of the embodiment. The embodiment may be modified. Modifications are shown below.

[0110] In the second embodiment, an example has been described in which future shipment volumes are predicted using a shipment volume prediction model selected based on the characteristics of past shipment volumes of a product. The selection unit 206 may select which model to generate in the learning phase. Then, the generation unit 204 may generate the selected model.

[0111] This concludes the description of the modified examples. Each embodiment and each modified example may be combined. In the embodiments, the shipment volume forecasting systems 10 and 20 may be configured to include some of the functional units and information.

[0112] Furthermore, the embodiments are not limited to the above examples and can be modified in various ways. Furthermore, the configuration of the shipment volume prediction systems 10 and 20 in the embodiments is not particularly limited. For example, the shipment volume prediction systems 10 and 20 may be realized by a single device, such as a single server. When each functional unit of the shipment volume prediction systems 10 and 20 is realized by a single device, the single device may be called, for example, a shipment volume prediction device, an information processing device, or the like, and is not particularly limited. Alternatively, the shipment volume prediction systems 10 and 20 in the embodiments may be realized by different devices for different functions or data. For example, each functional unit may be configured by multiple servers and realized as the shipment volume prediction systems 10 and 20. For example, the shipment volume prediction system 10 may be realized by a database server including each DB (Database) and a server having each functional unit. Furthermore, in the embodiments, each piece of information may include some of the above-mentioned information. Furthermore, each piece of information may include information other than the above-mentioned information. Each piece of information may be divided into multiple pieces of information in more detail, or may be organized into a single DB. Furthermore, a system including the shipment volume forecasting systems 10 and 20 and the terminal device 21 may be configured.

[0113] Furthermore, the process of generating information to be displayed on the terminal device 21 may be performed by the output units 102 and 202. This process may also be performed by the terminal device 21.

[0114] (Example of Computer Hardware Configuration) Next, an example of a hardware configuration in which each device, such as the shipment quantity forecasting systems 10 and 20 and the terminal device 21 described in the embodiment, is implemented by a computer will be described. Fig. 15 is an explanatory diagram showing an example of a computer hardware configuration. For example, some or all of each device can be implemented using any combination of a computer 80 and a program as shown in Fig. 15.

[0115] The computer 80 includes, for example, a processor 801, a ROM (Read Only Memory) 802, a RAM (Random Access Memory) 803, and a storage device 804. The computer 80 also includes a communication interface 805 and an input / output interface 806. The components are connected to each other, for example, via a bus 807. The number of each component is not particularly limited, and there may be one or more of each component.

[0116] The processor 801 controls the entire computer 80. The processor 801 may be, for example, a central processing unit (CPU), a digital signal processor (DSP), a graphics processing unit (GPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, or a combination thereof, and is not particularly limited.

[0117] The computer 80 also includes a ROM 802, a RAM 803, and a storage device 804 as storage units. Examples of the storage device 804 include semiconductor memory such as flash memory, a hard disk drive (HDD), and a solid state drive (SSD). For example, the storage device 804 stores an operating system (OS) program, application programs, and programs related to the embodiments. Alternatively, the ROM 802 stores application programs and programs related to the embodiments. The RAM 803 is used as a work area for the processor 801.

[0118] The processor 801 also loads programs stored in the storage device 804, ROM 802, etc. The processor 801 then executes each process coded in the program. The processor 801 may also download various programs via the communication network NT. The processor 801 also functions as a part or all of the computer 80. The processor 801 may then execute the processes or instructions in the illustrated flowchart based on the program.

[0119] The communication interface 805 is connected to a communication network NT such as a LAN (Local Area Network) or a WAN (Wide Area Network) via a wireless or wired communication line. The communication network NT may be composed of multiple communication networks NT. As a result, the computer 80 is connected to external devices and external computers 80 via the communication networks NT. The communication interface 805 serves as an interface between the communication network NT and the inside of the computer 80. The communication interface 805 also controls the input and output of data from external devices and external computers 80.

[0120] Furthermore, the input / output interface 806 is connected to at least one of an input device, an output device, and an input / output device. The connection method may be wireless or wired. Examples of the input device include a keyboard, a mouse, and a microphone. Examples of the output device include a display device, a lighting device, and an audio output device that outputs audio. Examples of the input / output device include a touch panel display. Note that the input device, output device, and input / output device may be built into the computer 80 or may be external.

[0121] The hardware configuration of the computer 80 is an example. The computer 80 may have some of the components shown in FIG. 15 . The computer 80 may have components other than those shown in FIG. 15 . For example, the computer 80 may have a drive device or the like. The processor 801 may then read programs and data stored on a recording medium attached to the drive device or the like into the RAM 803. Examples of non-transitory tangible recording media include optical disks, flexible disks, magneto-optical disks, and USB (Universal Serial Bus) memories. As described above, the computer 80 may have input devices such as a keyboard and a mouse. The computer 80 may have an output device such as a display. The computer 80 may also have an input device, an output device, and an input / output device.

[0122] The computer 80 may also include various sensors (not shown). The types of sensors are not particularly limited. The computer 80 may also include an imaging device capable of capturing images or videos.

[0123] This concludes the description of the hardware configuration of each device. There are various variations in the method of realizing each device. For example, each device may be realized by any combination of a different computer and program for each component. Furthermore, multiple components of each device may be realized by any combination of a single computer and program.

[0124] Furthermore, some or all of the components of each device may be realized by circuits for specific applications. Furthermore, some or all of the components of each device may be realized by general-purpose circuits such as FPGAs (Field Programmable Gate Arrays). Furthermore, some or all of the components of each device may be realized by a combination of circuits for specific applications and general-purpose circuits. These circuits may be a single integrated circuit. Alternatively, these circuits may be divided into multiple integrated circuits. The multiple integrated circuits may be connected via a bus or the like.

[0125] Furthermore, when some or all of the components of each device are realized by a plurality of computers, circuits, etc., the plurality of computers, circuits, etc. may be centrally located or distributed.

[0126] The shipment volume forecasting method described in the embodiment is realized by being executed by the shipment volume forecasting systems 10 and 20. Also, for example, the shipment volume forecasting method is realized by having a computer such as a server or a terminal device 21 execute a program prepared in advance.

[0127] The programs described in the embodiments are recorded on a computer-readable recording medium such as a HDD, SSD, flexible disk, optical disk, magneto-optical disk, or USB memory. The programs are then read from the recording medium and executed by a computer. The programs may also be distributed via a communication network NT.

[0128] The functions of each of the components of the shipment volume forecasting systems 10 and 20 according to the embodiments described above may be realized by dedicated hardware, such as a computer. Alternatively, each component may be realized by software. Alternatively, each component may be realized by a combination of hardware and software.

[0129] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above embodiments. The configuration and details of each of the present disclosures may include embodiments to which various modifications that would be apparent to those skilled in the art are applied within the scope of the present disclosure. The present disclosure may include embodiments in which the details described herein are appropriately combined or substituted as necessary. For example, details described using a particular embodiment may also be applied to other embodiments to the extent that no contradiction occurs. For example, although multiple operations are described in sequence in the form of a flowchart, the order of description does not limit the order in which the multiple operations are performed. Therefore, when implementing the embodiments, the order of the multiple operations may be changed as long as it does not interfere with the content.

[0130] Some or all of the above-described embodiments can be described as follows: However, some or all of the above-described embodiments are not limited to the following.

[0131] (Supplementary Note 1) A shipment volume forecasting system comprising: a forecasting means for forecasting a future shipment volume of a product based on one or more shipment volume forecasting models generated from time-series data including an explanatory variable that is information about the distribution of the product according to a sales period of the product, and a target variable that is the shipment volume of the product; and an output means for outputting the predicted future shipment volume. (Supplementary Note 2) The shipment volume forecasting system according to Supplementary Note 1, wherein the forecasting means sets the future shipment volume to a shipment volume predicted by one of the models selected based on an error between the shipment volume for an evaluation period predicted for each of the one or more shipment volume forecasting models and the actual shipment volume for the evaluation period. (Supplementary Note 3) The shipment volume forecasting system according to Supplementary Note 1, wherein the forecasting means sets the future shipment volume to a shipment volume predicted by one of the models selected based on a ratio of the level of actual shipment volume for an evaluation period to the level of actual shipment volume for a learning period corresponding to the sales period. (Supplementary Note 4) The shipment volume forecasting system according to any one of Supplementary Note 1 to Supplementary Note 3, wherein the explanatory variables are at least any of calendar information, information on sales of the product, information on past shipment volumes of the product, information on the environment, and information on the impact of the environment on shipment volumes of the product. (Supplementary Note 5) The shipment volume forecasting system according to any one of Supplementary Note 1 to Supplementary Note 4, further comprising: a generation means for generating the one or more shipment volume forecast models based on the time-series data, wherein the forecasting means predicts the future shipment volume based on the one or more shipment volume forecast models generated by the generation means. (Supplementary Note 6) The shipment volume forecasting system according to Supplementary Note 5, further comprising: an exclusion means for excluding, from the time-series data, data for a period during the sales period in which shipment volume is fluctuating due to factors other than those usual, wherein the generation means generates the one or more shipment volume forecast models based on the excluded time-series data. (Supplementary Note 7) The shipment volume forecasting system according to Supplementary Note 6, wherein the fluctuating period is the period during which the product is initially released. (Supplementary Note 8) The shipment volume forecasting system according to Supplementary Note 6 or Supplementary Note 7, wherein the period of fluctuation is a predetermined period during which the shipment volume is equal to or less than a predetermined volume.(Supplementary Note 9) The shipment volume forecasting system according to Supplementary Note 8, wherein the fluctuating period is a period other than a period during which the shipment volume is equal to or less than the predetermined volume at the same time in different years. (Supplementary Note 10) The shipment volume forecasting system according to any of Supplementary Notes 6 to 9, wherein the fluctuating period is a period during which a difference between the shipment volume and a statistical quantity of the shipment volume is equal to or greater than a predetermined value. (Supplementary Note 11) The shipment volume forecasting system according to Supplementary Note 10, wherein the statistical quantity is an average or median of the shipment volume. (Supplementary Note 12) The shipment volume forecasting system according to any of Supplementary Notes 1 to 11, wherein the longer the sales period, the greater the types of explanatory variables available, and the shorter the sales period, the fewer the types of explanatory variables available. (Supplementary Note 13) A shipment volume forecasting method comprising: forecasting a future shipment volume of a product based on one or more shipment volume forecasting models generated from time-series data including explanatory variables that are information regarding the distribution of the product according to the sales period of the product, and a target variable that is the shipment volume of the product; and outputting the predicted future shipment volume. (Supplementary Note 14) A computer-readable non-transitory recording medium having recorded thereon a program that causes a computer to execute the following process: predicting a future shipment volume of a product based on one or more shipment volume prediction models generated from time-series data including explanatory variables that are information related to the distribution of the product according to a sales period of the product and a target variable that is the shipment volume of the product, and outputting the predicted future shipment volume. (Supplementary Note 15) A computer-readable non-transitory recording medium having recorded thereon a program that causes a computer to execute the following process: predicting a future shipment volume of a product based on one or more shipment volume prediction models generated from time-series data including explanatory variables that are information related to the distribution of the product according to a sales period of the product and a target variable that is the shipment volume of the product, and outputting the predicted future shipment volume.

[0132] 10, 20 Shipping volume forecasting system 21 Terminal device 80 Computer 101, 201 Forecasting unit 102, 202 Output unit 203 Acquisition unit 204 Generation unit 205 Exclusion unit 206 Selection unit 801 Processor 802 ROM 803 RAM 804 Storage device 805 Communication interface 806 Input / output interface 807 Bus NT Communication network

Claims

1. An explanatory variable that is information about the distribution of the product according to the sales period of the product; a prediction means for predicting a future shipment volume of the product based on one or more shipment volume prediction models generated from time-series data including a target variable that is the shipment volume of the product; an output means for outputting the predicted future shipping volume; A shipment volume forecasting system comprising:

2. the prediction means sets the future shipment volume to a shipment volume predicted by one of the one or more shipment volume prediction models selected based on an error between the shipment volume for the evaluation period predicted for each of the one or more shipment volume prediction models and the actual shipment volume for the evaluation period. The shipping volume forecasting system according to claim 1 .

3. the prediction means sets the future shipment volume to a shipment volume predicted by one of the models selected based on a ratio of a level of actual shipment volume in an evaluation period to a level of actual shipment volume in a learning period corresponding to the sales period; The shipping volume forecasting system according to claim 1 .

4. The explanatory variables are at least one of calendar information, information on sales of the product, information on past shipment volume of the product, information on the environment, and information on the impact of the environment on the shipment volume of the product. The shipping volume forecasting system according to claim 1 or 2.

5. generating means for generating the one or more shipment volume forecast models based on the time-series data; the prediction means predicts the future shipment volume based on the one or more shipment volume prediction models generated by the generation means. The shipping volume forecasting system according to claim 1 or 2.

6. an exclusion means for excluding data for a period during the sales period in which the shipment volume fluctuates due to factors different from normal from the time-series data; the generation means generates the one or more shipment volume forecast models based on the excluded time-series data. The shipping volume forecasting system according to claim 5 .

7. The fluctuating period is the initial period when the product is released. The shipping volume forecasting system according to claim 6 .

8. The period during which the shipment volume fluctuates is a predetermined period during which the shipment volume is equal to or less than a predetermined volume. The shipping volume forecasting system according to claim 6 .

9. An explanatory variable that is information about the distribution of the product according to the sales period of the product; predicting a future shipment volume of the product based on one or more shipment volume prediction models generated from time-series data including a target variable that is the shipment volume of the product; outputting the predicted future shipping volume; Shipment volume forecasting method.

10. On the computer, An explanatory variable that is information about the distribution of the product according to the sales period of the product; predicting a future shipment volume of the product based on one or more shipment volume prediction models generated from time-series data including a target variable that is the shipment volume of the product; outputting the predicted future shipping volume; A program that executes a process.