Prediction device, prediction method, and ordering support device

Abstract

According to the present invention, a prediction device comprises: a first prediction unit for predicting a first predicted number of sales in a future prediction period on the basis of a past actual number of sales of a commodity; a reference value setting unit for setting a reference value on the basis of specific information indicating a specific situation related to a change in the number of sales in the prediction period; and a second prediction unit for determining, on the basis of the reference value, whether or not to set a second predicted number of sales, to be output as a predicted number of sales of the commodity, to the same value as the first predicted number of sales.

Description

Prediction device, prediction method, and order support device

[0001] The present disclosure relates to a prediction device, a prediction method, and an order support device.

[0002] Patent Document 1 discloses an ordering support system that includes a demand forecasting model that has learned the actual past demand for a product, and a demand forecasting unit that calculates the predicted demand for the product using the demand forecasting model.

[0003] Japanese Patent No. 6814302

[0004] However, because the demand forecasting model is based on learning from past long-term sales performance, if an unusual situation occurs that affects the demand for a product, the predicted demand calculated using the demand forecasting model may deviate significantly from the actual demand.

[0005] An object of the present disclosure is to provide a technology that can predict demand while taking into account cases where unusual circumstances that affect the demand for a product occur.

[0006] A prediction device according to one embodiment includes a first prediction unit that predicts a first predicted sales volume for a future prediction period based on actual past sales volumes of a product; a reference value setting unit that sets a reference value based on specific information that indicates specific circumstances related to fluctuations in sales volumes during the prediction period; and a second prediction unit that determines, based on the reference value, whether a second predicted sales volume to be output as the predicted sales volume of the product should be the same as the first predicted sales volume.

[0007] A prediction method according to one embodiment predicts a first predicted sales number for a future prediction period based on the actual past sales numbers of a product, sets a reference value based on specific information indicating specific circumstances related to fluctuations in sales numbers during the prediction period, and determines, based on the reference value, whether a second predicted sales number to be output as the predicted sales number of the product should be the same as the first predicted sales number.

[0008] An ordering support device according to one embodiment supports ordering of the product using the prediction method described above.

[0009] These comprehensive or specific aspects may be realized as a system, an apparatus, a method, an integrated circuit, a computer program, or a recording medium, or may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

[0010] According to the present disclosure, it is possible to predict the demand quantity while taking into consideration the occurrence of unusual circumstances that affect the demand quantity of a product.

[0011] FIG. 1 is a block diagram showing a configuration example of an ordering support system according to the first embodiment; FIG. 2 is a graph showing an example of a case where the safety rate, sales volume, and inventory volume change as predicted, according to the first embodiment; FIG. 3 is a diagram showing an example of a product list screen according to the first embodiment; FIG. 4 is a diagram showing an example of an ordering support screen for manual ordering according to the first embodiment; FIG. 5 is a diagram showing an example of an ordering support screen for automatic ordering according to the first embodiment; FIG. 6 is a diagram showing an example of an ordering support screen for fixed ordering according to the first embodiment;

[0012] Hereinafter, embodiments of the present disclosure will be described in detail with appropriate reference to the drawings. However, more detailed description than necessary may be omitted. For example, detailed descriptions of well-known matters and redundant descriptions of substantially identical configurations may be omitted. This is to avoid unnecessary redundancy in the following description and to facilitate understanding by those skilled in the art. Note that the accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0013] First Embodiment <System Configuration> FIG. 1 is a block diagram showing an example of the configuration of an order support system 10 according to a first embodiment.

[0014] The ordering support system 10 is a device that supports a purchaser 1 of a product in determining the order quantity of a product having an expiration date so that opportunity loss and waste loss are minimized.

[0015] For example, the ordering support system 10 displays on the terminal 12 the recommended order quantity for each product that minimizes opportunity loss and waste loss. The purchaser 1 inputs the order quantity for each product into the ordering device 4, referring to the recommended order quantity for each product displayed on the terminal 12. The ordering device 4 transmits the order quantity for each product input by the purchaser 1 to the product delivery center 3. The product delivery center 3 delivers each product in the order quantity received from the ordering device 4 to, for example, the store of the purchaser 1.

[0016] In this embodiment, the product is a food or drink with an expiration date that is provided to customers at a store, and the orderer 1 is described as a person who orders the food or drink (for example, a store manager). However, the product is not limited to this. For example, the product may be a food or drink with an expiration date that is provided to patients hospitalized in a hospital, and the orderer 1 may be a person who coordinates orders for the food or drink. For example, the product may be a medicine with an expiration date (for example, eye drops) sold at a pharmacy, and the orderer 1 may be a person who orders the medicine.

[0017] The ordering support system 10 will be described in detail below.

[0018] The ordering support system 10 includes an actual sales quantity DB (DataBase) 41, a sales quantity prediction unit 31, a predicted sales quantity DB 42, an expiration date master 47, a safety rate calculation unit 33, an inventory quantity DB 43, an order quantity DB 44, a recommended order quantity calculation unit 32, and an input display control unit 34.

[0019] The actual sales number DB 41 is a DB that stores the number of items actually sold in the past (hereinafter referred to as the actual sales number) for each product. For example, the actual sales number DB 41 stores, for each product, a product ID, a product name, a past date, and the actual sales number of the product on that date, in association with each other.

[0020] The sales volume prediction unit 31 calculates the sales volume of a product at each future date and time (hereinafter referred to as the predicted sales volume) based on the past actual sales volume of the product stored in the actual sales volume DB 41. The sales volume prediction unit 31 calculates the predicted sales volume for each product. The sales volume prediction unit 31 registers the calculated predicted sales volume at each future date and time in the predicted sales volume DB 42.

[0021] The sales volume prediction unit 31 may calculate the predicted sales volume according to the following (A1) and / or (A2).

[0022] (A1) The sales number prediction unit 31 calculates predicted sales numbers based on statistical analysis. For example, the sales number prediction unit 31 statistically analyzes past actual sales numbers of a product stored in the actual sales number DB 41, and calculates predicted sales numbers of the product at a future date and time based on the analysis results. For example, the sales number prediction unit 31 statistically analyzes the actual sales numbers for multiple past weeks to identify sales trends for the week (e.g., sales trends by day of the week), and calculates predicted sales numbers for a future week based on the identified sales trends.

[0023] (A2) The sales number prediction unit 31 calculates the predicted sales number using AI (artificial intelligence). For example, the sales number prediction unit 31 calculates the predicted sales number for a target date and time based on weather information, etc., for a future target date and time using a learning model (e.g., a deep learning model) that learns the relationship between actual sales numbers and weather information using past actual sales numbers of the product and past weather information as learning data. Weather information may include, for example, weather forecasts such as sunny, cloudy, rainy, and snowy, temperature forecasts, humidity forecasts, and atmospheric pressure forecasts. The sales number prediction unit 31 may acquire weather information for the target date and time from a predetermined weather information server 5. Furthermore, the sales number prediction unit 31 calculates the predicted sales number for a future target date and time using a learning model that learns the relationship between actual sales numbers and calendar information using past actual sales numbers of the product and past calendar information as learning data. Calendar information may include, for example, dates, days of the week, holidays, national events, anniversaries, and festivals. The sales forecasting unit 31 may acquire calendar information for the target date and time from a predetermined calendar information server 6 .

[0024] The sales volume prediction unit 31 may register both the predicted sales volume calculated by the method (A1) above and the predicted sales volume calculated by the method (A2) above in the predicted sales volume DB 42.

[0025] The predicted sales volume DB 42 holds the predicted sales volume of each product at each future date and time registered by the sales volume prediction unit 31. For example, for each product, the predicted sales volume DB 42 holds a product ID, a product name, a future date and time, and a predicted sales volume of the product at that date and time, in association with each other.

[0026] The expiration date master 47 holds the expiration date of each product. For example, the expiration date master 47 holds the product ID, product name, and expiration date for each product in association with each other.

[0027] The inventory quantity DB 43 stores the actual inventory quantity of each product in the store. For example, the inventory quantity DB 43 stores the product ID, product name, date and time, and the inventory quantity of the product at that date and time in association with each other.

[0028] The order quantity DB 44 stores the number of items already ordered for each product in the store. For example, the order quantity DB 44 stores the item ID, the item name, the order date and time, the stocking date and time, and the order quantity of the item ordered on the order date and time (i.e., the stocking quantity of the item on the stocking date and time) in association with each other.

[0029] The safety rate calculation unit 33 calculates the safety rate of the product based on the product's expiration date stored in the expiration date master 47, the predicted sales volume of the product stored in the predicted sales volume DB 42, and the inventory volume of the product stored in the inventory volume DB 43. Next, the safety rate will be described assuming that the expiration date of the product is 48 hours, for example.

[0030] <About the safety margin> If there is no inventory at the start of work on day X and the number of items received on day X is less than the actual number of items sold on day X, an opportunity loss will occur. Therefore, to prevent this opportunity loss from occurring, inventory at the start of work on day X (i.e., inventory that has been carried over from the previous day and that has not passed its expiration date) is used. This means that no opportunity loss will occur unless the actual number of items sold on day X exceeds the sum of the number of items in stock at the start of work on day X and the number of items received on day X. On the other hand, if the inventory at the start of work on day X is sold within its expiration date (i.e., within 48 hours of the items being received), it will not become a waste loss.

[0031] Therefore, if the following conditions can be maintained: "All inventory at the start of business is sold within its expiration date" and "On any given day, the number of products sold will never exceed the total number of products in stock at the start of business and the number of products received that day," then both opportunity loss and waste loss can be reduced.

[0032] Therefore, a safety factor is defined as an index related to the inventory quantity at the start of business hours. The safety factor is calculated using the following formula (1A), formula (1B), or formula (1C): Safety factor on day X = (actual inventory quantity at the start of business hours on day X) / (actual sales quantity on day X) (1A) Safety factor on day X = (actual inventory quantity at the start of business hours on day X) / (predicted sales quantity on day X) (1B) Safety factor on day X = (predicted inventory quantity at the start of business hours on day X) / (predicted sales quantity on day X) (1C) The safety factor calculation unit 33 calculates the safety factor for each date of each product using the above formula (1A), formula (1B), or formula (1C). The safety factor calculation unit 33 may use the above formula (1A) when calculating the safety factor for a past date, the above formula (1B) when calculating the safety factor for the current date, and the above formula (1C) when calculating the safety factor for a future date.

[0033] This safety factor indicates the extent to which the demand for the predicted sales quantity can be met with the inventory at the start of business on day X. Note that inventory that has expired on day X is discarded, so the inventory quantity at the start of business on day X, which is used to define the safety factor for day X, indicates the inventory quantity that will not expire on day X. Therefore, if the condition "all of the inventory at the start of business will be sold on day X" is met (in terms of the forecast), the condition "all of the inventory at the start of business will be sold within the expiration date" can be met (in terms of the forecast). In other words, the condition "all of the inventory at the start of business will be sold within the expiration date" can be met reliably. In other words, if the safety factor is less than 100%, no waste loss will occur in the forecast. Note that if the expiration date is expected to arrive after day X, even if the safety factor exceeds 100%, it may be possible to sell the inventory after day (X + 1) and before the expiration date. However, since there is little benefit in securing excess inventory at the risk of incurring waste loss, it is generally desirable to manage inventory so that the safety factor is below 100%. Furthermore, in order to meet (in terms of predicted values) the condition that "the sales volume for any given day will not exceed the total of the inventory at the start of the day and the inventory for that day," it is best to avoid a safety margin of 0%. This is because if the safety margin were 0%, the entire sales volume for that day would need to be met solely by the inventory for that day, but maintaining this state is extremely difficult. The reason for this will be explained in detail below.

[0034] Generally, it takes a certain period of time (hereinafter referred to as α days) from ordering to receiving the goods, so the quantity received on that day will be the quantity ordered α days in advance. In other words, if you try to meet the demand for day X with a safety margin of 0%, you will have to perfectly predict the sales volume for day X α days in advance. Note that if you can operate with a safety margin of 0%, the inventory will never expire, so if you can perfectly predict the sales volume, it is desirable to maintain a safety margin of 0%. However, since it is extremely difficult to make such a perfect prediction, a realistic approach is to place orders so that the safety margin is highly likely to reduce both opportunity loss and waste loss.

[0035] The recommended order quantity calculation unit 32 calculates the recommended order quantity for today's order based on the forecast sales quantity for the target period from today (the order date) to the date the ordered product is received, which is obtained from the forecast sales quantity DB 42, and the safety factor for each date in the target period, which is obtained from the safety factor calculation unit 33. Details of the method for calculating the recommended order quantity will be described later (see FIGS. 2A and 2B).

[0036] The input display control unit 34 displays an order support screen including the safety factor and the recommended order quantity on the terminal 12. The input display control unit 34 may be connected to the terminal 12 via a predetermined communication network (for example, a wired LAN (Local Area Network) or a wireless LAN).

[0037] The terminal 12 displays an order support screen including the safety factor and the recommended order quantity based on instructions from the input / display control unit 34. Examples of the terminal 12 include a PC, smartphone, tablet terminal, etc. The purchaser 1 determines the order quantity of products based on the recommended order quantity displayed on the terminal 12 and inputs the determined order quantity into the ordering device 4. This allows the purchaser 1 to order products in an order quantity that minimizes opportunity loss and waste loss. Note that because the purchaser 1 can determine the order quantity if they can at least confirm the recommended order quantity, the terminal 12 may omit displaying information other than the recommended order quantity (e.g., safety factor, etc.). Furthermore, the terminal 12 may be connected to the ordering device 4 via a communication network, and the ordering device 4 may automatically order the recommended order quantity of products based on instructions from the terminal 12. In this case, the purchaser 1 does not need to determine the order quantity, so the terminal 12 may omit displaying the recommended order quantity.

[0038] The terminal 12 may appropriately reflect the actual sales number of the product, which changes over time, in the actual sales number DB 41, and may appropriately reflect the inventory number of the product, which changes over time, in the inventory number DB 43. The terminal 12 may also reflect the order number of the product for which an order has been completed in the order number DB 44. Alternatively, the actual sales number and inventory number may be automatically reflected in the actual sales number DB and inventory number DB 43 in response to an order from a customer.

[0039] At least one of the sales result DB 41, inventory DB 43, and order quantity DB 44 may be located in a so-called cloud or may be located locally (for example, in a store). Alternatively, a DB that is a copy or synchronized version of at least one of the sales result DB 41, inventory DB 43, and order quantity DB 44 located in the cloud may be located locally. In this case, the terminal 12 may access the locally located DB instead of the DB located in the cloud.

[0040] Furthermore, a robot that automatically retrieves products from inventory in response to customer orders may be installed in a store that sells products. This reduces mistakes by store staff, as the staff only need to provide the products retrieved by the robot to the customer.

[0041] 2A is a diagram showing an example of the order support screen 100 according to embodiment 1. The order support screen 100 may be generated by the input / display control unit 34 and displayed on the terminal 12.

[0042] In this example, the start time is 8:00, the expiration date of the product is 48 hours, and the ordered product will be received at 15:00, two days after the order date. However, these start time, expiration date, and receiving time are only examples. Furthermore, the order time and receiving time do not have to be once a day, but may be two or more times a day.

[0043] The order support screen 100 includes a product list area 101 , a graph area 102 , and an information area 103 .

[0044] A list of products is displayed in the product list area 101. The purchaser 1 selects one product from the product list area 101.

[0045] The graph area 102 displays a line graph (hereinafter referred to as a sales number graph 104) showing the change in sales volume over time for a product selected from the product list area 101, and a line graph (hereinafter referred to as a stock quantity graph 105) showing the change in inventory volume over time. Note that the line graph is just an example, and other types of graphs may also be used. The sales number graph 104 and the inventory quantity graph 105 may be calculated by the recommended order quantity calculation unit 32 and drawn in the graph area 102.

[0046] The safety factor and the inventory quantity for each date are displayed in the information area 103. The input display control unit 34 displays the safety factor for each date calculated by the safety factor calculation unit 33 in the information area 103. The input display control unit 34 displays the inventory quantity for each date in the information area 103 based on the order quantity for each date acquired from the order quantity DB 44.

[0047] Next, the safety factor and inventory quantity will be described in detail with reference to FIG. 2A.

[0048] On February 9th (yesterday), the safety margin was 12% and the number of plates in stock was 46. The safety margin for February 9th was calculated by dividing the actual number of plates in stock at the start of work on February 9th by the number of plates sold on February 9th. The number of plates in stock on February 9th corresponds to the number of plates ordered on February 7th.

[0049] As of 8:00 AM on February 10th (today), the safety margin is 56% and the number of plates in stock is 40. The safety margin for February 10th is calculated by (actual inventory at the start of work on February 10th) / (forecast sales for February 10th). The actual inventory at the start of work on February 10th matches the actual inventory at the end of work on February 9th. Additionally, the number of plates in stock on February 10th corresponds to the number of orders placed on February 8th.

[0050] At 8:00 on February 11th (one day later), the safety margin is 38% and the number of plates in stock is 40. The safety margin for February 11th is calculated by (predicted inventory at the start of work on February 11th) / (predicted sales on February 11th). The predicted inventory at the start of work on February 11th is calculated by (actual inventory at the start of work on February 10th) + (inventory on February 10th) - (predicted sales on February 10th). Note that the predicted inventory at the start of work on February 11th is the same as the predicted inventory at the end of work on February 10th. The inventory on February 11th corresponds to the order placed on February 9th (yesterday).

[0051] At 8:00 on February 12th (two days later), the safety margin is 13% and the number of plates in stock is 60. The safety margin for February 12th is calculated by (predicted inventory at the start of work on February 12th) / (predicted sales on February 12th). The predicted inventory at the start of work on February 12th is calculated by (predicted inventory at the start of work on February 11th) + (inventory on February 11th) - (predicted sales on February 11th). Note that the predicted inventory at the start of work on February 12th is the same as the predicted inventory at the end of work on February 11th. The inventory on February 12th corresponds to the order quantity scheduled for February 10th (today).

[0052] Here, the recommended order quantity calculation unit 32 calculates the recommended order quantity for February 10 (today) (i.e., equivalent to the inventory quantity for February 12) based on the 13% safety margin for February 12 and the predicted sales quantity for February 12. For example, the recommended order quantity calculation unit 32 calculates the recommended order quantity for inventory on February 12 so that it is equal to or greater than the predicted sales quantity for February 12 minus the quantity that can be covered by the 13% safety margin. The input display control unit 34 displays the calculated recommended order quantity on the terminal 12. The purchaser 1 may input the order quantity into the ordering device 4 with reference to the recommended order quantity displayed on the terminal 12. This allows the purchaser 1 to easily place an order that minimizes the sum of opportunity loss and waste loss.

[0053] As described above, to prevent opportunity loss on February 12, the recommended order quantity must be equal to or greater than the predicted sales quantity for February 12 minus the quantity that can be covered with a safety margin of 13% (hereinafter referred to as the "daily shortage quantity"). However, if an order is placed exactly for this daily shortage quantity, there is a risk of opportunity loss on February 12 if the actual sales quantity on February 12 exceeds the predicted sales quantity. However, the recommended order quantity calculation unit 32 calculates the recommended order quantity to be stocked on February 12 so as to satisfy a predetermined safety margin for the following day, February 13. This adds a certain margin to the recommended order quantity beyond the daily shortage quantity. Therefore, according to this embodiment, even if the actual sales quantity on February 12 slightly exceeds the predicted sales quantity, this margin prevents opportunity loss. On the other hand, even if the actual sales quantity is slightly lower than the predicted sales quantity and the inventory quantity at the start of business on February 13 is higher than expected, as long as the inventory quantity is equal to or less than the actual sales quantity on February 13, the product will be consumed within the expiration date (48 hours) on February 13, and no waste will occur.

[0054] Furthermore, if the purchaser 1 places more importance on reducing waste loss than on opportunity loss, the recommended order quantity may be a number lower than the predicted sales quantity minus the number that can be covered by the safety margin.

[0055] In addition, by checking the sales quantity graph 104 and inventory quantity graph 105 shown in Figure 2A, the purchaser 1 can visually confirm the relationship between the changes in inventory quantity at the start of business corresponding to the safety factor on each date and the inventory quantity after storage.

[0056] FIG. 2B is a graph showing an example of a case in which the safety rate, sales volume, and inventory volume change as predicted according to the first embodiment.

[0057] As shown in FIG. 2B, the recommended order quantity calculation unit 32 may calculate the recommended order quantity so that the daily safety rate is a predetermined value (for example, 30%).

[0058] For example, if a product has a 48-hour expiration date, it is received on March 7th, and the remaining inventory at the end of work on March 7th must be consumed by the end of work on the following day, March 8th. The inventory at the end of work on March 7th will be the inventory at the start of work on the following day, March 8th, which corresponds to a safety margin of 30% on March 8th, and will be sold preferentially on March 8th. Therefore, no waste will be generated. Also, a portion of a product received at 3:00 p.m. on March 8th will be consumed on March 8th, and will become partial inventory at the end of work on March 8th without incurring any opportunity loss. The inventory at the end of work on March 8th will be the inventory at the start of work on the following day, March 9th, which corresponds to a safety margin of 30% on March 9th, and will be sold preferentially on March 9th. This process is repeated daily, so as shown in FIG. 2B, if orders are placed to satisfy an appropriate safety margin (e.g., 30%) and the sales and inventory quantities change as predicted, no waste or opportunity loss will be generated.

[0059] Although the above description is based on an example in which the expiration date is 48 hours, the same applies when the expiration date is longer than 48 hours (for example, 72 hours, 96 hours, etc.). In other words, by setting an appropriate safety margin so that the inventory at the start of each day is consumed within the subsequent expiration date, and placing orders to satisfy that safety margin, if the sales volume and inventory volume change as predicted, neither waste nor opportunity loss will occur.

[0060] 3 is a diagram showing an example of the product list screen 120 according to Embodiment 1. The product list screen 120 may be generated by the input display control unit 34 and displayed on the terminal 12.

[0061] The product list screen 120 has a calendar area 121 and a graph area 122 .

[0062] A monthly calendar is displayed in the calendar area 121. The purchaser 1 selects a date for which the purchaser wishes to display a graph in the graph area 122 from the calendar displayed in the calendar area 121.

[0063] The graph area 122 displays a bar graph 123 showing the order quantity of each product on the date selected in the calendar area 121. The size of the frame of the bar graph 123 indicates the number of orders on the selected date. In other words, the frame of the bar graph 123 becomes larger as the number of orders for a product increases.

[0064] The breakdown of the bar graph 123 indicates the percentage of inventory, the percentage of opportunity loss, and the percentage of waste loss of the order quantity. In the bar graph 123, each percentage may be shown in a different color or pattern. For example, the percentage of inventory may be shown in green, the percentage of opportunity loss may be shown in red, and the percentage of waste loss may be shown in purple. If it is predicted that none of the order quantity will result in opportunity loss or waste loss (i.e., all will be sold), the percentage of inventory in the bar graph 123 will be 100%. If a1% of the order quantity is predicted to result in opportunity loss, the percentage of opportunity loss in the bar graph 123 will be a1%. If a2% of the order quantity is predicted to result in waste loss, the percentage of waste loss in the bar graph 123 will be a2%. Note that the bar graph is just an example, and other types of graphs (e.g., pie charts) may also be used.

[0065] 4 is a diagram showing an example of an order support screen 140 for manual ordering according to Embodiment 1. The order support screen 140 may be generated by the input / display control unit 34 and displayed on the terminal 12.

[0066] The ordering support screen 140 for manual ordering includes a graph area 141, a manual ordering tab 142, an automatic ordering tab 143, a fixed ordering tab 144, a recommended order quantity area 151, an order quantity input area 152, an inventory quantity at the end of business hours area 153, a waste loss area 154, an opportunity loss area 155, and a total loss area 156. The ordering support screen 140 for manual ordering shown in FIG. 4 is displayed when the manual ordering tab 142 is selected.

[0067] The graph area 141 displays, for each date, a line graph showing the change in predicted sales quantity over time (hereinafter referred to as predicted sales quantity graph 145) and a bar graph showing the change in predicted inventory quantity over time (hereinafter referred to as predicted inventory quantity graph 146). Furthermore, the graph area 141 displays, for each date, a line graph showing the change in the number of waste losses over time (hereinafter referred to as waste loss graph 147) and a bar graph showing the change in the number of opportunity losses over time (hereinafter referred to as opportunity loss graph 148). Note that the bar graph and line graph are merely examples, and other types of graphs may be used.

[0068] The recommended order quantity area 151 displays the recommended order quantity for the target date and time (April 4th in FIG. 4) calculated by the recommended order quantity calculation unit 32.

[0069] The order quantity is input by the purchaser 1 in the order quantity input area 152 .

[0070] The stock quantity at closing time area 153 displays the predicted stock quantity at the closing time of the day before the target date and time (i.e., the opening time of the target date and time).

[0071] The waste loss area 154 displays the number and amount of waste loss that is predicted when the order quantity input in the order quantity input area 152 is placed.

[0072] The opportunity loss area 155 displays the number and amount of opportunity loss that is predicted when the order quantity input in the order quantity input area 152 is placed.

[0073] The total loss area 156 displays the total of the predicted amount of waste loss and the amount of opportunity loss.

[0074] When the purchaser 1 inputs an order quantity in the order quantity input area 152 while referring to the recommended order quantity displayed in the recommended order quantity area 151, the predicted stock quantity graph 146, waste loss graph 147, and opportunity loss graph 148 in the graph area 141 change according to the input order quantity. This allows the purchaser 1 to visually check the changes in waste loss and opportunity loss and determine an order quantity that will reduce waste loss and opportunity loss.

[0075] The manual ordering shown in FIG. 4 is suitable for ordering when a group reservation has already been made, or for ordering products for which the accuracy of predicted sales figures is not very high.

[0076] Fig. 5 is a diagram showing an example of an order support screen 160 for automatic ordering according to embodiment 1. The order support screen 160 for automatic ordering shown in Fig. 5 is displayed when the automatic ordering tab 143 is selected. The order support screen 160 may be generated by the input display control unit 34 and displayed on the terminal 12.

[0077] The order support screen 160 for automatic ordering includes a graph area 141, a manual ordering tab 142, an automatic ordering tab 143, a fixed ordering tab 144, a recommended order quantity area 151, an order quantity area 161, an inventory quantity at the end of business hours area 153, a waste loss area 154, an opportunity loss area 155, and a total loss area 156.

[0078] A graph similar to that shown in FIG. 4 is displayed in the graph area 141 .

[0079] The recommended order quantity area 151 displays the recommended order quantity for the target date and time (April 4th in FIG. 4) calculated by the recommended order quantity calculation unit 32.

[0080] The order quantity area 161 automatically displays the order quantity corresponding to the recommended order quantity displayed in the recommended order quantity area 151 .

[0081] The stock quantity at closing time area 153 displays the predicted stock quantity at the closing time of the day before the target date and time (i.e., the opening time of the target date and time).

[0082] The waste loss area 154 displays the number and amount of waste loss that is predicted when the order quantity displayed in the order quantity area 161 is placed.

[0083] The opportunity loss area 155 displays the number and amount of the opportunity loss that is predicted when the order quantity displayed in the order quantity area 161 is placed.

[0084] The total loss area 156 displays the total of the predicted amount of waste loss and the amount of opportunity loss.

[0085] The recommended order quantity calculation unit 32 displays in the graph area 141 a predicted stock quantity graph 146, a waste loss graph 147, and an opportunity loss graph 148 according to the order quantity displayed in the order quantity area 161. This allows the purchaser 1 to visually confirm changes in waste loss and opportunity loss when an order is automatically placed based on the recommended order quantity.

[0086] The automatic ordering shown in FIG. 5 is suitable for ordering products with a relatively short expiration date and / or products that are sold in large quantities per day.

[0087] 6 is a diagram showing an example of an order support screen 180 for fixed orders according to embodiment 1. The order support screen 180 for fixed orders shown in FIG. 6 is displayed when the fixed order tab 144 is selected. The order support screen 180 may be generated by the input / display control unit 34 and displayed on the terminal 12.

[0088] The order support screen 180 for fixed orders includes a graph area 141, a manual order tab 142, an automatic order tab 143, a fixed order tab 144, a recommended order quantity area 151, a lower limit input area 181, a fixed order input area 182, an inventory quantity at the end of business hours area 153, a waste loss area 154, an opportunity loss area 155, and a total loss area 156.

[0089] A graph similar to that shown in FIG. 4 is displayed in the graph area 141 .

[0090] The recommended order quantity area 151 displays the recommended order quantity for the target date and time (April 4th in FIG. 4) calculated by the recommended order quantity calculation unit 32.

[0091] In the lower limit input area 181, the purchaser 1 inputs the lower limit of the stock quantity.

[0092] In the fixed order input area 182, the purchaser 1 inputs the order quantity to be automatically ordered when the stock quantity falls below the lower limit value input in the lower limit input area 181.

[0093] The stock quantity at closing time area 153 displays the predicted stock quantity at the closing time of the day before the target date and time (i.e., the opening time of the target date and time).

[0094] The waste loss area 154 displays the number and amount of waste loss that is predicted when an order is placed for the number of orders entered in the fixed order input area 182 .

[0095] The opportunity loss area 155 displays the number and amount of opportunity loss that is predicted when the order quantity input in the fixed order input area 182 is placed.

[0096] The total loss area 156 displays the total of the predicted amount of waste loss and the amount of opportunity loss.

[0097] When the purchaser 1 inputs an order quantity in the fixed order input area 182 while referring to the recommended order quantity displayed in the recommended order quantity area 151, the predicted stock quantity graph 146, waste loss graph 147, and opportunity loss graph 148 in the graph area 141 change according to the input order quantity. This allows the purchaser 1 to determine a fixed order quantity while visually checking the changes in waste loss and opportunity loss.

[0098] The fixed order shown in FIG. 6 is suitable for ordering products with a relatively long expiration date and / or products that are not sold in large numbers per day.

[0099] <Use of Short-Term and Long-Term Actual Sales Volumes> Next, a method for calculating the predicted sales volume using the short-term and long-term actual sales volumes will be described.

[0100] The sales result DB 41 may store short-term sales results and long-term sales results. The short-term sales results may be sales results for a period of, for example, one week or one month. The long-term sales results may be sales results for a period of, for example, one year or two years.

[0101] The sales volume prediction unit 31 calculates a short-term predicted sales volume based on the short-term actual sales volume and registers it in the predicted sales volume DB 42. The sales volume prediction unit 31 also calculates a long-term predicted sales volume based on the long-term actual sales volume and registers it in the predicted sales volume DB 42. The short-term predicted sales volume may be read as the first predicted sales volume, and the long-term predicted sales volume may be read as the second predicted sales volume.

[0102] In addition, the sales volume prediction unit 31 may calculate the short-term predicted sales volume based on statistical analysis as described above in (A1), and may calculate the long-term predicted sales volume using AI as described above in (A2).

[0103] The safety factor calculation unit 33 calculates a short-term safety factor using the short-term predicted sales volume registered in the predicted sales volume DB 42. The safety factor calculation unit 33 also calculates a long-term safety factor using the long-term predicted sales volume registered in the predicted sales volume DB 42. The short-term safety factor may be read as a first safety factor, and the long-term safety factor may be read as a second safety factor.

[0104] The recommended order quantity calculation unit 32 calculates the short-term recommended order quantity using the short-term safety factor, and calculates the long-term recommended order quantity using the long-term safety factor. Note that the short-term recommended order quantity may be read as the first recommended order quantity, and the long-term recommended order quantity may be read as the second recommended order quantity. The recommended order quantity calculation unit 32 may calculate the recommended order quantity using the following method (B1) or (B2).

[0105] (B1) The recommended order quantity calculation unit 32 compares the short-term recommended order quantity with the long-term recommended order quantity and determines the recommended order quantity based on the comparison result. For example, if the difference between the short-term recommended order quantity and the long-term recommended order quantity is less than a predetermined threshold, the recommended order quantity calculation unit 32 selects either the short-term recommended order quantity or the long-term recommended order quantity and sets it as the recommended order quantity. For example, if the difference between the short-term recommended order quantity and the long-term recommended order quantity is equal to or greater than a predetermined threshold, the recommended order quantity calculation unit 32 calculates the recommended order quantity by varying the weighting of the short-term recommended order quantity and the long-term recommended order quantity, as shown in (B2) below.

[0106] (B2) The recommended order quantity calculation unit 32 determines the recommended order quantity using the following formula: Recommended order quantity = (β1 × recommended short-term order quantity) + (β2 × recommended long-term order quantity) The short-term coefficient β1 and the long-term coefficient β2 may be determined, for example, by the following (B2-1) or (B2-2).

[0107] (B2-1) For example, in normal mode, β1 = 0.5 and β2 = 0.5 (i.e., the short-term coefficient and the long-term coefficient are equal), if the most recent accuracy is quite high, β1 = 0.8 and β2 = 0.2 (i.e., the short-term coefficient is larger than the long-term coefficient), and if this year's climate is significantly different from last year's climate, β1 = 1.0 and β2 = 0 (i.e., the long-term coefficient is 0). The smaller the difference between the recommended short-term and long-term order quantities, the higher the most recent accuracy may be determined to be.

[0108] (B2-2) Determine the short-term coefficient β1 and the long-term coefficient β2 based on the correlation between the sales figures for a certain period in the past (e.g., April 2021) and the sales figures for the same period this year (e.g., April 2022). For example, if the correlation coefficient r is equal to or greater than a predetermined threshold (e.g., 0.5), the normal mode is set to β1 = 0.5 and β2 = 0.5 (i.e., the short-term coefficient and the long-term coefficient are equal), and if the most recent accuracy is quite high, the normal mode is set to β1 = 0.8 and β2 = 0.2 (i.e., the short-term coefficient is larger than the long-term coefficient). If the correlation coefficient r is less than a predetermined threshold (e.g., 0.5), the normal mode is set to β1 = 0.9 and β2 = 0.1 (i.e., the short-term coefficient is larger than the long-term coefficient) to avoid being influenced by past data.

[0109] Second Embodiment In a second embodiment, the same reference numerals are used for the components already described in the first embodiment, and the description thereof may be omitted.

[0110] FIG. 7 is a block diagram showing an example of the configuration of an order support system 10 according to the second embodiment.

[0111] The order support system 10 according to the second embodiment includes a store information DB 48 in addition to the components included in the order support system 10 shown in FIG.

[0112] The store information DB 48 is a DB that stores store information, which is information about each store that sells products. Examples of information included in the store information include the location conditions of the store, the surrounding environment of the store, the distance from the product delivery center 3 to the store, weather information for the store location, events taking place around the store, etc.

[0113] The sales number prediction unit 31 calculates the predicted sales number for each store, taking into consideration the store information in the store information DB 48. For example, if the store information includes information that an event that will attract many people will be held near a certain store on a target prediction date and time, the sales number prediction unit 31 calculates the predicted sales number for the target date and time to be higher than the predicted sales number under normal circumstances.

[0114] <When Data in the Sales Amount DB is Insufficient> The sales amount of a new product immediately after its release is not stored in the sales amount DB 41. In such a case, the sales amount prediction unit 31 may generate a tentative sales amount of the new product using the sales amount of an existing product similar to the new product, and use the tentative sales amount to predict the sales amount of the new product. In this way, the sales amount prediction unit 31 can also predict the sales amount of the new product.

[0115] Furthermore, the actual sales figures for the new store are not stored in the actual sales figure DB 41. In such cases, the sales figure prediction unit 31 may generate tentative actual sales figures for the new store using the actual sales figures for existing stores of a similar size to the new store or existing stores with similar location conditions to the new store, and use the tentative actual sales figures to predict the sales figures for each product in the new store. This allows the sales figure prediction unit 31 to predict the sales figures for each product in the new store.

[0116] (Other Modifications) In the above-described embodiments, the safety factor is defined by the following formulas (1A) to (1C): Safety factor for day X = (actual inventory quantity at the start of business on day X) / (actual sales quantity for day X) (1A) Safety factor for day X = (actual inventory quantity at the start of business on day X) / (forecasted sales quantity for day X) (1B) Safety factor for day X = (forecasted inventory quantity at the start of business on day X) / (forecasted sales quantity for day X) (1C). However, these formulas (1A) to (1C) are simplified formulas. More specifically, the safety factor is a value indicating the ratio of the amount of inventory secured in advance before the product is received to the sales quantity that should be covered by the product corresponding to the current order. Therefore, the safety factor is more specifically defined by the following formula (2). Safety factor = (predicted inventory quantity "before the current order is received") / (predicted sales quantity from "after the current order is received" to "before the next order is received") (2) This formula (2) can be simplified based on several assumptions to obtain the above-mentioned formulas (1A) to (1C). For example, in formula (2), if we assume that the product ordered on day (X-α) is scheduled to be received on day X, and that the product ordered after day (X-α) will arrive on day (X+1), the numerator of formula (2) is the "inventory quantity at the start of work on day X," and the denominator is the predicted sales quantity from day X to before receiving on day (X+1) = the predicted sales quantity on day X. As a result, the safety factor on day X is as shown in formula (1C).

[0117] (Embodiment 3) In embodiment 3, a modified example of the sales number prediction unit 31 shown in embodiment 1 or 2 will be described. Note that the sales number prediction unit 31 may be a function or device included in the order support system 10. Alternatively, the sales number prediction unit 31 may be a prediction device or prediction system independent of the order support system 10, and may be called by the order support system 10 for use.

[0118] <Configuration of Sales Number Prediction Unit> FIG. 8 is a block diagram showing an example configuration of the sales number prediction unit 31 according to the third embodiment.

[0119] The sales volume prediction unit 31 includes a prediction model 200 , a first prediction unit 201 , a specific information management unit 202 , a reference value setting unit 203 , a reliability determination unit 204 , a second prediction unit 205 , and a specific information DB 206 .

[0120] The forecasting model 200 is a model for predicting the sales volume (demand volume) of a product for a certain future period (hereinafter referred to as the forecasting period). The forecasting model 200 may be a model trained based on the actual sales volume for each past day of the week and / or date and time stored in the actual sales volume DB 41. The forecasting model 200 may be, for example, a moving average model, an exponential smoothing model, an ARIMA model, a multiple regression model, or a deep learning model. Note that, although an example using a model realized by machine learning is described here, sales volume prediction may also be realized by a method that does not use a machine learning model. For example, sales volume may be predicted by inputting actual sales volume into a predetermined function. A simple example would be to use the average value of actual sales volume as the predicted sales volume, but techniques using more complex functions are also known.

[0121] The first prediction unit 201 predicts the sales volume of a product during a prediction period using the prediction model 200. Hereinafter, the sales volume predicted by the first prediction unit 201 will be referred to as the first predicted sales volume. For example, the first prediction unit 201 inputs the prediction period into the prediction model 200 and obtains the first predicted sales volume from the prediction model 200.

[0122] The specific information DB 206 manages the specific information 210 (see FIG. 9) as a DB.

[0123] 9 is a diagram showing an example of the special information 210 according to embodiment 3. Next, the special information 210 will be described with reference to FIG.

[0124] The specific information 210 is information indicating a specific situation regarding fluctuations in sales volume that is difficult to predict using the prediction model 200. The specific information 210 has items such as a period 211, a classification 212, a cause 213, and a response method 214.

[0125] The period 211 indicates a period during which a particular situation related to fluctuations in sales volume may occur. Note that the period 211 may be one day, multiple consecutive days, or a time period.

[0126] The category 212 indicates a category of a specific situation. Examples of the category 212 include "an event expected to attract more customers," "an event expected to attract less customers," "a holiday or public holiday that differs from the calendar," and "a change in customer demographics."

[0127] Cause 213 indicates the cause of the occurrence of the unique situation indicated by category 212 during period 211. Examples of cause 213 for category 212 "event expected to attract more customers" include "discount event at store" and "event held at surrounding facility." Examples of cause 213 for category 212 "event expected to attract less customers" include "construction at the nearest station," "construction on surrounding roads," and "construction in the tenant building where the store is located." Examples of cause 213 for category 212 "holiday or public holiday that differs from the calendar" include "newly established public holiday," "abolition of an existing public holiday," and "regionally specific public holiday." Examples of cause 213 for category 212 "change in customer demographics" include "summer vacation at a nearby school," "long vacation in a certain foreign country," and "holiday at a nearby large company."

[0128] The response method 214 indicates a response method to the first predicted sales volume depending on the cause 213. The response method 214 may be set differently depending on the reliability of the first predicted sales volume, which will be described later. The response method 214 may define the processing of the second prediction unit 205, which will be described later. The response method 214 will be described in detail later.

[0129] Returning to the description of FIG.

[0130] The specific information management unit 202 manages the specific information 210. The specific information management unit 202 may generate the specific information 210 and register it in the specific information DB 206. The specific information management unit 202 may change or delete the specific information 210 registered in the specific information DB 206. For example, the specific information management unit 202 may display, via the input display control unit 34, candidates for the classification 212, cause 213, and response method 214 that can be set in the above-mentioned specific information 210, and the user may be able to easily generate or change the specific information 210 by selecting the displayed candidate.

[0131] The reference value setting unit 203 sets a reference value for the prediction period based on the actual sales volume DB 41 and the specific information DB 206. For example, the reference value setting unit 203 acquires specific information 210 corresponding to the prediction period from the specific information DB 206 and sets a reference value based on the cause 213 of the specific information 210. For example, if the cause 213 of the specific information 210 is a "discount event," the reference value setting unit 203 sets the average value of actual sales volumes for similar "discount events" held in the past as the reference value. For example, if the cause 213 of the specific information 210 is a "newly established holiday," the reference value setting unit 203 sets the average value of actual sales volumes for past holidays as the reference value. For example, if the cause 213 of the specific information 210 is a "neighboring school's summer vacation," the reference value setting unit 203 sets the actual sales volume for past summer vacation periods as the reference value. In this case, the reference value may be a change in each date during past summer vacation periods. This allows the reference value setting unit 203 to calculate a predicted sales volume (i.e., a reference value) that is unlikely to deviate significantly during the prediction period, using a method different from that used by the prediction model 200. In other words, the reference value does not necessarily have to be a single value, but may be a set of multiple values.

[0132] The reliability determination unit 204 determines the reliability of the first predicted sales volume based on a comparison between the reference value set by the reference value setting unit 203 and the first predicted sales volume predicted by the first prediction unit 201 .

[0133] For example, if the difference value D between the reference value and the first predicted sales number is less than a predetermined threshold value Th1 (D<Th1), the reliability determination unit 204 may assign the highest reliability of "1" to the first predicted sales number; if the difference value D is equal to or greater than the threshold value Th1 and less than a threshold value Th2 (Th1≦D<Th2), the reliability determination unit 204 may assign the second highest reliability of "2" to the first predicted sales number; and if the difference value D is equal to or greater than the threshold value Th2 (Th2≦D), the reliability determination unit 204 may assign the lowest reliability of "3." In other words, the reliability determination unit 204 may determine that the smaller the difference value D, the higher the reliability of the first predicted sales number.

[0134] Alternatively, the reliability determination unit 204 may set the reliability of the first forecasted sales quantity to the highest level of "1" when the correlation value R between the reference value and the first forecasted sales quantity during the forecast period is equal to or greater than a predetermined threshold value Th3 (R≧Th3), set the reliability of the first forecasted sales quantity to the second highest level of "2" when the correlation value R is less than the threshold value Th3 and equal to or greater than a threshold value Th4 (Th3>R≧Th4), and set the reliability to the lowest level of "3" when the correlation value R is less than the threshold value Th4 (Th4>R). In other words, the reliability determination unit 204 may determine that the reliability of the first forecasted sales quantity is higher when the correlation value R is larger.

[0135] The reliability ranking is not limited to three levels as in this embodiment, but may be two levels or four or more levels.

[0136] The second prediction unit 205 determines the second predicted sales number based on the specific information 210 used to determine the reference value, the reliability determined by the reliability determination unit 204, and the first predicted sales number predicted by the first prediction unit 201. Here, the second predicted sales number is a value output as the predicted sales number of the product, and as will be described later, it may be the same as the first predicted sales number or may be a different value.

[0137] For example, if the first row of the specific information 210 shown in FIG. 9 is used to set the reference value, the second prediction unit 205 performs the following processing based on the response method 214 included in the specific information 210. If the reliability is "1," the second prediction unit 205 sets the first predicted sales number as the second predicted sales number. If the reliability is "2," the second prediction unit 205 corrects the first predicted sales number to approach the reference value set by the reference value setting unit 203, and sets the second predicted sales number. For example, if the first predicted sales number is smaller than the reference value, the second prediction unit 205 adds half of the difference α between the first predicted sales number and the reference value to the first predicted sales number, and sets the second predicted sales number. For example, if the first predicted sales number is larger than the reference value, the second prediction unit 205 subtracts half of the difference α between the first predicted sales number and the reference value from the first predicted sales number, and sets the second predicted sales number. This makes it possible to calculate a second predicted sales number that is suited to specific circumstances while making use of the prediction (first predicted sales number) made by the prediction model 200. When the reliability is "3," the second prediction unit 205 sets the reference value set by the reference value setting unit 203 as the second predicted sales number. When the reliability is "3," there is a possibility that the first predicted sales number is fundamentally incorrect, and therefore, whether the first predicted sales number is used as is as the second predicted sales number or the first predicted sales number is corrected and used as the second predicted sales number, there is a high risk that the prediction will be significantly off during the prediction period.

[0138] For example, if the fourth line of the specific information 210 shown in FIG. 9 is used to set the reference value, the second prediction unit 205 performs the following processing based on the response method 214 included in the specific information 210: If the reliability is "1," the second prediction unit 205 sets the first predicted sales number as the second predicted sales number. If the reliability is "2," the second prediction unit 205 corrects the first predicted sales number for a children's product so that it approaches the reference value set by the reference value setting unit 203, and sets the corrected number as the second predicted sales number. This is because the cause 213 is "school summer vacation," and it is estimated that the reference value calculated based on actual sales numbers during past summer vacations is more reliable for predicted sales numbers of children's products. This makes it possible to calculate a second predicted sales number tailored to the specific situation while utilizing the prediction (first predicted sales number) made by the prediction model 200. If the reliability is "3," the second prediction unit 205 requests the user to input predicted sales numbers, and sets the sales number input by the user as the second predicted sales number. When the reliability level is "3," there is a possibility that the first forecast sales number is fundamentally incorrect, and therefore, whether the first forecast sales number is used as is for the second forecast sales number, or the first forecast sales number is revised and used as the second forecast sales number, there is a high risk that the forecast will be significantly off during the forecast period.

[0139] The second prediction unit 205 registers the determined second predicted sales number in the predicted sales number DB 42 as the "predicted sales number" described in the first or second embodiment. This makes it possible to prevent the predicted sales number from being significantly off the target during a period in which unique circumstances related to fluctuations in sales number occur, which are difficult to predict using the prediction model 200. In other words, according to this embodiment, sales number can be stably predicted even when various changes in circumstances occur.

[0140] In the above description, the reference value is calculated by the reference value setting unit 203 using the specific information 210. However, the reference value itself may be recorded as the specific information 210. The reference value does not have to be a value comparable to the first predicted sales number. For example, the period during which a specific event is held may be recorded as the reference value, and if today's date is included in the reference value, the first predicted sales number may be revised. In this case, information on the judgment conditions indicating how the reference value is to be used for judgment may be recorded as the specific information. Specific examples will be described below with reference to the drawings.

[0141] FIG. 10 is a diagram showing an example of special information including information on the determination condition according to the third embodiment.

[0142] As shown in FIG. 10, the specific information 210 including the information on the determination condition has the following items: number, comparison target, reference value, determination condition, response method, and remarks.

[0143] For example, the row numbered "1" in the table shown in Figure 10 indicates that if today's date and time (comparison target) is within the range (determination condition) of May 1, 2023 to May 4, 2023 (reference value), the second predicted sales number will be doubled (response method). The reason for this is that "an event will be held at the facility and an increase in customers is expected" (note).

[0144] For example, the row numbered "2" in the table shown in Figure 10 indicates that if today's date and time (comparison target) matches August 1st (reference value) (judgment condition), the first predicted sales number will be replaced with the actual sales number for the holiday (response method) to become the second predicted sales number. The reason for this is that "this day is a newly established holiday" (note).

[0145] For example, the row numbered "4" in the table shown in FIG. 10 indicates that when today's date and time (comparison target) is within the range (judgment condition) of January 21, 2023 to January 29, 2023 (reference value), for menu items whose menu type (comparison target) is within the range (judgment condition) of Menu Group A (reference value), the first predicted sales number is replaced with the actual sales number for the holiday (response method), and the reason for this is that "only a specific customer group is on holiday during this period, and an increase in orders for menu items suitable for this customer group is predicted" (note). An example of a specific customer group is foreign children from a country with a nearby school.

[0146] <Display and Modification of Information> The input display control unit 34 may display the reference value set by the reference value setting unit 203 on the terminal 12. The user may modify the displayed reference value by operating the terminal 12. The input display control unit 34 may reflect the reference value modified by the user in the reference value setting unit 203. This allows the user to manually modify the reference value if they think the reference value is inappropriate.

[0147] The input display control unit 34 may display the reliability determined by the reliability determination unit 204 on the terminal 12. The user may modify the displayed reliability by operating the terminal 12. The input display control unit 34 may reflect the reliability modified by the user in the reliability determination unit 204. This allows the user to manually modify the reliability if they think the reliability is inappropriate.

[0148] The input display control unit 34 may display on the terminal 12 the response method 214 selected by the second prediction unit 205 according to the reliability. The user may change the displayed response method 214 to another response method 214 by operating the terminal 12. The input display control unit 34 may reflect the response method 214 changed by the user in the second prediction unit 205. In this case, the second prediction unit 205 applies the response method 214 changed by the user to the first predicted sales number and determines the second predicted sales number. This allows the user to change the response method 214 to another appropriate response method 214 if they consider the response method 214 to be inappropriate.

[0149] FIG. 11 is a diagram showing a display example of the performance forecast comparison screen according to the third embodiment.

[0150] The input display control unit 34 may display a performance forecast comparison screen 300 including a graph 301 comparing the first forecast sales volume and the actual sales volume for a predetermined period in the past (e.g., one week), as shown in Fig. 11. The input display control unit 34 may also include a graph 302 showing changes in precipitation and solar radiation over the same predetermined period, and a calendar 303 showing dates on which events were held during the same predetermined period, on the performance forecast comparison screen 300. By viewing these graphs, the user can determine whether the prediction by the first prediction unit 201 is appropriate.

[0151] Furthermore, the input display control unit 34 may include response method options 304 that can be taken when the user determines that the prediction is invalid on the performance prediction comparison screen 300. This allows the user to select from the response method options 304 and easily change the response method 214 to another appropriate response method 214 when the user determines that the prediction by the first prediction unit 201 is invalid.

[0152] Furthermore, the input / display control unit 34 may output an alert when the error rate between the first predicted sales volume and the actual sales volume for a predetermined period is greater than a predetermined threshold value. This allows the user to be quickly notified when the prediction by the first prediction unit 201 is highly likely to be invalid.

[0153] The input display control unit 34 may also display the following information based on the statistical values ​​of the error between the first predicted sales volume and the actual sales volume for a predetermined period: - If the variance of the error is smaller than a certain threshold and there are many samples in which the first predicted sales volume is smaller than the actual sales volume, the input display control unit 34 displays information such as "The predicted value tends to be smaller than the actual value." - If the variance of the error is smaller than a certain threshold and there are many samples in which the first predicted sales volume is larger than the actual sales volume, the input display control unit 34 displays information such as "The predicted value tends to be larger than the actual value." - If the variance of the error is larger than a certain threshold, the input display control unit 34 displays information such as "There is a large deviation between the actual value and the predicted value on X month X day" for the sample with the largest error.

[0154] FIG. 12 is a diagram showing a display example of a predicted value change screen according to the third embodiment.

[0155] The input display control unit 34 may display a forecast value change screen 400, as shown in FIG. 12 , including a graph 401 showing past forecasted sales and actual sales, and a graph 402 showing second forecasted sales for a predetermined future period. The input display control unit 34 may provide a UI that accepts changes to the graph 402 of second forecasted sales. If the user deems the displayed graph 402 of second forecasted sales inappropriate, the user can change the graph 402 of second forecasted sales while referring to the graph 401 showing past forecasted sales and actual sales. This allows the user to easily change the second forecasted sales. The input display control unit 34 may also generate a new response method 214 based on changes to the second forecasted sales. For example, the input display control unit 34 may generate a new response method 214, such as multiplying the first forecasted sales by α under the changed day's conditions (e.g., day of the week, weather, etc.), and add the new response method 214 to the specific information 210.

[0156] <Review of specific information> The specific information management unit 202 may review the contents of the specific information 210 used to determine the second predicted sales volume based on the first predicted sales volume and second predicted sales volume predicted for a certain target period and the actual sales volume for that target period.

[0157] For example, the specific information management unit 202 may compare the error E1 between the first predicted sales number and the actual sales number for the target period with the error E2 between the second predicted sales number and the actual sales number for the target period, and if the error E2 is greater than the error E1, perform at least one of the following (C1) and (C2). If the error E2 is greater than the error E1, it is presumed that the original first predicted sales number has higher prediction accuracy than the second predicted sales number determined using the specific information 210, and the content of the specific information 210 is not very appropriate. (C1) The specific information management unit 202 deletes the specific information 210 used to determine the second predicted sales number from the specific information DB 206. (C2) The specific information management unit 202 notifies the user to review the content of the specific information 210 used to determine the second predicted sales number. Upon receiving this notification, the user may, for example, review the content of the response method 214 for the specific information 210 or delete the specific information 210.

[0158] FIG. 13 is a diagram showing an example of a review screen for specific information including information on the determination conditions according to the third embodiment.

[0159] For example, if the error E2 between the second forecast sales number and the actual sales number for the target period calculated by applying row number "1" in the table shown in Figure 10 is greater than the error E1 between the first forecast sales number and the actual sales number for the target period (error E2 > error E1), the specific information management unit 202 may display a specific information review screen 500 such as that shown in Figure 13 via the input display control unit 34.

[0160] 13, the specific information review screen 500 may display the applied specific information 501, a button 502 for instructing to use the specific information 501, a button 503 for instructing not to use the specific information 501 this time, and a button 504 for instructing not to use the specific information 501 this time and to delete the specific information 501 from the specific information DB 206. By selecting one of these buttons 502, 503, and 504, the user can easily review the specific information 501 that does not seem to have very good accuracy.

[0161] <Revision of Reference Value> The reference value setting unit 203 may revise (review) the current reference value based on the error E3 between the second predicted sales volume calculated for a specified period in the most recent past (e.g., the most recent week) and the actual sales volume for the same specified period. For example, the reference value setting unit 203 revise the current reference value if the error E3 is equal to or greater than a specified threshold. This is because the accuracy of the original reference value may be insufficient. For example, the reference value setting unit 203 may revise the average actual sales volume (i.e., the original reference value) of similar "discount events" held in the past so that the average value approaches the actual sales volume for the specified period in the most recent past (e.g., the most recent week), and use this as the current reference value. This prevents the continued use of an inaccurate reference value. Additionally, it is possible to prevent the setting of a reference value that is significantly different from the actual sales volume of the most recent period, thereby improving the reliability of the reference value.

[0162] The method of correcting the reference value is not limited to the method of bringing the reference value closer to the actual sales volume for a predetermined period in the most recent past. For example, the reference value setting unit 203 may correct (review) the current reference value based on the actual sales volume for a predetermined period in the more distant past (e.g., one month) rather than the most recent period, or the actual sales volume for the same season in the past.

[0163] Alternatively, the reference value setting unit 203 may be realized by machine learning. In this case, the reference value setting unit 203 learns information indicating the situation in which the error E3 occurred and a correction value that cancels out the error E3. When the reference value setting unit 203 predicts that a situation similar to the situation in which the error E3 occurred will occur during the prediction period, the reference value setting unit 203 sets the reference value using the learned correction value that cancels out the error E3. By repeating this process, the reference value setting unit 203 becomes able to calculate a reference value with a smaller error as the learning progresses.

[0164] <Modification> The second prediction unit 205 may correct the first predicted sales number obtained from the first prediction unit 201 as needed based on the trend (e.g., upward or downward trend) of sales performance over the most recent period (e.g., several weeks), and use this as the second predicted sales number. This will be described in detail below.

[0165] 14 is a diagram illustrating a method for correcting the first forecasted sales number based on trend information according to embodiment 3. In the graph shown in Fig. 14, the horizontal axis represents the date, and the vertical axis represents the sales number.

[0166] The specific information 210 may set the following for a certain period: Classification 212: Reflection of the trend in the determined performance Cause 213: An upward or downward trend occurs in the sales performance Response method 214 Reliability 1: Use the first predicted sales number as is Reliability 2: Replace the first predicted sales number with a corrected value calculated based on the trend coefficient and the first predicted sales number.

[0167] The reference value setting unit 203 calculates the slope of an approximate line (regression line) of the sales volume for a predetermined period (for example, the past month) as a reference value (i.e., trend coefficient) representing the sales trend.

[0168] The reliability determination unit 204 determines the reliability depending on the magnitude of the reference value. For example, if the absolute value of the reference value is less than a predetermined threshold, the reliability determination unit 204 determines the reliability to be "1," and if the absolute value of the reference value is equal to or greater than the predetermined threshold, the reliability determination unit 204 determines the reliability to be "2."

[0169] If the reliability determined by the reliability determination unit 204 is "2", the second prediction unit 205 replaces the first predicted sales number with a corrected value calculated based on the trend coefficient and the first predicted sales number, and sets this as the second predicted sales number.

[0170] Next, an example of the above-described processing will be described with reference to FIG. 14 . For example, the order support system 10 performs the following processing: (S101) The reference value setting unit 203 calculates the average value A of the sales figures for the most recent past month (March 1st to March 28th). (S102) The reference value setting unit 203 calculates the slope B of the approximation line (regression line) of the sales figures for the most recent past month (March 1st to March 28th) as trend information obtained from the sales figures for that period. (S103) The reliability determination unit 204 compares the amount of change C (= slope B × 7 days) for one week based on the trend information with the average value A of the sales figures and selects a response method. For example, if the amount of change C for one week is less than (10% of average value A), the reliability determination unit 204 determines the reliability to be "1." If "the amount of change C for one week is equal to or greater than (10% of the average value A)," the reliability determination unit 204 determines the reliability to be "2." (S104) The second prediction unit 205 refers to the specific information 210 described above, and if the reliability is "1," the second prediction unit 205 sets the first predicted sales number as the second predicted sales number for the prediction period (March 29th to April 5th) as is. If the reliability is "2," the second prediction unit 205 calculates the second predicted sales number for the prediction period (March 29th to April 5th) using the following formula: second predicted sales number = first predicted sales number + one-week change C Note that (first predicted sales number + one-week change C) is an example of the correction value described above.

[0171] As a result, the predicted sales figures for March 29th to April 5th in the graph shown in the upper part (a) of Figure 14 are corrected to match the predicted sales figures for March 29th to April 5th in the graph shown in the lower part (b) of Figure 14.

[0172] Using only the prediction model 200, there is a risk that the sales forecast will be significantly off if a major change in trend occurs. However, by calculating or correcting the sales forecast taking into account recent trend changes (e.g., an upward or downward trend) in this way, this risk can be reduced. Note that various regression analysis methods other than those for finding a regression line are known for analyzing a set of data to determine a function indicating a trend. Therefore, in the third embodiment, instead of using a regression line, it is also possible to estimate a function indicating a trend using these regression analysis methods, and use the parameters of that function as trend coefficients.

[0173] (Hardware Configuration) The functional blocks of the order support system 10 described above can be realized by a computer program.

[0174] FIG. 15 is a diagram showing the hardware configuration of a computer that implements the functional blocks of the order support system 10 according to the present disclosure using a computer program.

[0175] The computer 1000 includes a processor 1001, a memory 1002, a storage 1003, an input device 1004, an output device 1005, a communication device 1006, a graphics processing unit (GPU) 1007, a reading device 1008, and a bus 1009. These devices are connected to the bus 1009 and can send and receive data bidirectionally via the bus 1009.

[0176] The processor 1001 is a device that executes a computer program stored in the memory 1002 and realizes the above-mentioned functional blocks. Examples of the processor 1001 include a central processing unit (CPU), a micro processing unit (MPU), a controller, a large scale integration (LSI), an application specific integrated circuit (ASIC), a programmable logic device (PLD), and a field-programmable gate array (FPGA).

[0177] The memory 1002 is a device that stores computer programs and data used by the computer 1000. The memory 1002 may include a read-only memory (ROM) and a random access memory (RAM).

[0178] The storage 1003 is configured as a non-volatile storage medium, and is a device that stores computer programs and data used by the computer 1000. Examples of the storage 1003 include a hard disk drive (HDD) and a solid state drive (SSD).

[0179] The input device 1004 is a device that accepts data to be input to the processor 1001. Examples of the input device 1004 include a keyboard, a mouse, a touchpad, and a microphone.

[0180] The output device 1005 is a device that outputs data generated by the processor 1001. Examples of the output device 1005 include a display and a speaker.

[0181] The communication device 1006 is a device that transmits and receives data to and from other devices (for example, the weather information server 5 or the terminal 12) via a communication network. The communication device 1006 may support either wired communication or wireless communication. An example of wired communication is Ethernet (registered trademark). Examples of wireless communication are Wi-Fi (registered trademark), Bluetooth, LTE, 4G, 5G, etc.

[0182] The GPU 1007 is a device that performs high-speed image rendering processing. The GPU 1007 may be used for artificial intelligence (AI) processing (e.g., deep learning).

[0183] The reading device 1008 is a device that reads data from a recording medium such as a Digital Versatile Disk Read Only Memory (DVD-ROM) or a Universal Serial Bus (USB) memory.

[0184] (Summary of Third Embodiment) The third embodiment discloses the following techniques.

[0185] <Technology 1> A prediction device (e.g., sales volume prediction unit 31) according to this embodiment includes a first prediction unit (201) that predicts a first predicted sales volume for a future prediction period based on the actual past sales volume of a product, a reference value setting unit (203) that sets a reference value based on specific information (210) that indicates specific circumstances related to fluctuations in sales volume during the prediction period, and a second prediction unit (205) that determines, based on the reference value, whether a second predicted sales volume to be output as the predicted sales volume of the product should be the same as the first predicted sales volume. This makes it possible to obtain a predicted sales volume appropriate to the specific circumstances related to fluctuations in sales volume.

[0186] <Technology 2> The prediction device described in Technology 1 further includes a reliability determination unit (204) that determines the reliability of the first predicted sales number based on a comparison between a reference value and the first predicted sales number, and a second prediction unit (205) that determines the second predicted sales number based on the reliability and the first predicted sales number. This makes it possible to obtain an appropriate predicted sales number according to the reliability of the first predicted sales number.

[0187] <Technology 3> In the prediction device according to Technology 1 or 2, the specific information (210) includes information indicating fluctuations in the number of customers or customer demographics during the prediction period. This makes it possible to obtain appropriate predicted sales figures that take into account fluctuations in the number of customers or customer demographics.

[0188] <Technology 4> In the prediction device according to any one of techniques 1 to 3, the specific information (210) includes at least one of information indicating events that are expected to attract more customers, events that are expected to attract less customers, and changes in customer demographics during the prediction period. This makes it possible to obtain appropriate predicted sales figures that take into account the holding of events or changes in customer demographics.

[0189] <Technology 5> In the prediction device according to any one of Techniques 2 to 4, the reliability determination unit (204) determines the reliability based on a difference value or a correlation value between a reference value and the first predicted sales number. This makes it possible to obtain an appropriate predicted sales number according to the reliability of the first predicted sales number.

[0190] <Technology 6> In the prediction device according to any one of claims 2 to 5, the second prediction unit (205) sets the first predicted sales number to the second predicted sales number when the reliability is equal to or greater than a predetermined threshold, and modifies the first predicted sales number to the second predicted sales number when the reliability is less than the threshold. As a result, when the reliability of the first predicted sales number is relatively low, the second predicted sales number is used, thereby preventing the predicted sales number from being significantly off.

[0191] <Technology 7> In the prediction device described in Technology 6, the second prediction unit (205) determines a method for correcting the first predicted sales number depending on the reliability. This makes it possible to obtain an appropriate predicted sales number depending on the reliability of the first predicted sales number.

[0192] <Technology 8> In the prediction device described in Technology 7, the second prediction unit (205) determines a method for correcting the first predicted sales volume based on the specific information (210) used to determine the reference value. This makes it possible to obtain an appropriate predicted sales volume according to the reliability of the first predicted sales volume.

[0193] <Technology 9> The prediction device according to any one of technologies 2 to 8 further includes an input display control unit (34) that displays at least one of the first predicted sales number, the reference value, the reliability, and the second predicted sales number, and accepts input or modification of at least one of the reference value and the reliability. This allows a user to visually check at least one of the first predicted sales number, the reference value, the reliability, and the second predicted sales number, and to input or modify at least one of the reference value and the reliability.

[0194] <Technology 10> The prediction device according to any one of technologies 1 to 9 further includes a unique information management unit (202) that, when an error between the second predicted sales volume and the actual sales volume during the prediction period is equal to or greater than a predetermined threshold, modifies or deletes the unique information used to determine the second predicted sales volume. This makes it possible to modify or delete, as necessary, the unique information that may have a relatively large error.

[0195] <Technology 11> In the prediction device described in Technology 10, the predetermined threshold is a value based on the error between the first predicted sales volume and the actual sales volume during the prediction period. This makes it possible to correct or delete, as necessary, specific information that may have a relatively large error.

[0196] <Technology 12> In the prediction device according to any one of techniques 1 to 11, the reference value setting unit (203) corrects the reference value when the difference between the second predicted sales volume and the actual sales volume for a predetermined period in the past is equal to or greater than a predetermined threshold. This corrects the reference value when the difference is relatively large, making it possible to obtain an appropriate predicted sales volume.

[0197] Technique 13: A prediction method according to this embodiment predicts a first predicted sales number for a future prediction period based on the actual past sales number of a product, sets a reference value based on specific information indicating specific circumstances related to fluctuations in sales number during the prediction period, and determines whether or not to make a second predicted sales number output as the predicted sales number of the product the same as the first predicted sales number based on the reference value. This makes it possible to obtain a predicted sales number appropriate to the specific circumstances related to fluctuations in sales number.

[0198] <Technology 14> An order support device (e.g., order support system 1) according to this embodiment supports product ordering using the prediction method described in Technology 13. This makes it possible to support product ordering based on an appropriate predicted sales volume according to a specific situation related to fluctuations in sales volume.

[0199] Although the embodiments have been described above with reference to the accompanying drawings, the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications, alterations, substitutions, additions, deletions, and equivalents within the scope of the claims, and it is understood that these also fall within the technical scope of the present disclosure. Furthermore, the components in the above-described embodiments may be combined in any manner without departing from the spirit of the invention.

[0200] This application is based on a Japanese patent application (Patent Application No. 2023-121719) filed on July 26, 2023, the contents of which are incorporated by reference into this application.

[0201] The technology of the present disclosure is useful for forecasting demand for products.

[0202] 1 Orderer 3 Product delivery center 4 Ordering device 5 Weather information server 6 Calendar information server 10 Ordering support system 12 Terminal 31 Sales quantity prediction unit 32 Recommended order quantity calculation unit 33 Safety factor calculation unit 34 Input display control unit 41 Actual sales quantity DB 42 Forecasted sales quantity DB 43 Inventory quantity DB 44 Order quantity DB 47 Expiry date master 48 Store information DB 100 Ordering support screen 101 Product list area 102 Graph area 103 Information area 104 Sales quantity graph 105 Inventory quantity graph 120 Product list screen 121 Calendar area 122 Graph area 123 Bar graph 140 Ordering support screen 141 Graph area 142 Manual order tab 143 Automatic order tab 144 Fixed order tab 145 Forecasted sales quantity graph 146 Forecasted inventory quantity graph 147 Waste loss graph 148 Opportunity loss graph 151 Recommended order quantity area 152 Order quantity input area 153 Stock quantity at end of business area 154 Waste loss area 155 Opportunity loss area 156 Total loss area 160 Order support screen 161 Order quantity area 180 Order support screen 181 Lower limit input area 182 Fixed order input area 200 Forecast model 201 First forecasting section 202 Specific information management section 203 Reference value setting section 204 Reliability determination section 205 Second forecasting section 206 Specific information DB 210 Specific information 211 Period 212 Classification 213 Cause 214 Countermeasure method 300 Actual performance forecast comparison screen 301, 302 Graph 303 Calendar 400 Forecast value change screen 401, 402 Graph 500 Specific information review screen 501 Specific information 502, 503, 504 Button 1000 Computer 1001 Processor 1002 Memory 1003 Storage 1004 Input device 1005 Output device 1006 Communication device 1007 GPU 1008 Reading device 1009 Bus

Claims

1. A prediction device comprising: a first prediction unit that predicts a first predicted sales volume for a future prediction period based on the past actual sales volume of a product; a reference value setting unit that sets a reference value based on specific information that indicates specific circumstances related to fluctuations in sales volume during the prediction period; and a second prediction unit that determines, based on the reference value, whether a second predicted sales volume to be output as the predicted sales volume of the product should be the same as the first predicted sales volume.

2. The prediction device described in claim 1, further comprising a reliability determination unit that determines the reliability of the first predicted sales number based on a comparison between the reference value and the first predicted sales number, and the second prediction unit determines the second predicted sales number based on the reliability and the first predicted sales number.

3. The prediction device according to claim 2, wherein the specific information includes information indicating fluctuations in the number of customers or customer demographics during the prediction period.

4. The prediction device according to claim 3, wherein the specific information includes at least one of information indicating events that are expected to attract more customers, events that are expected to attract less customers, and changes in customer demographics during the prediction period.

5. The prediction device according to claim 2, wherein the reliability determination unit determines the reliability based on a difference value or a correlation value between the reference value and the first predicted sales volume.

6. The prediction device described in claim 3, wherein the second prediction unit sets the first predicted sales number to the second predicted sales number when the reliability is equal to or greater than a predetermined threshold, and modifies the first predicted sales number to set it as the second predicted sales number when the reliability is less than the threshold.

7. The prediction device according to claim 6, wherein the second prediction unit determines a method of correcting the first predicted sales volume depending on the reliability.

8. The prediction device according to claim 7, wherein the second prediction unit determines a method of correcting the first predicted sales volume based on the specific information used in determining the reference value.

9. The prediction device described in claim 3, further comprising an input display control unit that displays at least one of the first predicted sales number, the reference value, the reliability, and the second predicted sales number, and accepts input or modification of at least one of the reference value and the reliability.

10. The prediction device described in claim 3, further comprising a unique information management unit that modifies or deletes the unique information used to determine the second forecast sales volume when an error between the second forecast sales volume and actual sales volume during the prediction period is equal to or greater than a predetermined threshold.

11. The prediction device according to claim 10, wherein the predetermined threshold is a value based on an error between the first predicted sales volume and actual sales volume during the prediction period.

12. The prediction device according to claim 3, wherein the reference value setting unit modifies the reference value when an error between the second predicted sales volume and actual sales volume during a specified past period is equal to or greater than a specified threshold value.

13. A prediction method comprising: predicting a first predicted sales volume for a future prediction period based on the actual past sales volume of a product; setting a reference value based on specific information indicating specific circumstances related to fluctuations in sales volume during the prediction period; and determining, based on the reference value, whether or not a second predicted sales volume output as the predicted sales volume of the product should be set to the same value as the first predicted sales volume.

14. An ordering support device that uses the prediction method according to claim 13 to support ordering of the product.

Images