Processing device, processing program, processing method, and processing system

Abstract

A processing device, a processing program, a processing method, and a processing system capable of more appropriately calculating the recommended number of products to be displayed are provided. [Solution] Demand forecast information relating to the demand for the product at any given time is generated using a demand forecast model generated based on demand information for products displayed in a store, and processing is performed to generate recommendation information indicating the recommended number of products to display at any given time based on the generated demand forecast information and display-related information related to the display of the products.

Description

[Technical field]

[0001] The present disclosure relates to a processing device, a processing program, a processing method, and a processing system capable of executing processing related to products displayed in a store. [Background technology]

[0002] Conventionally, a processing system that checks the current display status of products in a store and requests the display of new products has been known. For example, Patent Document 1 describes a "shipment management system consisting of an image recognition camera unit arranged in a store, a weight sensor mat unit, a producer's terminal device, and a management server device, the management server device including a reception unit that receives image data from the image recognition camera unit and weight data from the weight sensor mat unit, an image analysis unit that analyzes the image data and identifies products and display quantities, a storage unit that stores at least a shipping request threshold value, a production item weight, an approximation range, the image data, and the weight data, an acquisition unit that acquires the shipping request threshold value, the production item weight, the approximation range, the display quantity identified by the image analysis, and the like, and a processing ... The document describes a shipping management system having a display quantity calculation unit that divides the weight related to the weight data received by an attachment unit by the weight of the produced item, determines whether the divided value is within the approximation range, and if it is within the approximation range, updates the display quantity to the display quantity identified by an image analysis unit, and if it is outside the approximation range, updates the display quantity to the display quantity calculated by the division; a shipping determination unit that compares the display quantity calculated by the display quantity calculation unit with a shipping threshold value to determine whether or not to ship; and a shipping request unit that makes a shipping request based on the determination result of the shipping determination unit, and the producer's terminal device has a receiving unit that receives a notification related to the shipping request and a display unit that displays the notification. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-189989 Summary of the Invention [Problem to be solved by the invention]

[0004] Therefore, in consideration of the above-mentioned technology, the present disclosure aims to provide a processing device, a processing program, a processing method, and a processing system that can more appropriately calculate the recommended number of products to display in various embodiments. [Means for solving the problem]

[0005] According to one aspect of the present disclosure, there is provided "a processing device having at least one processor, wherein the at least one processor is configured to execute processing to generate demand forecast information related to the demand amount of the product at any time using a demand forecast model generated based on demand information of the product displayed in a store, and to generate recommendation information indicating the recommended number of the product to be displayed at the any time based on the generated demand forecast information and display-related information related to the display of the product."

[0006] According to one aspect of the present disclosure, there is provided a processing program that, when executed by at least one processor, causes the at least one processor to function as follows: generate demand forecast information related to the demand for a product at any given time using a demand forecast model generated based on demand information for the product displayed in a store, and generate recommendation information indicating the recommended number of the product to be displayed at the any given time based on the generated demand forecast information and display-related information related to the display of the product.

[0007] According to one aspect of the present disclosure, there is provided a "processing method executed by at least one processor, the processing method including: generating demand forecast information relating to a demand amount of a product at any time using a demand forecast model generated based on demand information of the product displayed in a store; and generating recommendation information indicating a recommended number of the product to be displayed at the any time based on the generated demand forecast information and display-related information related to the display of the product."

[0008] According to one aspect of the present disclosure, the processing system includes "the processing device and a terminal device that is arranged in a store where products are displayed so as to be able to communicate with the processing device, receives recommendation information from the processing device indicating a recommended number of products to be displayed, and is configured to output the received recommendation information." Effect of the Invention

[0009] According to the present disclosure, it is possible to provide a processing device, a processing program, a processing method, and a processing system that are capable of more appropriately calculating the recommended number of products to display.

[0010] It should be noted that the above effects are merely illustrative for the convenience of explanation and are not limiting. In addition to or instead of the above effects, any effect described in this disclosure or any effect obvious to a person skilled in the art may be achieved. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a block diagram showing a configuration of a processing system 1 according to an embodiment of the present disclosure. [Figure 2A] FIG. 2A is a block diagram showing a configuration of a processing device 100 according to an embodiment of the present disclosure. [Figure 2B] FIG. 2B is a block diagram showing a configuration of a terminal device 200 according to an embodiment of the present disclosure. [Figure 3A] FIG. 3A is a diagram conceptually illustrating a product management table stored in the processing device 100 according to an embodiment of the present disclosure. [Figure 3B] FIG. 3B is a diagram conceptually illustrating demand forecast information stored in the processing device 100 according to an embodiment of the present disclosure. [Figure 3C] FIG. 3C is a diagram conceptually illustrating a recommended information management table stored in the processing device 100 according to an embodiment of the present disclosure. [Figure 3D] FIG. 3D is a diagram conceptually showing a work information management table stored in the processing device 100 according to an embodiment of the present disclosure. [Figure 4] FIG. 4 is a diagram showing a process flow executed in the processing device 100 according to an embodiment of the present disclosure. [Diagram 5] FIG. 5 is a diagram showing a process flow executed in the processing device 100 according to an embodiment of the present disclosure. [Figure 6] FIG. 6 is a diagram conceptually illustrating a process in which recommendation information at time t' is generated by the processing device 100 according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Various embodiments of the present invention will be described below with reference to the accompanying drawings. Note that common components in the drawings are given the same reference numerals. Also, please note that components shown in one drawing may be omitted in another drawing for convenience of explanation. Furthermore, please note that the attached drawings are not necessarily drawn to scale.

[0013] The various systems, methods, and devices described in this disclosure should not be construed as being limited in any way. Indeed, the present disclosure is directed to all novel features and aspects of each of the various disclosed embodiments, combinations of these various embodiments with each other, and combinations of parts of these various embodiments with each other. The various systems, methods, and devices described in this disclosure are not limited to a specific aspect, a specific feature, or a combination of such a specific aspect and a specific feature, and the products and methods described in this disclosure do not require that one or more specific advantages exist or problems be solved. Moreover, various features or aspects of the various embodiments described in this disclosure, or parts of such features or aspects, may be used in combination with each other.

[0014] Although the operations of some of the various methods disclosed in this disclosure have been described in a particular order for convenience, it should be understood that description in such a manner encompasses rearranging the order of the operations, unless a particular order is required by specific text below. For example, operations described in a sequence may in some cases be rearranged or performed simultaneously. Moreover, for purposes of simplicity, the accompanying drawings do not show various ways in which the various items and methods described in this disclosure may be used in conjunction with other items and methods.

[0015] Any theories of operation, scientific principles, or other theoretical descriptions presented in this disclosure relating to the devices or methods of the present disclosure are provided for the purpose of better understanding and are not intended to limit the scope of the technology, and the devices and methods in the appended claims are not limited to devices and methods that operate in a manner described by such theories of operation.

[0016] Any of the various methods disclosed in this disclosure may be implemented using computer-executable instructions stored on one or more computer-readable media and executed in a computer. The one or more media may be a non-transitory computer-readable storage medium, such as at least one optical media disk, volatile memory components, or non-volatile memory components, where the volatile memory components include, for example, DRAM or SRAM, and the non-volatile memory components include, for example, hard drives and solid-state drives (SSD). The computer may include any computer available in the market, including, for example, smartphones and other mobile devices having hardware to perform computations.

[0017] Any such computer-executable instructions for implementing the techniques disclosed in this disclosure, along with any data generated and used during the implementation of the various embodiments disclosed in this disclosure, may be stored in one or more computer-readable media (e.g., non-transitory computer-readable storage media). Such computer-executable instructions may be, for example, part of a separate software application, or part of a software application accessed or downloaded via a web browser or other software application (such as a remote computing application). Such software may be executed, for example, on a single local computer (e.g., as a process executed on any suitable computer available on the market) or in a network environment (e.g., the Internet, a wide area network, a local area network, a client-server network (such as a cloud computing network), or other such network) using one or more networked computers.

[0018] For clarity, only certain selected aspects of various software-based implementations are described. Other details that are well known in the art are omitted. For example, the techniques disclosed in this disclosure are not limited to a specific computer language or program. For example, the techniques disclosed in this disclosure may be implemented by software written in C, C++, Java, or any other suitable programming language. Similarly, the techniques disclosed in this disclosure are not limited to a specific computer or a specific type of hardware. Specific details of suitable computers and hardware are well known and need not be described in detail in this disclosure.

[0019] Moreover, any of the various such software-based embodiments (e.g., including computer-executable instructions for causing a computer to perform any of the various methods disclosed in this disclosure) may be uploaded, downloaded, or accessed in a remote manner by any suitable communication means, including, for example, the Internet, the World Wide Web, an intranet, a software application, cable (including fiber optic cable), magnetic communication, electromagnetic communication (including RF, microwave, infrared), electronic communication, or other such communication means.

[0020] 1. Overview of Processing System 1 In a store, one or more products are displayed. The number of such products displayed decreases as the products are purchased by consumers who visit the store. However, in general, each product is set with a deadline for which it can be displayed in order to maintain freshness, such as a use-by date or a best-before date. Since it is difficult to predict when and how much a product will be purchased by a consumer, if the product is displayed without planning, products that have passed the deadline will not be purchased by consumers and will be removed from the display and discarded. Such discarding not only increases the environmental load but also generates economic losses. Therefore, maintaining an appropriate number of products displayed in a store is a very important factor in preventing an increase in the environmental load and economic losses.

[0021] The processing system 1 according to the present disclosure is used to generate recommendation information indicating the number of products recommended to be displayed at any given time. In other words, by generating the recommendation information, the processing system 1 makes it possible to maintain an appropriate number of products displayed in a store, and further makes it possible to prevent an increase in environmental load and economic loss.

[0022] Specifically, the processing system 1 generates demand forecast information related to the demand amount of the product at any time using a demand forecast model generated based on demand information of the product displayed in the store. Then, the processing system 1 generates recommendation information indicating the recommended number of products to be displayed at any time based on the generated demand forecast information and display related information related to the display of the products.

[0023] The processing system 1 also generates the recommendation information for each product. The processing system 1 also generates recommendation information for each set time at a predetermined interval, such as the time set next to the arbitrary time. Therefore, the processing system 1 can generate recommendation information indicating the recommended number of products to be displayed at each time for each product.

[0024] In the present disclosure, a store may be any store capable of displaying products. Examples of such stores include department stores (large department stores and other department stores), general merchandise stores (large general merchandise stores and medium-sized general merchandise stores), specialty supermarkets (clothing specialty supermarkets, food specialty supermarkets, home-related specialty supermarkets, etc.), convenience stores (all-day stores and others), drug stores, other supermarkets, specialty stores (clothing specialty stores, food specialty stores, home-related specialty stores, etc.), and central stores (clothing-focused stores, food-focused stores, home-related stores, etc.). Stores are not limited to the above, and may also include, as examples, restaurants, traditional Japanese restaurants, inns, hotels, etc., as long as they have space capable of displaying products. In the following, convenience stores will be mainly used as an example of stores, but stores are not limited to these.

[0025] In addition, in this disclosure, a "product" may be any product that can be displayed in a store. Textiles, clothing, and accessories such as kimonos, clothing fabrics, bedding, men's clothing, women's clothing, children's clothing, shoes, footwear (excluding shoes), bags, sacks, underwear, Western goods, accessories, other clothing, and accessories Food and beverage products, including vegetables, fruits, meat, eggs, poultry, fresh fish, alcohol, confectionery (manufactured), confectionery (non-manufactured), bread (manufactured), bread (non-manufactured), milk, beverages (excluding milk, including tea beverages), tea, cooked foods, rice, processed foods such as tofu and kamaboko, dried foods, and other food and beverage products. · New cars, used cars, car parts and accessories, motorcycles, bicycles and other cars and bicycles Electrical machinery and equipment (excluding second-hand items), electrical office machinery and equipment (excluding second-hand items), second-hand electrical products, other machinery and equipment, etc. Furniture, building materials, tatami mats, religious implements, metal objects, hardware, ceramics and glassware, other household items, over-the-counter medicines, medical drugs, cosmetics, agricultural machinery and equipment, seedlings and seeds, fertilizer and feed, fuel (gas stations), fuel (excluding gas stations), books and magazines (excluding second-hand books), used books, newspapers, paper and stationery, sporting goods, toys and entertainment goods, musical instruments, cameras and photographic materials, watches, glasses and optical equipment, tobacco and smoking accessories, flowers and plants, building materials, jewellery, pets and pet supplies, antiques, second-hand goods (excluding antiques), etc. Among these, the product is preferably a food or beverage, more preferably a food or beverage cooked by a staff member at a store. Note that, in the following, the product is described using a food or beverage cooked by a staff member at a store (for example, a so-called FF product such as fried chicken), but the product is not limited to this.

[0026] In the present disclosure, the "display-related information" may be any information related to the display of each product. For example, the display-related information may preferably include cooking count information indicating the number of times each product can be cooked at each time, cooking cost information indicating the cooking cost of the displayed products, the location of the store, and cooking time required to cook the products, and more preferably includes cooking count information and cooking cost information.

[0027] Fig. 1 is a block diagram showing a configuration of a processing system 1 according to an embodiment of the present disclosure. According to Fig. 1, the processing system 1 includes a processing device 100 and a terminal device 200, and each device is connected to each other so as to be able to communicate with each other via a wired or wireless network.

[0028] In the present disclosure, the processing device may be any device capable of performing the processing executed by the processing device 100. In other words, various devices such as an on-premise server device, a cloud server device, a smartphone, a tablet device, a laptop PC, and a desktop PC can be used as the processing device. Also, the terminal device 200 can function as the processing device. Furthermore, in the present disclosure, the storage and processing performed by the processing device may be distributed to other processing devices, a server device, a database device, and the like. In other words, the processing device is not limited to only those configured from a single housing, but also includes combinations of the various devices exemplified above.

[0029] In addition, in the present disclosure, the terminal device may be any device capable of performing the processing executed by the terminal device 200. The terminal device 200 may be a smartphone, a tablet device, a laptop PC, a desktop PC, a cash register device, a POS device, or other various devices. Furthermore, in the present disclosure, the processing performed by the terminal device may be distributed to other terminal devices, processing devices, server devices, database devices, or the like. In other words, the terminal device is not limited to devices configured from a single housing, but may also include combinations of the various devices exemplified above.

[0030] 1 shows only one terminal device, the processing system 1 can include multiple terminal devices corresponding to the number of stores or staff, for example. For example, multiple stores or multiple staff can share a single terminal device, or one store or one staff can use multiple terminal devices.

[0031] FIG. 2A is a block diagram showing a configuration of a processing device 100 according to an embodiment of the present disclosure. According to FIG. 2A, the processing device 100 includes a processor 111, a memory 112, and a communication interface 113. These components are electrically connected to each other via a control line and a data line. The processing device 100 does not need to include all of the components shown in FIG. 2A, and can be configured by omitting some of them, or can add other components. For example, it is possible to use an external memory, a database device, a server device, etc., which are communicably connected as a memory. It is also possible to distribute some of the processing to processing devices including other server devices and execute them. That is, the processing device 100 is not limited to only a single device, but also includes a case where it is distributed to multiple devices depending on the handling of the information and the processing load.

[0032] The processor 111 functions as a control unit that controls other components of the processing system 1 based on a processing program stored in the memory 112. The processor 111 is mainly composed of one or more CPUs, but may be appropriately combined with a GPU, an FPGA, or the like. The processor 111 generates demand forecast information for products based on the processing program stored in the memory 112, and generates recommendation information indicating the recommended number of products to be displayed. Specifically, the processor 111 executes, based on the processing program stored in the memory 112, "processing of generating demand forecast information related to the demand amount of products at any time using a demand forecast model generated based on demand information of products displayed in a store" and "processing of generating recommendation information indicating the recommended number of products to be displayed at any time based on the generated demand forecast information and display-related information related to the display of products."

[0033] The memory 112 is composed of RAM, ROM, non-volatile memory, HDD, SSD, etc., and functions as a storage unit. The memory 112 stores instructions and commands for various controls of the processing system 1 according to this embodiment as a processing program. Specifically, the memory 112 stores a program for the processor 111 to execute "a process of generating demand forecast information related to the demand amount of a product at any time by a demand forecast model generated based on demand information of the product displayed in the store" and "a process of generating recommendation information indicating the number of products recommended to be displayed at any time based on the generated demand forecast information and display related information related to the display of the products". In addition to the program, the memory 112 also stores various information stored in a product management table, a recommendation information management table, a work information management table, etc. Note that these pieces of information do not need to be stored in the memory 112 in the processing device 100 at all times, and may be stored in a database device or a server device installed remotely. In that case, the database device or the server device is also included in the memory 112.

[0034] The communication interface 113 functions as a notification unit for transmitting and receiving various information to and from the terminal device 200 or other processing devices connected via a wired or wireless network. Examples of the communication interface 113 include connectors for wired communication such as USB and SCSI, wireless communication transmitting and receiving devices for broadband wireless communication such as wireless LAN, Bluetooth (registered trademark), and LTE, infrared, and various connection terminals for printed circuit boards and flexible circuit boards.

[0035] Fig. 2B is a block diagram showing a configuration of a terminal device 200 according to an embodiment of the present disclosure. According to Fig. 2B, the terminal device 200 includes a processor 211, a memory 212, an input interface 213, an output interface 214, and a communication interface 215. These components are electrically connected to each other via a control line and a data line. Note that the terminal device 200 does not need to include all of the components shown in Fig. 2B, and may be configured with some of the components omitted or with other components added.

[0036] The processor 211 functions as a control unit that controls other components of the terminal device 200 based on a program stored in the memory 212. The processor 211 is mainly composed of one or more CPUs, but may be appropriately combined with a GPU, an FPGA, or the like. The processor 211 executes a process for outputting recommended information received from the processing device 100 based on a processing program stored in the memory 112. Specifically, the processor 211 executes a "process for receiving recommended information from the processing device 100 via the communication interface 215" and a "process for outputting recommended information received via the output interface 214" based on the program stored in the memory 212.

[0037] The memory 212 is composed of a RAM, a ROM, a non-volatile memory, a HDD, etc., and functions as a storage unit. The memory 212 stores instructions and commands for various controls of the processing system 1 according to this embodiment as programs. Specifically, the memory 212 stores programs to be executed by the processor 211, such as "processing for receiving recommended information from the processing device 100 via the communication interface 215" and "processing for outputting recommended information received via the output interface 214."

[0038] The input interface 213 functions as an input unit that accepts a user's operation input to the terminal device 200. Examples of the input interface 213 include physical key buttons and a touch panel having an input coordinate system corresponding to the display coordinate system of the display. In the case of a touch panel, icons are displayed on the display, and the operator performs an operation input via the touch panel to select each icon. The detection method of the target person's operation input by the touch panel may be any method such as a capacitive type or a resistive film type. The input interface 213-2 does not always need to be physically provided in the terminal device 200, and may be connected as necessary via a wired or wireless network. Therefore, in addition to the above, a mouse, a keyboard, etc. can also be used as the input interface 213-2.

[0039] The output interface 214 functions as an output unit for outputting various information. An example of the output interface 214 is a display, but is not limited to this and may be composed of other liquid crystal panels, organic EL displays, plasma displays, printers, etc. Also, a display is not necessarily provided, and for example, an interface for connecting to a display or the like that can be connected to the processing device 100 via a wired or wireless network can function as the output interface 214 that outputs display data to the display or the like.

[0040] The communication interface 215 functions as a communication unit for transmitting and receiving information to and from the processing device 100, other terminal devices 200, and other processing devices. Examples of the communication interface 215 include connectors for wired communication such as USB and SCSI, wireless communication transmitting and receiving devices for broadband wireless communication such as wireless LAN, Bluetooth (registered trademark), and LTE, infrared, and various connection terminals for printed circuit boards and flexible circuit boards.

[0041] 3. Various information used in processing in the processing system 1 Figures 3A, 3C, and 3D show various tables in which information is stored in the processing device 100 and transmitted to and received from the processing device 100 as the processing progresses. This information is updated and stored as needed as the processing progresses. Note that each piece of information shown in Figures 3A, 3C, and 3D may be stored in the memory 112 of the processing device 100, or may be stored in a database device or server device installed remotely and read out as needed as the processing progresses.

[0042] 3A is a diagram conceptually illustrating a product management table stored in the processing device 100 according to an embodiment of the present disclosure. According to Fig. 3A, the product management table stores information on the number of dishes to be cooked, the number of dishes to be discarded, the number of dishes sold, the number of dishes displayed, demand forecast information, recommendation information, and cooking cost information in association with product ID information.

[0043] "Product ID information" is information unique to products displayed in a store and is used to identify each product. As an example, product ID information is generated by a store staff member or a product manager each time a new type of product is registered. Note that any product ID information that can identify the product may be used, and various information such as the product name may be used.

[0044] The "cooking number information" is information indicating the number of products displayed in a store that are actually cooked at each time divided by an arbitrary interval. The information is used as one of the demand information of products displayed in a store and identified by each product ID information. The information is newly stored each time the number of products for each time is acquired as a result. The information may be deleted after a certain period of time has passed, or may be constantly added and stored. Specifically, the cooking number information is stored in association with each time and the number of products cooked, such as "2 at 0:00 on June 1, 2024", "8 at 5:00 on June 1, 2024", "10 at 8:00 on June 1, 2024", "10 at 11:00 on June 1, 2024", "5 at 14:00 on June 1, 2024", "10 at 17:00 on June 1, 2024", and "2 at 20:00 on June 1, 2024".

[0045] The "discarded number information" is information indicating the number of products actually discarded at each time divided at an arbitrary interval. The information is used as one of the demand information of the products displayed in the store and identified by each product ID information. The information is newly stored each time the number of each time is acquired as a performance. The information may be deleted after a certain period of time has passed, or may be constantly added and stored. Specifically, the discarded number information is stored in such a manner that each time is associated with the number of discarded products, such as "1 at 0:00 on June 1, 2024," "2 at 5:00 on June 1, 2024," "3 at 8:00 on June 1, 2024," "1 at 11:00 on June 1, 2024," "4 at 14:00 on June 1, 2024," "3 at 17:00 on June 1, 2024," and "0 at 20:00 on June 1, 2024."

[0046] The "sales number information" is information indicating the number of products actually sold at each time divided at an arbitrary interval. The information is used as one of the demand information of the products displayed in the store and identified by each product ID information. The information is newly stored each time the number of each time is acquired as a performance. The information may be deleted after a certain period of time has passed, or may be constantly added and stored. Specifically, the sales number information is stored in association with each time and the number of products sold, such as "2 units at 0:00 on June 1, 2024", "4 units at 5:00 on June 1, 2024", "8 units at 8:00 on June 1, 2024", "5 units at 11:00 on June 1, 2024", "6 units at 14:00 on June 1, 2024", "3 units at 17:00 on June 1, 2024", and "6 units at 20:00 on June 1, 2024".

[0047] The "display number information" is information indicating the number of products actually displayed at each time divided at any interval. The information is used as one of the demand information of the products displayed in the store and identified by each product ID information. The information is newly stored each time the number of each time is acquired as a performance. The information may be deleted after a certain period of time has passed, or may be constantly added and stored. Specifically, the display number information is stored in association with each time and the number of products displayed, such as "2 units at 0:00 on June 1, 2024," "4 units at 5:00 on June 1, 2024," "2 units at 8:00 on June 1, 2024," "6 units at 11:00 on June 1, 2024," "1 unit at 14:00 on June 1, 2024," "5 units at 17:00 on June 1, 2024," and "1 unit at 20:00 on June 1, 2024."

[0048] "Demand forecast information" is information related to the demand of each product predicted at each future time separated by an arbitrary interval. For example, the information is obtained as an output by inputting each piece of demand information into a demand forecast model by the processing device 100. Here, FIG. 3B is a diagram conceptually showing the demand forecast information stored in the processing device 100 according to an embodiment of the present disclosure. Specifically, FIG. 3B is a diagram showing demand forecast information including a predicted demand amount and its accuracy for a product with product ID information "A1" at a time specified by time information "J1". Specifically, according to FIG. 3B, for a product with product ID information "A1", demand forecast information and its accuracy are stored in correspondence with each other, such as demand forecast information F1-1 (e.g., 0 units) and its accuracy I1 (e.g., 5%) at the time specified by time information "J1", demand forecast information F1-2 (e.g., 1 unit) and its accuracy I2 (e.g., 7.5%), demand forecast information F1-3 (e.g., 2 units) and its accuracy I3 (e.g., 8.3%), demand forecast information F1-4 (e.g., 3 units) and its accuracy I4 (e.g., 23%), demand forecast information F1-5 (e.g., 4 units) and its accuracy I5 (e.g., 34%), and demand forecast information F1-n (e.g., n units) and its accuracy I1 (e.g., n%).

[0049] Although not shown in Fig. 3B, for the product with the product ID information "A1", similar demand time information is generated for other future times other than the time information "J1". Also, similar demand time information is generated for each time for other products other than the product ID information "A1".

[0050] The demand forecast information may be any information indicating the demand amount (sales volume) at each time. Here, the demand forecast information is calculated from a demand forecast model in the processing device 100 as an example, but it may simply be information indicating any classification and its accuracy, such as "demand is expected to increase," "demand is expected to decrease," "demand is expected to increase significantly," and "demand is expected to decrease significantly."

[0051] Returning to FIG. 3A again, the "recommended information" is information indicating the number of displays of each product recommended at each future time divided at an arbitrary interval. Here, FIG. 3C is a diagram conceptually illustrating a recommended information management table stored in the processing device 100 according to an embodiment of the present disclosure. Specifically, FIG. 3C is a diagram illustrating a table for managing recommended information stored in association with product ID information "A1" for a product having the product ID information "A1", and similar tables are generated for each product ID information in addition to the table associated with the product ID information "A1" in FIG. 3C. According to FIG. 3C, the recommended information is stored in association with each future time (time J1 to time Jn) divided at an arbitrary interval, and indicates the number of displays recommended for the product having the product ID information "A1". As an example, recommendation information indicating each time and the recommended number of displays is stored in association with each other, such as "4 items" recommended information for the time information of "June X, 2024, 0:00", "8 items" recommended information for the time information of "June X, 2024, 5:00", "3 items" recommended information for the time information of "June X, 2024, 8:00", "8 items" recommended information for the time information of "June X, 2024, 11:00", "10 items" recommended information for the time information of "June X, 2024, 14:00", "10 items" recommended information for the time information of "June X, 2024, 17:00", and "3 items" recommended information for the time information of "June X, 2024, 20:00".

[0052] The recommendation information may be any information indicating the number of products recommended to be displayed at each time. In the example of Fig. 3C, the recommended display number determined by the processing device 100 is shown, but it is also possible to use the number that needs to be replenished by subtracting the number currently displayed from the recommended display number as the recommendation information. In addition, even information that simply notifies the need for replenishment, such as "refill required", can be used as the recommendation information.

[0053] Returning to FIG. 3A again, the "cooking cost information" is information indicating the cost required for cooking each product identified by the product ID information. This information is used as one of the display-related information related to the display of the product identified by each product ID information. This information may be registered in advance by a store staff member or a product manager when a new product is registered, or may be updated each time the actual cost is calculated. As described above, the cooking cost information may be any information indicating the cost required for cooking, and may include various information such as the purchase cost of cooking utensils, the operating cost of cooking utensils, labor costs, or material costs. In this way, the more times a product is cooked, the higher the cooking cost becomes, so the lower the cooking cost, the more desirable it is. However, by using the cooking cost information, the number of times cooking is reduced, making it possible to efficiently operate a store.

[0054] FIG. 3D is a diagram conceptually illustrating a work information management table stored in a processing device 100 according to an embodiment of the present disclosure. Specifically, FIG. 3D is a diagram illustrating a table for managing work information in which the person in charge ID information of a person in charge working at a store with store ID information "Z1" and the working hours of the person in charge are associated with each other, and in addition to the table associated with the store ID information "Z1" in FIG. 3D, a similar table is generated for each store ID information of each store. Note that the work information managed in the work information management table is information used as one of the display-related information related to the display of the product identified by each product ID information.

[0055] The "time information" is information indicating each time period divided at an arbitrary interval. The time information is preferably the same as or corresponding to each time period divided at an arbitrary interval used in the cooking number information, waste number information, sales number information, display number information, and recommendation information shown in FIG. 3A. Specifically, the time information includes past times such as "June 1, 2024, 0:00", "June 1, 2024, 5:00", "June 1, 2024, 8:00", "June 1, 2024, 11:00", "June 1, 2024, 14:00", "June 1, 2024, 17:00", and "June 1, 2024, 20:00". Future times such as "June X, 2024, 0:00", "June X, 2024, 5:00", "June X, 2024, 8:00", "June X, 2024, 11:00", "June X, 2024, 14:00", "June X, 2024, 17:00", and "June X, 2024, 20:00" are stored. As an example of the information, the shift change time of the person in charge is set.

[0056] "Staff member ID information" is information unique to each store staff member and is used to identify each staff member. As an example, staff member ID information is generated each time a new staff member is registered by the store staff member or product manager. Any staff member ID information can be used as long as it can identify the staff member, and various information such as the name of the staff member can be used. In other words, staff member ID information is stored in association with each piece of time information, and staff member ID information for one or more staff members working at each time indicated by each piece of time information is stored. This makes it possible to manage who is working at each time indicated by each piece of time information. The "cooking count information" is information that indicates the number of times each product can be cooked in each time period specified by the time information. For example, the cooking count is limited to a certain number of times in each time period, such as once or twice, and the cooking count information is used to understand how many cooking times remaining are possible in comparison with this limit.

[0057] 4. Processing flow executed by the processing device 100 4 and 5 are diagrams showing processing flows executed in the processing device 100 according to an embodiment of the present disclosure. Specifically, FIG. 4 is a diagram showing a processing flow for processing related to generation of a demand prediction model and a recommendation information generation model. Also, FIG. 5 is a diagram showing a processing flow for generating recommendation information indicating the number of products recommended to be displayed at any given time. Each processing flow is mainly performed by the processor 111 of the processing device 100 reading and executing a program stored in the memory 112.

[0058] (A) Process flow for generating a demand forecast model and a recommendation information generation model 4, the processor 111 reads out various demand information used to generate a demand forecast model from the product management table (S111). Examples of the demand information to be read out include sales number information indicating the number of products actually sold at each time, and display number information indicating the number of products actually displayed in the store at each time.

[0059] The processor 111 generates a demand forecasting model based on the various demand information that has been read out (S112). As an example, the demand forecasting model performs machine learning using the read-out sales volume information and display number information as feature quantities, and predicts the demand amount (sales volume) of the product at any time in the future and its accuracy as demand forecast information. The demand forecasting model is preferably generated for each product. Therefore, the demand amount and accuracy are generated using the demand forecasting model generated for each product. Such a demand forecasting model is preferably generated based on a Poisson regression model according to the Poisson distribution, as will be described in detail below, but may of course be a model generated by another method.

[0060] The processor 111 stores the generated demand forecast model in the memory 112 (S113).

[0061] Next, the processor 111 reads out various demand information used to generate a recommendation information generation model from the product management table (S114). Examples of the demand information to be read out include sales volume information, cost rate information, and display number information. The cost rate information indicating the cost rate of a product is acquired by reading out an arbitrary value set in advance by a user or manager regardless of the product. However, for example, in the product management table shown in FIG. 3A, the cost rate of each product may be stored in association with each product ID information, and read out from the product management table in the same way as the sales volume information, etc.

[0062] The processor 111 generates a recommendation information generation model for outputting recommendation information indicating the number of recommended products to be displayed at each time interval divided by an arbitrary interval based on the various demand information read out (S115). As an example, the recommendation information generation model performs machine learning using the various demand information read out as features and the recommendation information as an objective function, and predicts the number of recommended products to be displayed at each time interval divided by an arbitrary interval as recommendation information. The recommendation information generation model is preferably generated for each product. Therefore, the recommendation information generation model generated for each product is used to generate recommendation information indicating the number of recommended products to be displayed at each time interval divided by an arbitrary interval. Such a recommendation information generation model is generated by a method using reinforcement learning, as described in detail below, but may of course be a model generated by another method.

[0063] The processor 111 stores the generated recommendation information generation model in the memory 112 (S116). This ends the processing flow.

[0064] 4, the demand forecasting model or the recommendation information generation model is generated by the processor 111 of the processing device 100, but each of these processes can also be performed by a processing device or a server device other than the processing device 100. Also, in FIG. 4, the generated demand forecasting model or the recommendation information generation model is stored in the memory 112, but it is also possible to store the demand forecasting model in another processing device or a server device.

[0065] (A1-1) Example of a demand forecast model based on the Poisson distribution An example of a demand forecasting model for predicting the demand volume at any time for each product displayed in a store based on demand information (sales volume information and display volume information) is described below. In this demand forecasting model, the probabilistic behavior of the sales volume indicated by the sales volume information is taken into account, and the approach is to estimate the demand volume forecast as a probability distribution rather than as a single estimated value. Therefore, this demand forecasting model is generated based on a Poisson regression model. Specifically, it is assumed that the sales volume at any time for each product follows a Poisson distribution, and a parameter μ of this Poisson distribution is estimated. In this case, it is desirable that the time granularity for calculating the predicted value in the demand forecasting model is the same as the time granularity for generating the recommendation information.

[0066] In addition, in demand forecasting, it is important to predict the number of sales assuming that there is a sufficient number of displays for each product. In a situation where there is a shortage of displays, even if there is demand from users, their purchasing behavior is restricted. Therefore, the true demand may exceed the actual sales number, and the true demand and the actual sales number may not necessarily match. Therefore, in this demand forecasting model, it is possible to estimate the true demand more precisely by considering the display number specified by the display number information, rather than simply using the sales number specified by the sales number information. Therefore, as shown in the demand forecasting model of formula (1), a demand forecasting model is constructed for each case where the sales number and the true demand are likely to match and the case where they are not likely to match. Specifically, for each time i divided into any intervals, the number of displays m specified by the display number information is calculated. iUsing the actual data of i M i In this case, the number of sales is m i The likelihood function l(μ; k, m) shown in the lower part of the formula (1) is generated and learning is performed. In addition, the sales volume k specified by the sales volume information at time i is i The number of displays is m i If k,m is smaller than k,m, the likelihood function l(μ;k,m) shown in the upper part of formula (1) is generated and learning is performed. In this way, by using a demand forecasting model that is trained separately according to the display number and sales number, it is possible to predict demand more accurately.

[0067]

number

[0068] Among the parameters used in the above formula (1), the parameter μ i is calculated by the following formula (2). Specifically, the feature amount of the training data at time i is x i ∈R M , the parameter μ of the Poisson distribution at time i i A model for predicting is generated as shown in the following formula (2). As shown in the following formula (2), the formula (2) includes day of the week information, which is one of the demand information, as an explanatory variable. In other words, since the number of products sold is also affected by the day of the week, by using the day of the week information, it is possible to predict demand more precisely. Note that s represents the effect common to all days of the week, β represents the effect for each day of the week, and x i indicates the day of the week for the i-th data item.

[0069]

number

[0070] In the above formula (2), θ is a parameter used in the model for predicting μ, and θ = {s, β}(s ∈ R, β ∈ R M) Therefore, the likelihood function L(θ) is expressed as in the following formula (3), which calculates the parameters θ and β in the demand forecasting model based on the Poisson distribution in formula (1).

[0071]

number

[0072] In addition, x in the above formula (2) i However, instead of or in addition to this, it is also possible to use event information indicating an event (e.g., a sales promotion event) held during the display of the product, or period information specifying the season or time when the product is displayed.

[0073] In this way, by utilizing the demand forecasting model of equation (1) with each parameter obtained from equations (2) and (3), it is possible to obtain the demand volume (sales volume) and its accuracy.

[0074] (A1-2) Examples of other demand forecasting models The processing device 100 can generate a demand forecasting model using other machine learning or the like in addition to the demand forecasting model exemplified in (A1-1) above. For example, the processor 111 of the processing device 100 acquires, as learning time information, information on each time divided at an arbitrary interval specified by the sales number information shown in FIG. 3A. Next, the processor 111 assigns the sales number information at each time specified by the sales number information as correct answer label information to each learning time information. The processor 111 executes a step of performing machine learning of a prediction pattern of a demand amount (sales number) at each future time using the learning time information and the sales number information at that time. As an example, the machine learning is performed by giving a set of learning time information and sales number information at that time (correct answer label) to a neural network combining neurons, and repeating learning while adjusting the parameters of each neuron so that the output from the neural network is the same as the correct answer label information. Then, when the above adjustment is completed, the processor 111 acquires the adjusted model as a demand forecasting model.

[0075] Note that the demand forecasting model can also be generated using machine learning techniques such as neural networks, convolutional neural networks, multi-layer Herceptrons (MLP), LSTM (Long Short Term Memory), GRU (Gated Recurrent Unit), GNN (Graph Neural Network), Transformer, etc., gradient boosting decision trees (GBDT) such as LightGBM (Light Gradient Boosting Machine), XGBoost, CatBoost, etc., ridge regression, logistic regression, support vector regression (SVR), nearest neighbor methods, decision trees, regression trees, random forests, etc.

[0076] (A1-3) Examples of other demand forecasting models Although the above (A1-1) and (A1-2) show examples of techniques using machine learning, it is also possible to generate a model using a logic circuit approach based on past performance data such as sales volume information stored in the product management table shown in Fig. 3. As one example, it is also possible to identify a past time that meets predetermined conditions, such as the same day of the week as an arbitrary time to be predicted, or a day on which the same event as that held at an arbitrary time to be predicted is held, from the information stored in the product management table, and to construct a demand forecast model such that the sales volume information for that time is used as demand forecast information.

[0077] (A2) Example of recommendation information generation model An example of a recommendation information generation model for generating recommendation information indicating the recommended number of products to display at each time t (t=0, 1, 2, ..., T) divided at any interval for each product by utilizing the demand prediction model generated in S112 of Fig. 4 is described below. The recommendation information generation model uses a reinforcement learning model that uses waste number information and sales number information, which are types of demand information, as feature quantities, and therefore it is possible to output the optimal recommended number of products to display that can reduce food waste and increase profits.

[0078] To generate a recommendation information generation model, we first formulate the recommended number of displays as an objective function J. Specifically, we define the optimal number of additional displays b t , the number of products sold is n t , the selling price is γ, and the cost rate is KThe objective function J is expressed as shown in formula (4). The processor 111 reads out and uses, as the display-related information, display number information indicating the number of additional displays, sales number information indicating the number of products sold, cost rate information indicating the cost rate, and sales price information indicating the sales price. The display number information and sales number information are information read out from the product management table in FIG. 3A. Although not particularly shown in FIG. 3A, the sales price information is information stored in the product management table in association with the product ID information of each product. The cost rate information is information acquired by reading out an arbitrary value set in advance by a user or manager regardless of the product. However, for example, in the product management table shown in FIG. 3A, the cost rate of each product may be stored in association with each product ID information, and read out from the product management table in the same manner as the sales number information and the like.

[0079]

number

[0080] Here, the first term in equation (4) represents the sum of the sales amount for each time period separated by an arbitrary interval, and the second term represents the sum of the display costs for each of those times. The sum of the sales amount for each time period is calculated from the number of items sold and the sales price. The sum of the display costs for each time period is calculated from the number of items displayed, the sales price, and the cost rate. Furthermore, E[·] represents the expected value regarding the true demand n' (i.e., the amount of sales profit). Furthermore, K can be interpreted as a hyperparameter that controls the trade-off between food waste and profit. In other words, the objective function J is the amount of sales profit obtained from displaying the product in question.

[0081] Next, the recommendation information, which is the recommended number of displays at time t separated by an arbitrary interval, can be shown according to a Markov decision process, and the objective function J in the above formula (4) is converted as shown in formula (5). Specifically, if the value' (i.e., the expected profit amount) when the display state at time t is s and the display plan at time t is π(t) items is r(s,t,π(t)), and the probability that the display state at time t+1 will be s' when the display state at time t is s and the display plan at time t is π(t) items is p(s'|s,t,π(t)), the objective function J in the above formula (4) is converted as shown in formula (5).

[0082]

number

[0083] The display state s refers to the state of the products displayed at time t. As an example, it is information indicating when the products displayed at time t will be discarded, such as 3 products to be discarded at time t+1 and 2 products to be discarded at time t+2. The information is acquired by reading out information such as discard number information as display-related information from the product management table shown in FIG. 3A. The display plan π(t) is information indicating the recommended number of products to be displayed at time t.

[0084] Here, E π [·|t] represents the value (i.e., the expected profit amount) associated with the display state s at any time interval t when following the display plan π. In other words, by finding a display plan π that maximizes this objective function, it is possible to optimize the reduction of food waste and the improvement of profits.

[0085] Next, we formulate r(s, t, π(t)) and p(s'|s, t, π(t)) used in the above formula (5). Specifically, we first define the number of sales at a time t divided by an arbitrary interval as n' t , and its probability distribution is f t , display status s t , the recommended number of displays is at where s t indicates the time until disposal of each displayed item in the form of an array. Based on this setting, (s t ,a t ) to s t+1 The probabilistic transition to (t=0,1,2,...,T-1) is expressed as follows: 1:b t =max(a t -|s t |,0) ·2:n' t ~f t 3:n t =min(n' t ,|s t |) 4:s t+1 =g(s t ,n t ,b t )

[0086] Furthermore, |s| is the total number of exhibits in s, and the function g(s t ,n t ,b t ) by s t The following operations are performed on Step 1: n in descending order t Eliminate individual elements Step 2: Decreases the value of all elements by 1 Step 3: Eliminate elements with a value of 0 Step 4: Set the elements with value h to b t Add

[0087] Here, h indicates the time from when a product is put on display until it is discarded. Step 1 indicates that products that were put on display the longest will be sold first. Step 2 indicates that the deadline until disposal will become shorter. Step 3 indicates that products that have reached their deadline will be discarded. Step 4 indicates that new products will be added to the display.

[0088] Based on the above, r(s,t,π(t)) and p(s'|s,t,π(t)) can be formulated as shown in Equation (6) and Equation (7), respectively. Here, f t(n) is the probability that the sales volume at time t, separated by an arbitrary interval, will be n, and uses the demand (sales volume n) and its accuracy (occurrence probability), which are the demand forecast information output from the demand forecasting model. I(·) is a function that takes the value of 1 when the equation in parentheses holds, and takes the value of 0 when it does not. In other words, r(s,t,π(t)) in equation (6) is expressed as a reward function in which the expected value of sales amount at time t is used as the loss function, and the display cost at time t is used as the penalty term.

[0089]

number

number

[0090] Next, the objective function J formulated as above is optimized with respect to the display plan π. If the approach is to work backwards from time T, which is the end of a period of time (t=0,1,2,...,T) separated by an arbitrary interval, it is possible to calculate the recommended number of displays for each (s,t)=(display state, time), but it is not possible to calculate the recommended number of displays for each time t. On the other hand, when considering the recommended number of displays for each time t, it is possible to calculate the recommended number of displays at a specific time by fixing the recommended display plans for other times. Specifically, by fixing the display plans for t=0,1,2,...,t'-1,t'+1,t'+2,...,T, the recommended number of displays for t=t' can be calculated as follows.

[0091] <Process 1> By fixing the display plan for t=t'+1, t'+2, ..., T, the value of the display plan at t=t'+1 (i.e., the amount of sales profit from time t'+1 onwards) is calculated using a backward recurrence formula. <Process 2> By fixing the display plan for t=0, 1, 2, ..., t'-1, the occurrence probability of the display state at t=t' is calculated using a forward recurrence formula.

[0092] Here, it is possible to determine the recommended number of displays at time t=t' by combining the value of the display plan from time t'+1 onwards (i.e., the amount of sales profit from time t'+1 onwards) calculated by the above processes 1 and 2, the occurrence probability of the display state at t=t', and demand forecast information by a demand forecast model at time t=t'. Then, processes 1 and 2 are repeatedly subjected to reinforcement learning until the value of the display plan π at times t separated by any interval converges, whereby the number of displays π at time t is calculated. Furthermore, by repeatedly performing reinforcement learning at times other than time t in the same manner until the value of the display plan π converges, the optimal display plan π for all times, i.e., the recommended number of products to be displayed at each time, is calculated.

[0093] Therefore, the processes shown in the above processes 1 and 2 are formulated as shown in equations (8) to (10) and used as a recommendation information generation model.

[0094] <Process 1A> For each t=T, T-1, ..., 1, 0, V is generated based on the recommendation information generation model shown as the backward recursion formula in (8). π Calculate. <Process 2A> For each time t=0,1,...,T-1, we use the forward recursion formula (9) and (10) to generate recommendation information. π Calculate (s|t) and π(t).

[0095]

number

number

number

[0096] In addition, V π(s,t) indicates the value of the display plan when the display state is s at time t and the display plan π is followed from time t onwards (i.e., the amount of sales profit from time t'+1 onwards). π (s|t) indicates the probability of the display state when following the display plan π up to time t.

[0097] Specifically, in process 1A, a random value is set for the display plan π at time t, and V, which indicates the value of the display plan at time t'+1 (i.e., the amount of sales profit after time t'+1) when the display state is s at time t, is calculated using equation (8). π At this time, as shown in formula (6) and formula (7), the demand forecast information output from the demand forecast model, that is, the demand amount (sales volume n) and its accuracy (occurrence probability), are used in calculating the values ​​of r(s, t, π(t)) and p(s'|s, t, π(t)).

[0098] Next, in process 2A, the display plan π at time t is set to an arbitrary value the same as in process 1A, and p π (s|t) is calculated. Then, based on the value of the display plan calculated using equation (8) (sales profit amount) and the probability calculated using equation (9), the number of displays a that maximizes the value of the display plan (sales profit amount) is calculated using equation (10). This series of processes is repeated as reinforcement learning until the display plan π converges. Similarly, reinforcement learning is also performed for other times other than time t. This allows the recommended number of displays a to be calculated for each time period separated by any time interval.

[0099] (B) Processing flow for generating recommendation information According to FIG. 5, the processor 211 of the terminal device 200 accepts an operation input by a person in charge or an administrator via the input interface 213 and selects a product and a time (for example, a date) for which recommendation information is to be output. The processor 211 of the terminal device 200 transmits a generation request for recommendation information to the processing device 100 via the communication interface 215 together with product ID information indicating the selected product and information indicating the time. When the processor 111 of the processing device 100 receives the generation request or the like via the communication interface 113, it inputs information indicating the desired time to a demand forecasting model generated corresponding to the received product ID information among the demand forecasting models generated in S112 of FIG. 4 (S211). As an example, the information is input to a likelihood function L(θ) which is a demand forecasting model shown in Equation (3). Note that the demand forecasting model used here is merely an example, and it is naturally possible to use other demand forecasting models described above in (A1-2) and (A1-3), etc.

[0100] Next, the processor 111 obtains, from the demand forecast model, the demand amount and its accuracy for each time period obtained by dividing the input time period into arbitrary intervals, as demand forecast information (S212). In S212, it is sufficient to obtain the demand forecast information for each time period, and the demand amount and its accuracy calculated by the person in charge or the manager may be input via the input interface 213 of the terminal device 200 to obtain the demand forecast information. The processor 111 stores the obtained demand forecast information in the product management table in association with the product ID information.

[0101] Next, processor 111 estimates the value of the display plan (i.e., the amount of sales profit) at each time separated by an arbitrary interval (S213). As an example, as described above as process 1A, the demand forecast information, etc. acquired in S212 is input to a recommendation information generation model shown as a backward recurrence formula of formula (8), and the value of the display plan (i.e., the amount of sales profit) at time t is acquired as an output. Similarly, the value of the display plan (i.e., the amount of sales profit) at times other than time t is also acquired as an output from the recommendation information generation model.

[0102] Next, the processor 111 estimates the occurrence probability of the display state s at each time interval (S214). As an example, as described above as process 1B, by inputting a time t among the times separated by an arbitrary interval and the demand forecast information acquired in S212 into the recommendation information generation model shown as a forward recurrence formula of formula (9), the occurrence probability p π (s|t) is obtained as the output. Similarly, the occurrence probability of the display state (e.g., s-1) at time other than t (e.g., p π (s|t-1)) is also obtained as output from the recommendation information generation model.

[0103] Next, processor 111 determines the number of products to be displayed at each time based on the value of the display plan (i.e., the amount of sales profit) at each time segmented by an arbitrary interval acquired in S213 and the occurrence probability of the display state at each time segmented by an arbitrary interval acquired in S214. As an example, as described as process 1B above, the value of the display plan at time t acquired in S213 (i.e., the amount of sales profit) and the occurrence probability p πFrom (s|t), the recommended display number a of products at time t is calculated. Then, processor 111 repeats the processes of S213 to S215 until the display plan π converges, and determines the finally obtained display number a as the recommended display number a of products. Similarly, the recommended display number a of products at times other than time t is determined.

[0104] Here, FIG. 6 is a diagram conceptually showing a process of generating recommendation information at time t by the processing device 100 according to an embodiment of the present disclosure. Specifically, it is a diagram conceptually showing the value of the display plan π (i.e., the amount of sales profit) calculated by the recurrence formula shown in the above step 1, and the occurrence probability calculated by the recurrence formula shown in step 2. According to FIG. 6, as shown in S213 of FIG. 5, the value of the display plan at time t (i.e., the total amount of sales profit from time t+1 onward) is calculated by fixing the display plan π at each time t+1, t+2, ..., T. That is, based on the recurrence formula of formula (8), in order to maximize the amount of sales profit at time T, which is the terminal time, the display number that should be at the time immediately before that is calculated from the display number that should be. Then, by repeating this calculation, the display number that should be at time t is calculated from the display number that should be at time t+1.

[0105] 5, the occurrence probability of display plan s at time t is calculated by fixing display plan π at each time t=0, 1, 2, ..., t-1. That is, based on the recurrence formula of formula (9), the occurrence probability of display plan s at time 0, which is the start point, is calculated as 0 The probability of occurrence of the display state s1 occurring at the next time, time 1 (i.e., the state of the products displayed at time 1) is calculated from the display state s1 at time t-1. Then, by repeating this calculation, the display state S t-1 The probability of occurrence of display state s at the next time t is p π (s|t) is calculated.

[0106] Then, based on the value of the display plan (sales profit amount) calculated using equation (8) and the probability calculated using equation (9), the number of displays a that maximizes the value of the display plan (sales profit amount) is calculated using equation (10).

[0107] 5, the processor 111 stores the calculated number of recommended products to be displayed at each time as recommendation information in association with product ID information in the product management table. The processor 111 then transmits the stored recommendation information via the communication interface 113 to the terminal device 200 that has transmitted the request to generate the recommendation information (S216). This ends the processing flow.

[0108] Although not shown in detail in Fig. 5, the terminal device 200 can display the recommended information received via the output interface 214 on a display or the like. This allows the person in charge or the manager to accurately grasp how many products should be displayed at what time, and further, how many should be cooked based on the current display quantity information.

[0109] In addition, although not shown in detail in Figure 5, for products other than the product for which recommendation information has been generated by the processing flow of Figure 5, it is possible to generate recommendation information for the other products by processing similar to the processing flow of Figure 5, using each demand forecasting model and recommendation information generation model generated corresponding to each of the other products.

[0110] In this way, according to this processing flow, the recommended display quantity for each product is calculated using not only a demand forecast but also display-related information such as display quantity information, sales quantity information, cost rate information, and sales price information. Therefore, not only can the recommended display quantity be calculated more accurately or efficiently, but it is also possible to reduce food waste and improve profits by operating the store according to the calculated display quantity.

[0111] (C) Example of using display-related information As described above, in the process shown in Figures 4 and 5, the recommended number of products to be displayed is calculated using display-related information such as display number information, sales number information, cost rate information, and sales price information, rather than simply demand forecasting. This not only makes it possible to calculate the recommended number of products to be displayed more accurately or efficiently, but also makes it possible to reduce food waste and improve profits by operating the store according to the calculated number of products to be displayed. However, in this process, by further calculating the recommended number of products to be displayed at each time based on other display-related information, the calculation can be made even more accurate or efficient, making it possible to further reduce food waste and improve profits.

[0112] 4 and 5, examples of the display-related information include cooking cost information indicating the cooking cost of the displayed products, cooking count information indicating the number of times each product can be cooked at each time, the location of the store, the cooking time required to cook the product, or a combination of these. Below, a case where cooking cost information or cooking count information is used as display-related information will be described.

[0113] Here, the cooking cost specified by the cooking cost information increases as the number of times the product is cooked increases, and is therefore an important factor in increasing profits. Therefore, by providing the cooking cost information as a penalty term in the recommendation information generation model, it becomes possible to generate recommendation information more accurately and efficiently.

[0114] In addition, in actual stores, cooking of products is often carried out according to the timing of the change in the working hours of the staff in charge, that is, the so-called shift time. In addition, for a staff member in a store who has various tasks, cooking multiple times during his / her working hours not only interferes with other tasks, but also increases cooking costs discharged as labor costs. Therefore, it is desirable to optimize cooking so that cooking is done only once in a time period divided by an arbitrary interval managed in the staff member's work information management table. Therefore, by using cooking count information indicating the number of times each product can be cooked in each time period as a variable of the recommendation information generation model, it is possible to generate recommendation information more accurately or efficiently.

[0115] The cooking cost and cooking count information are obtained from the cooking cost information shown in FIG. 3A or the work information management table shown in FIG. 3D, respectively.

[0116] (C-1) Example of using cooking cost information as display-related information First, an example in which cooking cost information is used as display-related information will be described. The processor 111 of the processing device 100 gives the cooking cost specified by the cooking cost information as a penalty term in the recommendation information generation model generated in S115 of FIG. 4. Here, as shown in the above formulas (6) and (7), the recommendation information generation model is generated by formulating r(s, t, π(t)), which is the value (i.e., the expected profit amount obtained) when the number of displays at time t is π(t), and p(s'|s, t, π(t)), which is the probability that the display state at time t+1 will be s' when the display state at time t is s and the display plan at that time is π(t). At this time, r(s, t, π(t)) in formula (6) is expressed as a regression model in which the sum of sales amounts at time t is the loss function and the sum of display costs at time t is the penalty term, and the cooking cost of the product at time t is further added as a penalty term.

[0117] Therefore, the value of the display plan (i.e., the amount of sales profit after time t'+1) is calculated by the backward recurrence formula shown in formula (8), the probability of the display state becoming s is calculated by the forward recurrence formula shown in formula (9), and the number of displays a that maximizes the value of the display plan (amount of sales profit) is calculated by formula (10), where each recurrence formula, which is the recommendation information generation model, is shown as formula (8a), formula (9a), and formula (10a). Note that the cooking cost is shown by φ.

[0118]

number

number

number

[0119] In this way, first, in process 1A, a random value is set for the display plan π at time t, φ is set as the cooking cost, and V, which indicates the value of the display plan at t'+1 (i.e., the amount of sales profit after time t'+1) when the display state is s at time t, is calculated using equation (8a). π In this case, as in the case of formula (8), the demand forecast information output from the demand forecast model, that is, the demand amount (sales volume n) and its accuracy (occurrence probability), are used in the calculation of each value of r(s, t, π(t)) and p(s'|s, t, π(t)) as shown in formulas (6) and (7).

[0120] Next, in process 2A, we set the display plan π at time t to an arbitrary value the same as in process 1A, and calculate p π(s|t) is calculated. Then, based on the value of the display plan calculated using equation (8a) (sales profit amount) and the probability calculated using equation (9a), the number of displays a that maximizes the value of the display plan (sales profit amount) is calculated using equation (10a). This series of processes is repeated as reinforcement learning until the display plan π converges. Similarly, reinforcement learning is also performed for other times other than time t. This allows the recommended number of displays a to be calculated for each time period separated by any time interval.

[0121] As described above, the generated recommendation information generation model can be similarly used to calculate the value of the display plan (i.e., the amount of sales profit from time t'+1 onwards) described in S213 and S214 of Figure 5, and to calculate the probability that the display state will be s.

[0122] Therefore, by providing cooking cost information as a penalty term in the recommendation information generation model, it is possible to generate recommendation information more accurately and efficiently.

[0123] (C-2) Example of using cooking frequency information as a variable in the recommendation information generation model Next, an example in which cooking count information is used as a variable of the recommendation information generation model will be described. The processor 111 of the processing device 100 refers to the work information table and calculates the remaining cooking count based on the cooking count information at each time specified by the time information. Here, as shown in the above formula (6) and formula (7), the recommendation information generation model is generated by formulating r(s, t, π(t)), which is the value (i.e., the expected profit amount) when the number of displays at time t is π(t), and p(s'|s, t, π(t)), which is the probability that the display state at time t+1 will be s' when the display state at time t is s and the number of displays at that time is π(t). At this time, p(s'|s, t, π(t)) in formula (7) is shown as the sum of the probability of sales occurring such that the display state becomes s' at time t+1, and the probability of transitioning to the remaining number of cooking times c' at time t+1 when the remaining number of cooking times at time t is c is further added as a variable.

[0124] Therefore, the value of the display plan (i.e., the amount of sales profit from time t'+1 onwards) is calculated by the backward recurrence formula shown in formula (8), the probability that the display state will be s is calculated by the forward recurrence formula shown in formula (9), and the number of displays a that maximizes the value of the display plan (amount of sales profit) is calculated by formula (10), where each recurrence formula, which is the recommendation information generation model, is shown as formulas (8b), (9b), and (10b). Note that the remaining number of times that cooking can be done at time t is shown by c, and the remaining number of times that cooking can be done at time t+1 is shown by c'.

[0125]

number

number

number

[0126] In this way, first, in process 1A, a random value is set for the display plan π at time t, and then, using equation (8b), when the display state at time t is s and the number of times cooking is possible is c, we calculate V π In this case, r(s,c,t,π(c,t)), which is the value (i.e., the expected profit amount) when the number of items on display at time t is π(t) and the number of times cooking can be done is c, is calculated using formula (6) based on the demand (volume sales number n) and its accuracy (occurrence probability), which are the demand forecast information output from the demand forecast model. In addition, when the display state at time t is s, the display plan at time t is π(t), and the remaining number of times cooking can be done is c, the probability that the display state at time t+1 will be s' is calculated using formula (6) based on the demand (volume sales number n) and its accuracy (occurrence probability), which are the demand forecast information output from the demand forecast model.

[0127] Next, in process 2A, the display plan π at time t is set to an arbitrary value the same as in process 1A, and p π (s,c|t) is calculated. Then, based on the value of the display plan calculated using equation (8b) (sales profit amount) and the probability calculated using equation (9b), equation (10b) is used to calculate the number of displays a that maximizes the value of the display plan (sales profit amount) when the number of cooking times possible at time t is c. This series of processes is repeated as reinforcement learning until the display plan π converges. Similarly, reinforcement learning is also performed for other times other than time t. This allows the recommended number of displays a to be calculated for each time period separated by any time interval.

[0128] In the above (C-1) and (C-2), the cooking cost information and cooking count information are used separately in the recommendation information generation model, but it is also possible to use both in combination. By using both in combination, it is possible to calculate the recommended display quantity particularly accurately.

[0129] As described above, in the present embodiment, it is possible to provide a processing device, a processing program, a processing method, and a processing system that are capable of more appropriately calculating the recommended number of products to display. [Explanation of symbols]

[0130] 1 Processing System 100 Processing equipment 200 Terminal Equipment

Claims

1. A processing device comprising at least one processor, the at least one processor: generating demand forecast information including the number of products sold at a given time by inputting an arbitrary time into a demand forecast model generated by machine learning based on learning demand information including information on the number of products displayed in a store at a given time and information on the number of products sold of the said products; generating recommendation information indicating the number of displays of the products recommended at the arbitrary time by inputting the arbitrary time and the generated demand forecast information into a recommendation information generation model generated by machine learning using, as an objective function, recommendation information for learning indicating the number of displays of the products recommended at the arbitrary time, the recommendation information being calculated based on display-related information including at least one of display number information, sales number information, cost rate information, and selling price information of the products at the arbitrary time; A processing unit configured to perform processing.

2. The processing device according to claim 1 , wherein the product is a food or beverage product that is prepared by a staff member at the store.

3. The processing device according to claim 2 , wherein the display-related information includes cooking frequency information indicating the number of times that the person in charge can cook the product within the given time period.

4. The processing device according to claim 2 , wherein the display-related information includes cooking cost information indicating a cost of preparing the product for displaying the product.

5. The processing device according to claim 1 , wherein the demand forecast information includes the sales volume and a probability of achieving the sales volume.

6. The processing device according to claim 1 , wherein the demand forecasting model is generated based on a Poisson regression model.

7. The processing device according to claim 1 , wherein the arbitrary time is determined based on work information of a person who prepares the product.

8. When executed by at least one processor, generating demand forecast information including the number of products sold at a given time by inputting an arbitrary time into a demand forecast model generated by machine learning based on learning demand information including information on the number of products displayed in a store at a given time and information on the number of products sold of the said products; generating recommendation information indicating the number of displays of the products recommended at the arbitrary time by inputting the arbitrary time and the generated demand forecast information into a recommendation information generation model generated by machine learning using, as an objective function, recommendation information for learning indicating the number of displays of the products recommended at the arbitrary time, the recommendation information being calculated based on display-related information including at least one of display number information, sales number information, cost rate information, and selling price information of the products at the arbitrary time; A processing program that causes the at least one processor to function in such a manner.

9. A method of processing executed by at least one processor, comprising: generating demand forecast information including the sales volume of the product at a given time by inputting a given time into a demand forecast model generated by machine learning based on learning demand information including information on the number of products displayed in a store at a given time and information on the number of products sold of the product; generating recommendation information indicating the number of displays of the products recommended at the arbitrary time by inputting the arbitrary time and the generated demand forecast information into a recommendation information generation model generated by machine learning using, as an objective function, recommendation information for learning indicating the number of displays of the products recommended at the arbitrary time, the recommendation information being calculated based on display-related information including at least one of display number information, sales number information, cost rate information, and selling price information of the products at the arbitrary time; A processing method comprising:

10. A processing device according to any one of claims 1 to 7, a terminal device that is arranged in a store where the products are displayed so as to be able to communicate with the processing device, and is configured to receive recommendation information indicating a recommended number of the products to be displayed from the processing device, and to output the received recommendation information; 13. A processing system comprising:

Images