Demand forecasting device and demand forecasting method

Abstract

This invention provides a demand forecasting device and demand forecasting method capable of predicting the demand for each type of product offered in stores and other establishments within a building. [Solution] A demand forecasting device for goods offered in a building, comprising: a floor attendance output unit that outputs floor attendance figures representing an indicator of the number of people present on each floor of the building; a menu feature extraction unit that selects or determines menu features for each item on the menu; a floor-specific menu preference storage unit that stores floor-specific menu preference figures representing the degree of preference for each menu feature on each floor; and a demand forecasting unit that outputs the predicted number of demands for each menu item on the forecast target day based on the floor attendance figures for a predetermined period on the forecast target day and the floor-specific menu preference figures.

Description

[Technical Field]

[0001] The present invention relates to a demand forecasting device and a demand forecasting method. [Background technology]

[0002] As a technology for predicting the number of items to be provided in stores and facilities located in office buildings and commercial buildings, for example, Patent Document 1 discloses a provision quantity prediction device having a first acquisition unit, a second acquisition unit, a third acquisition unit, and a prediction unit, wherein the first acquisition unit acquires the actual number of occupants, which is the number of people who were in the facility at a time prior to the time when items are provided in the facility, determined on a daily basis for a predetermined period in the past, and acquires the actual number of occupants for each day in the predetermined period, the second acquisition unit acquires the actual number of occupants for a target day, which is the day for which the number of items to be provided to occupants is to be predicted, the third acquisition unit acquires the actual number of items provided for each day in the predetermined period, and the prediction unit predicts the number of items to be provided for a target day based on the actual number of occupants for each day, the actual number of items provided for each day, and the actual number of occupants for the target day. Similarly, Patent Document 2 discloses a serving quantity prediction device having a occupancy count determination unit, a utilization rate estimation unit, and a prediction unit, wherein the occupancy count determination unit determines the aggregated number of occupants, which is the number of people who were in the facility at a time prior to the time when meals are served in a facility with a cafeteria; the utilization rate estimation unit determines the actual utilization rate of the cafeteria in the past based on the ratio of the number of people who have used the cafeteria in the past to the aggregated number of occupants in the past, and determines the estimated utilization rate of the cafeteria on the target day based on the actual utilization rate of the cafeteria in the past; and the prediction unit determines the predicted number of meals to be served on the target day based on the estimated utilization rate of the cafeteria on the target day and the aggregated number of occupants on the target day. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2018-205877 [Patent Document 2] Japanese Patent Publication No. 2019-32753 [Overview of the project] [Problems that the invention aims to solve]

[0004] When predicting the number of items to be provided (e.g., meals) based on the number of employees present in a building, as described in Patent Documents 1 and 2 above, there is a problem in that if multiple types of items (e.g., dishes and menus in a cafeteria) are provided, it is difficult to predict the number of items to be provided for each menu item based on the number of employees.

[0005] This invention has been made in view of these problems, and aims to provide a demand forecasting device and demand forecasting method that can predict the demand for goods in stores and other establishments within a building, as well as predict the demand for each type of goods offered. [Means for solving the problem]

[0006] The present invention includes several means for solving at least part of the above problems, one example being as follows: A demand forecasting device for goods provided in a building, comprising: a floor attendance output unit that outputs floor attendance, which represents an index of the number of people present on each floor of the building; a menu feature extraction unit that selects or determines menu features for each item on the menu; a floor-specific menu preference storage unit that stores floor-specific menu preference, which represents the preference level for each menu feature, for each floor; and a demand forecasting unit that outputs the predicted demand for each menu on the forecast target day based on the floor attendance for a predetermined period on the forecast target day and the floor-specific menu preference. [Effects of the Invention]

[0007] According to the present invention, it becomes possible to predict the demand for goods in stores and other establishments. Furthermore, it becomes possible to predict the demand for each type of goods offered in stores and other establishments.

[0008] Other issues, configurations, and effects not mentioned above will be clarified by the following description of the embodiments.

Brief Description of the Drawings

[0009] [Figure 1] It is a conceptual diagram showing the concept of predicting the number of meals in the first embodiment. [Figure 2] It is a diagram showing an example of the configuration of the meal number prediction device in the first embodiment. [Figure 3] It is a diagram showing an example of a flowchart of the meal number prediction process by the meal number prediction device in the first embodiment. [Figure 4] It is a table showing an example of elevator weight data and floor integrated weight in the first embodiment. [Figure 5] It is a table showing an example of menu data and menu feature quantities in the first embodiment. [Figure 6] It is a table showing an example of menu preference degrees by floor in the first embodiment. [Figure 7] It is a table showing an example of the floor integrated weight, menu feature quantities, menu preference degrees by floor input to the meal number prediction unit in the first embodiment, and the predicted meal number predicted by the meal number prediction unit. [Figure 8] It is a graph showing an example of the change in floor integrated weight according to time in the first embodiment. [Figure 9A] It is a diagram showing an example of a display screen of the predicted meal number output by the meal number prediction unit in the first embodiment. [Figure 9B] It is a diagram showing an example of a display screen of the predicted meal number output by the meal number prediction unit in the first embodiment. [Figure 10] It is a diagram showing an example of the configuration of the meal number prediction device in the second embodiment. [Figure 11] It is a diagram showing an example of a flowchart of the learning / updating process of menu preference degrees by floor by the learning unit in the meal number prediction device in the second embodiment. [Figure 12] It is a table showing an example of the floor integrated weight, menu feature quantities, predicted meal number, actual meal number on the learning target day input to the learning unit in the second embodiment, the prediction difference, and the floor-by-menu preference degree update value for each menu feature quantity calculated by the learning unit. [Figure 13] This figure shows an example of the configuration of the third embodiment of the meal consumption prediction device. [Figure 14] This figure shows an example of a flowchart of the feature term classification data creation process by the feature term classification data creation unit in the menu feature extraction unit of the third embodiment. [Figure 15] This table shows an example of the characteristic phrase classification data 111 in the third embodiment. [Figure 16] This table shows an example of the matching score as a menu feature, including the text data of the menu name in the menu data matched by the feature word matching unit of the third embodiment, the feature words registered in the feature word classification data and their corresponding related words, and the matching score as a menu feature. [Figure 17] This figure shows an example of the configuration of the fourth embodiment of the meal consumption prediction device. [Figure 18] This table shows an example of a menu image and an analysis score as a menu feature in the menu data analyzed by the menu image analysis unit of the fourth embodiment. [Figure 19] This figure shows an example of the configuration of the meal consumption prediction device according to the fifth embodiment. [Figure 20] This table shows an example of nutritional components in menu data classified by the nutrition category classification unit of the fifth embodiment, and a classification score as a menu feature. [Figure 21] This figure shows an example of the configuration of the sixth embodiment of the meal consumption prediction device. [Figure 22] This figure shows an example of a flowchart for the process of calculating the recommended purchase quantity by the recommended purchase quantity calculation unit of the sixth embodiment. [Figure 23] This figure shows an example of the configuration of the seventh embodiment of the meal consumption prediction device. [Figure 24] This figure shows an example of a flowchart for the process of creating a revised batch quantity proposal by the batch quantity difference calculation unit of the seventh embodiment. [Figure 25] This figure shows an example of the configuration of the eighth embodiment of the meal consumption prediction device. [Figure 26]This figure shows an example of a flowchart for the process of predicting the final number of meals consumed on a given day by the final number of meals consumed on a given day prediction unit of the seventh embodiment. [Figure 27] This graph shows an example of the trend in the average number of meals consumed for menu items where the food category, which is a menu feature in the eighth embodiment, is "meat". [Figure 28] This figure shows an example of a flowchart for the food surplus / deficiency notification process by the food surplus / deficiency notification unit of the eighth embodiment. [Figure 29] This figure shows an example of the configuration of the ninth embodiment of the meal consumption prediction device. [Figure 30A] This figure shows an example of the configuration of the cage weight sensor in the first embodiment. [Figure 30B] This figure shows an example of the configuration of the cage weight sensor in the first embodiment. [Modes for carrying out the invention]

[0010] Embodiments of the present invention will be described below with reference to the drawings. The embodiments are illustrative examples for explaining the present invention, and have been omitted and simplified as appropriate for clarity of explanation. The present invention can also be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

[0011] The position, size, shape, and extent of each component shown in the drawings may not represent the actual position, size, shape, and extent in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the position, size, shape, and extent disclosed in the drawings. When there are multiple components having the same or similar function, they may be described using the same reference numeral with different subscripts. Furthermore, when it is not necessary to distinguish between these multiple components, the subscripts may be omitted in the description.

[0012] In embodiments, processing performed by executing a program may be described. Here, the computer executes the program using a processor (e.g., CPU, GPU) and performs processing defined by the program using memory resources (e.g., memory) and interface devices (e.g., communication ports). Therefore, the main entity performing the processing by executing the program may be the processor. Similarly, the main entity performing the processing by executing the program may be a controller, device, system, computer, or node having a processor.

[0013] The main component of the processing performed by executing the program can be an arithmetic unit, and may include dedicated circuits for specific processing. Here, dedicated circuits include, for example, FPGAs (Field Programmable Gate Arrays), ASICs (Application Specific Integrated Circuits), and CPLDs (Complex Programmable Logic Devices).

[0014] The program may be installed on the computer from the program source. The program source may be, for example, a program distribution server or a storage medium readable by the computer. If the program source is a program distribution server, the program distribution server includes a processor and storage resources for storing the program to be distributed, and the processor of the program distribution server may distribute the program to other computers. In addition, in some embodiments, two or more programs may be implemented as a single program, or one program may be implemented as two or more programs.

[0015] In the following descriptions of embodiments of the present invention, we will mainly describe devices and methods for predicting the number of meals consumed in restaurants and the like located in buildings. However, the prediction target is not limited to the number of meals consumed in restaurants and the like. For example, the present invention can also be applied to predicting the demand (number of items provided and sold) for various goods in stores and facilities that provide and sell food products, clothing, miscellaneous goods, etc. [Examples]

[0016] Office buildings and commercial buildings may have dining facilities such as employee cafeterias (hereinafter referred to as "cafeterias") and facilities that provide and sell food, clothing, miscellaneous goods (hereinafter referred to as "shops") for employees and staff of companies and shops located in the building, as well as workers and cleaners who perform various tasks and cleaning within the building (hereinafter referred to as "employees"), and for visitors to the building (customers, related parties, contractors, etc.) and, in the case of buildings with attached residences, residents (hereinafter referred to as "occupants"). Below, we will explain using cafeterias as an example among such facilities (cafeterias and shops).

[0017] For example, in a cafeteria, the number of meals to be served (hereinafter referred to as "meal consumption") is predetermined to ensure that there is no shortage of meals during the busiest lunch hours, and ingredients are procured and meals are prepared accordingly. However, since cafeterias offer multiple types of meals, such as set meals, noodles, and curry rice (hereinafter referred to as "menus"), it is not easy to determine the number of meals consumed for each menu item in advance. Furthermore, due to the increase in teleworkers resulting from recent work-style reforms, the number of employees coming to the office fluctuates significantly from day to day, making it difficult to estimate the number of cafeteria users. In addition, from the perspective of reducing food waste, the need and importance of highly accurate meal consumption forecasts is increasing. Therefore, in the first embodiment, an example of predicting the number of meals consumed for each menu item offered in a cafeteria will be described. In the following explanation, cafeteria users will be described as employees, but the same applies even if cafeteria users include non-employees and other occupants (i.e., if all instances of "employees" in the following explanation are replaced with "occupants"). Furthermore, as stated above, while the menu is described as a type of meal, it may also include (or even be described as) types of goods that can be offered or sold in the store, such as clothing and miscellaneous goods. The same applies to the descriptions of other embodiments after the first embodiment.

[0018] Figure 1 is a conceptual diagram illustrating the concept of predicting the number of meals consumed in the first embodiment. As an example, Figure 1 shows an example of a three-story office building. Building 1 has a cafeteria 5 on the first floor, and workplaces 6 for employees, such as companies and shops, on the second and third floors. In addition, the basement floor contains a parking lot, an electrical room, and offices for various workers and cleaning staff. Building 1 is equipped with an elevator 2, and employees use the elevator 2 to enter workplaces 6 and offices on each floor, and to exit workplaces 6 and move to other floors. The elevator 2 has a car 3 for employees to board and alight, and the car 3 is equipped with a car weight sensor 4.

[0019] The car weight sensor 4 will be explained using Figures 30A and 30B. Figures 30A and 30B show an example of the configuration of the car weight sensor in the first embodiment. First, in Figure 30A, the car weight sensor 4 includes a sensor unit 301. The sensor unit 301 constantly measures the weight of workers getting on and off the car 3 (for example, at 1-second intervals, etc.) (hereinafter, the measured weight is referred to as the weight measurement value 302). The car weight sensor 4 further includes a weight increase / decrease calculation unit 303. The weight measurement value 302 measured by the sensor unit 301 is input to the weight increase / decrease calculation unit 303, and the weight increase / decrease for each floor (hereinafter, referred to as floor or floor) is calculated and output at predetermined time intervals (hereinafter, the weight increase / decrease for each floor at predetermined time intervals is referred to as elevator weight data 101). By accumulating this elevator weight data 101 for a predetermined period, it becomes possible to calculate the total weight (hereinafter referred to as the floor accumulated weight 102) of all employees who disembark on each floor and are present on each floor (in their respective workplaces 6 on each floor). Note that the above calculation of weight increases and decreases may also be performed by an information processing device 304 connected to the elevator car weight sensor 4 of the elevator 2's control device, as shown in Figure 30B. In this case, the weight measurement value 302 from the elevator car weight sensor 4 is input to the information processing device 304, which then calculates and outputs the elevator weight data 101.

[0020] Returning to the explanation of Figure 1, employees on each floor have preferences depending on the company and their work. For example, in the example in Figure 1, employees on the second floor tend to prefer fish dishes, while employees on the third floor tend to prefer meat dishes. Therefore, in the first embodiment, the number of meals consumed for each menu item (meat-based or fish-based set meals, etc.) offered in the cafeteria 5 is predicted based on the total floor weight for each floor and the preferences of the employees on each floor. For example, in the example in Figure 1, the total floor weight for the third floor is greater than that for the second floor (i.e., there are more employees in the workplace 6 on the third floor than on the second floor), and employees on the third floor tend to prefer meat dishes. Therefore, the number of meals consumed for meat-based menu items and the number of meals consumed for fish-based menu items offered in the cafeteria 5 is predicted to be higher.

[0021] Figure 2 shows an example of the configuration of the meal count prediction device of the first embodiment. In Figure 2, the meal count prediction device 10 includes a floor cumulative weight calculation unit 11, a menu feature extraction unit 12, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14.

[0022] The meal count prediction device 10 in this embodiment is implemented by an information processing device such as a personal computer (PC) or server installed in the dining hall 5 or in another location (such as the dining hall operator's or dining hall employee's office). The information processing device typically includes a CPU, memory, secondary storage device, input means such as a keyboard, mouse, or touch panel display, and output means such as a display or printer. The floor cumulative weight calculation unit 11, menu feature extraction unit 12, and meal count prediction unit 14 shown in Figure 2 are implemented by a floor cumulative weight calculation program, a menu feature extraction program, and a meal count prediction program stored in memory, respectively. The floor-specific menu preference storage unit 13 is implemented by a secondary storage device.

[0023] Figure 3 is a diagram showing an example of a flowchart of the meal count prediction process by the meal count prediction device 10 of the first embodiment. The operation of each component of the meal count prediction device 10 shown in Figure 2 will be explained using Figure 3. In Figure 3, the floor cumulative weight calculation unit 11 acquires (inputs) elevator weight data 101 for the prediction target day, which is output from the car weight sensor 4, as described above (S1001). The floor cumulative weight calculation unit 11 uses the acquired elevator weight data 101 to calculate and output the floor cumulative weight 102 for each floor over a predetermined period (S1002). The menu feature extraction unit 12 acquires (inputs) menu data 103 related to various menus offered in the cafeteria 5 on the prediction target day (S1003). The menu feature extraction unit 12 extracts menu features 104 for each menu from the acquired menu data 103, converts them numerically, and outputs them (S1004). The floor-specific menu preference storage unit 13 stores the preference levels of employees on each floor for each menu item's menu feature quantity 104 (hereinafter referred to as floor-specific menu preference 105). The meal count prediction unit 14 acquires (inputs) the floor cumulative weight 102, menu feature quantity 104, and floor-specific menu preference 105 for a predetermined period, and based on these, predicts the number of meals consumed for each menu item on the target day, and outputs the predicted meal count 110 (S1005). As will be described later, the predicted meal count 110 is displayed, for example, on a display provided by the above-mentioned information processing device, or on a display or monitor installed in a room such as a kitchen where cafeteria staff perform preparation work and cooking.

[0024] Figure 4 is a table showing an example of the elevator weight data 101 and floor cumulative weight 102 in the first embodiment. Figure 4 shows an example in which the elevator weight data 101 is input from the car weight sensor 4 to the floor cumulative weight calculation unit 11 at 5-minute intervals and accumulated.

[0025] As shown in Figure 4, the elevator weight data 101 includes timestamp, floor, and weight increase / decrease data. The timestamp indicates the date and time when the weight increase / decrease for each floor was calculated. In this example, the first timestamp is recorded from 8:00, the opening time of Building 1 on the prediction target date 2024 / 9 / 2, to 8:05, a predetermined time interval of 5 minutes, and thereafter, the date and time are recorded at 5-minute intervals. The floor indicates the number of floors from the basement 1st floor (B1F) to the 3rd floor (3F) of Building 1. The weight increase / decrease, as described above, is the value calculated for each floor based on the weight of the workers who boarded and alighted from elevator car 3 of elevator 2 at the date and time indicated by the timestamp. As shown in Figure 4, the increase / decrease in weight is calculated by taking the weight of employees who disembarked from car 3 (i.e., employees who will be in workplace 6) as a positive (+) weight and the weight of employees who boarded car 3 (i.e., employees who left workplace 6) as a negative (-) weight for each floor at the time indicated by each timestamp, and summing them up. This increase / decrease in weight allows us to determine the total weight of all employees present on each floor (in workplace 6 on each floor) at each time. Note that while Figure 4 shows weight (kg) as an example unit for the increase / decrease in weight, it may also be a percentage (%) of the rated weight of elevator 2, or any unit obtained by multiplying a percentage or weight by a predetermined constant.

[0026] On the other hand, the floor cumulative weight 102 represents the sum of all increases and decreases in weight for each floor prior to the timestamp indicated by that timestamp in the elevator weight data 101 (also called the total value or gross weight). In the example in Figure 4, for the first timestamp, "2024 / 9 / 2 08:00~08:05", there is no record of the time before that, so the floor cumulative weight 102 is the increase and decrease in weight for each floor at that timestamp. For the next timestamp, "2024 / 9 / 2 08:05~08:10", the floor cumulative weight 102 is the sum of the increase and decrease in weight for "2024 / 9 / 2 08:00~08:05" and the increase and decrease in weight for "2024 / 9 / 2 08:05~08:10". In this way, the floor cumulative weight 102 is calculated by the floor cumulative weight calculation unit 11 for each timestamp in the elevator weight data 101, and the floor cumulative weight 102 after a predetermined period has elapsed is output from the floor cumulative weight calculation unit 11. Here, the predetermined period can be any period, such as the first hour from 8:00, the opening time of building 1, or until the start of lunch break (for example, 12:00), and is predetermined and set in the floor cumulative weight calculation unit 11.

[0027] In the above explanation, it was stated that elevator weight data 101 is input from the car weight sensor 4 to the floor cumulative weight calculation unit 11 at predetermined time intervals (5-minute intervals in the example of Figure 4) and accumulated. However, elevator weight data 101 may be input to the floor cumulative weight calculation unit 11 at any interval or timing, without being limited to predetermined time intervals. For example, when a predetermined period has elapsed, the elevator weight data 101 for the predetermined period may be input to the floor cumulative weight calculation unit 11. In this case, the floor cumulative weight calculation unit 11 calculates and outputs the floor cumulative weight 102 by accumulating the increase / decrease in weight for each floor at all timestamps during the predetermined period. As for other timings, for example, elevator weight data 101 may be input to the floor cumulative weight calculation unit 11 as a trigger when the increase / decrease in weight exceeds a certain threshold, or when the number of times the elevator doors are opened and closed exceeds a certain threshold. This reduces the frequency of inputting elevator weight data 101 during times when elevator usage is low, and as a result, it reduces the communication load between the car weight sensor 4 and the floor cumulative weight calculation unit 11, as well as the processing load of the floor cumulative weight calculation unit 11.

[0028] Furthermore, while the above explanation states that the floor cumulative weight calculation unit 11 calculates and outputs the floor cumulative weight 102 for each floor over a predetermined period using elevator weight data 101, the floor cumulative weight calculation unit 11 can also use the floor cumulative weight 102 calculated from the elevator weight data 101 to estimate and output, for example, the number of people present on each floor (hereinafter referred to as the number of people on each floor). In this case, the number of people on each floor is an estimated value and is not necessarily an absolute value, but is output as an indicator of the number of people present on each floor. The floor cumulative weight calculation unit 11 may also be equipped with a floor cumulative weight output unit that outputs the floor cumulative weight 102 and a number of people on each floor output unit that outputs the number of people on each floor. In addition, the number of people on each floor may be obtained by sensors or access control devices, but by using elevator weight data 101, it becomes possible to estimate the number of people on each floor without introducing equipment such as gate devices.

[0029] Figure 5 is a table showing an example of the menu data 103 and menu features 104 in the first embodiment. As shown in Figure 5, the menu data 103 includes date, menu name, menu image, and calorie data. The date indicates the date on which each menu is served. The menu name indicates the name of each menu offered by the cafeteria 5. The menu image indicates an image of each menu. The calorie data indicates the calorie intake of each menu. In addition, the menu data 103 may also include other data, such as nutritional components (each nutrient (component) and value (amount)) contained in each menu, and data such as the weight and price of each menu. Note that individual menu items such as chicken cutlet and grilled mackerel may also be served as part of a set meal. Each of these data items is pre-input (registered) by the cafeteria operator or staff, for example, using the input means to the information processing device described above, and stored in a secondary storage device. Alternatively, the menu data 103 may be pre-entered (registered) on an information processing terminal (not shown) owned by the cafeteria operator or staff, such as a PC, tablet, or smartphone, and output from that terminal, separate from the information processing device described above. The menu data 103 may be registered in batches, for example, for a month or a week, by the first day of the month or week, or the menu for the day may be registered daily.

[0030] Menu Feature 104 represents the characteristics, properties, classifications, etc., of each menu item. For example, in the example in Figure 5, Menu Feature 104 shows the dish category (meat, fish, ramen, curry, etc.) for each menu item. In addition to dish categories, Menu Feature 104 may also include dish groups (staple food, main dish, side dish, fruit, salad, etc.) and nutritional balance categories (low carbohydrate, high protein, high fiber, etc.). Representative examples of individual categories and groups (meat, fish, ramen, curry, etc. in the dish category) are predetermined. The level of detail for individual categories and groups is arbitrary. For example, the dish category for meat could be defined as beef, pork, chicken, etc., and the dish category for fish could be defined as saltwater fish, freshwater fish, etc. Alternatively, the dish category could be defined as halal cuisine, vegan cuisine, low-allergen cuisine, etc., based on religious beliefs or allergies.

[0031] The menu feature extraction unit 12 is pre-configured to determine whether to use a dish category, dish group, nutritional balance category, etc., as the menu feature 104, and which of these categories or groups to use. The menu feature extraction unit 12 extracts, selects, classifies, or determines the menu feature 104 for each menu in the input menu data 103 according to this configuration. For example, as shown in the example in Figure 5, if the menu feature is set to use a dish category as the menu feature 104, the menu feature extraction unit 12 extracts words contained in each menu name in the input menu data 103 (for example, "chicken" in "chicken cutlet"), selects the dish category to which that word belongs (for example, "meat"), or classifies each menu into the dish category to which the extracted word belongs, thereby determining the menu feature 104 for each menu (details of the methods for extracting and determining the menu feature 104 by the menu feature extraction unit 12 will be described later).

[0032] Furthermore, the menu feature extraction unit 12 quantifies the menu features 104 for each menu determined as described above and outputs them as menu feature data (hereinafter, the quantified menu feature data is also referred to as menu feature 104). Various known methods can be used for quantification. For example, the menu feature extraction unit 12 may simply quantify the menu features 104 for each menu by assigning a number, or it may use a conversion method using dummy variables called One-Hot encoding to quantify the menu features 104 for each menu (the explanation of the One-Hot encoding method is omitted).

[0033] Figure 6 is a table showing an example of the floor-specific menu preference score 105 in the first embodiment. Figure 6 shows an example where the food category is used as the menu feature 104. As shown in Figure 6, the floor-specific menu preference score 105 includes the floor, the food category as the menu feature 104, and the menu feature 104 for each employee on each floor, i.e., the preference score for each food category. The preference score is a numerical representation of the employee's preference tendency, that is, the degree to which the employee prefers to select menus related to (or classified as) individual food categories, and a higher number indicates a stronger preference tendency. For example, in the example in Figure 6, the floor-specific menu preference score 105 shows that employees on the 3rd floor have a strong tendency to select meat-based menus related to (or classified as) the "meat" food category, while they do not tend to select noodle-based menus classified as "ramen" or "udon" very often.

[0034] The preference levels for each menu feature 104 of each floor's employees in the floor-specific menu preference score 105 are determined in advance as approximate values ​​by the cafeteria operator or staff, based on various parameters such as the company and job content of the employees on each floor, average weight, gender ratio, and age ratio. The determined preference levels for each menu feature 104 of each floor's employees are input (registered) into the information processing device described above and stored in the floor-specific menu preference score storage unit 13 (secondary storage device) as the floor-specific menu preference score 105. At this time, the menu feature 104 and floor data in the floor-specific menu preference score 105 are stored in the same way as above, that is, by assigning numbers to each menu feature 104 and each floor, or by storing them in a numerical state using one-hot encoding.

[0035] In the above explanation, the preference scores in the floor-specific menu preference score 105 were described as the preference scores for each menu feature 104 for each employee on each floor. However, it is not limited to employees; it could also be the preference score at the floor level or tenant level. The preference scores for each menu feature 104 at the floor level can be obtained, for example, by using the average or median preference scores of employees on each floor, or by looking at the sales history of each menu item during lunchtime on that floor.

[0036] Furthermore, if payment data (POS data, etc.) from the payment system and ticket issuance system installed in Cafeteria 5 is linked to identification data (ID data, etc.) from building access cards or employee ID cards owned by employees, the eating trends for each floor may be understood from past eating history based on this data, and the preference level may be determined according to those eating trends. If the cafeteria operator also operates cafeterias in other buildings, the preference level in Building 1 may be determined based on (and applied to) the results and experience from the cafeterias in other buildings. Alternatively, the cafeteria operator may conduct a survey of employees or cafeteria users in Building 1 and determine the preference level from the results. Also, although Figure 6 shows an example where the preference level is expressed as numerical data, the preference level may be expressed as a binary value (e.g., "True", "False", etc.) or a qualitative expression (e.g., "Like", "Neutral", "Dislike", etc.). (In the case of a qualitative expression, it will be quantified using the known methods described above.) Figure 7 is a table showing an example of the floor cumulative weight 102, menu features 104, and floor-specific menu preference 105 input to the meal count prediction unit 14, and the predicted number of meals consumed 110 predicted by the meal count prediction unit 14. In addition, to facilitate understanding of the contents of the table, Figure 7 also shows the date and menu name of the prediction target day in the menu data 103, and the menu features 104 are also shown in their pre-quantified state. As described above, the meal count prediction unit 14 acquires (inputs) the floor cumulative weight 102, menu features 104, and floor-specific menu preference 105 for a predetermined period, predicts the number of meals consumed for each menu on the prediction target day based on these, and outputs the predicted number of meals consumed 110. In Figure 7, the floor cumulative weight 102, menu features 104, and floor-specific menu preference 105 for the predetermined period are as described above. The predicted number of meals consumed 110 includes data for the predicted number of meals consumed per floor and the predicted number of meals consumed for all floors. The floor-specific predicted consumption figures represent the estimated number of meals consumed by staff on each floor for each menu item, while the total floor-specific predicted consumption figures represent the estimated number of meals consumed by staff on all floors for each menu item (the sum of the floor-specific predicted consumption figures for each menu item).

[0037] Here, the menu feature 104 (dish category in the example in Figure 7) is k, the floor is n, and the floor-specific menu preference 105 is A. k,n Floor total weight 102 W n If the predicted number of meals consumed per floor is p(k,n), the predicted number of meals consumed across all floors is P(k), and the meal weight coefficient is Z, then the predicted number of meals consumed per floor p(k,n) and the predicted number of meals consumed across all floors P(k) are calculated by the following equations (Equation 1) and (Equation 2).

[0038]

number

[0039]

number

[0040] In (Equation 2), the food consumption weight coefficient Z represents the number of meals consumed per unit of load weight. For example, if the unit of load weight is kg, the food consumption weight coefficient Z can be set to 65 kg, which is the average weight of Japanese people. Note that the food consumption weight coefficient Z does not need to be the same value for all floors; different values ​​(Z) can be used for each floor. n ) may be set. In this case, Z in (Equation 1) is Z n This is the result.

[0041] For example, in Figure 7, if we assign category numbers 1 to 3 to each of the menu feature categories "meat," "fish," and "ramen," and assign floor numbers 1 to 4 to each of the floors from B1F to 3F of Building 1, and expand (Equation 1) and (Equation 2) above with category number k and floor number n, and substitute the expanded (Equation 1) into (Equation 2), the formula for calculating the predicted number of meals consumed P(k) across all floors can be expressed as (Equation 3) below.

[0042]

number

[0043] By substituting the values ​​for floor-specific menu preference 105 and floor cumulative weight 102 shown in Figure 7 into (Equation 3), and setting the consumption weight coefficient Z = 65, the predicted number of meals consumed per floor is calculated, yielding the predicted number of meals consumed per floor in Figure 7. Summing the predicted number of meals consumed per floor for each food category yields the total predicted number of meals consumed for all floors in Figure 7. In this way, the meal consumption prediction unit 14 predicts (calculates) the number of meals consumed for each menu item on the target day. Note that while the consumption weight coefficient Z was used here, when predicting the number of meals consumed using the number of people per floor, the consumption weight coefficient is replaced by the number of meals consumed per person, so Z = 1.

[0044] The meal count prediction unit 14 preferably makes its prediction of meal counts before the opening time of the cafeteria 5 (for example, 12:00, the same as the start of lunch break), and more preferably before or during the preparation stage so that the predicted values ​​can be reflected in the preparation work for each menu item that the cafeteria 5 will offer that day. As described above, the meal count prediction unit 14 calculates the predicted meal counts for each floor and for all floors using the predetermined floor-specific menu preference scores 105 and the total floor weight 102 for the day. Therefore, in order to predict meal counts, it is necessary to predict the total floor weight 102 at the opening time of the cafeteria 5 with as much accuracy as possible.

[0045] Here, as is typical for office buildings, the floor weight 102 changes according to the passage of time after the building opens (e.g., 8:00). Figure 8 is a graph showing an example of how the floor weight 102 changes according to the time. In Figure 8, for example, on a floor where workplace 6 is located, the floor weight 102 increases significantly between the building's opening and the start of work (e.g., 9:00) because many employees arrive and use the elevator. After the start of work, the number of people arriving decreases, so the increase in floor weight 102 becomes more gradual. Then, around 11:00, most employees have finished arriving, so the change in floor weight 102 becomes smaller. In this way, the floor weight 102 changes according to the time, reflecting the characteristics of users on each floor.

[0046] Similar to the example of FIG. 7, assuming the floor is n, the date is d, and the time is T, and the floor integrated weight 102 for each floor is W n , the floor integrated weight W n with respect to the time change is f n (T). Then, the function f n (T) can be determined, for example, by averaging the time progression of the floor integrated weight 102 for each floor on the same date in the past (or the corresponding day (e.g., the first Monday in September, etc.)). At this time, the floor integrated weight W n can be expressed by the following (Equation 4) as a function of time T. Note that W n,d in (Equation 4) is the coefficient for the current day for each floor and is a parameter that reflects the number of people present on the prediction target day indicated by the date d. In the case of an office building, since the number of people present varies by day, the coefficient for the current day also changes by day.

[0047]

Equation

[0048] By introducing the function f n [[ID=26]](T), the eating number prediction unit 14 can predict the floor integrated weight W n at future times for each floor. For example, when the floor integrated weight W n = 1100 kg at 9:00 on the nth floor on the date d, substituting this value into (Equation 4) gives 1100 = W n,d × f n (9:00). Here, as described above, since the function f n (T) can be determined based on the floor integrated weight 102 on the same date (or the corresponding day) in the past, f n (9:00) is uniquely determined. For example, if f n (9:00) = 1000, then W n,d = 1.1, and the coefficient for the current day on the nth floor on the date d can be calculated. In this case, since the coefficient for the current day is greater than 1, it can be interpreted that more people are present on this date d than on a normal date.

[0049] By using the calculated coefficient for the current day, the floor integrated weight W at future times can be obtained.n This is predictable. For example, the total floor weight W of the nth floor at the opening time of cafeteria 5, 12:00. n is, f n If (12:00) = 3000, then W n =1.1 × 3000 = 3300 is predicted. In this way, the meal consumption prediction unit 14 calculates the floor cumulative weight W for each floor at an early time. n Using this, the cumulative floor weight W for each floor at a future time on the target date is calculated. n This makes it possible to predict the number of meals consumed, and using this predicted value, it is possible to calculate the predicted number of meals consumed per floor and for all floors combined (predict the number of meals consumed).

[0050] Figures 9A and 9B show examples of display screens for the predicted number of meals consumed 110, which are predicted (calculated) and output by the meal consumption prediction unit 14 as described above. The predicted number of meals consumed 110 output from the meal consumption prediction unit 14 is displayed on a display or monitor installed in a room such as a kitchen where cafeteria staff perform preparation work. Figure 9A shows an example of a display screen for the predicted number of meals consumed 110 at 9:00 on the prediction target day, 2024 / 9 / 2. In the example display screen of Figure 9A, the predicted number of meals consumed 110 at 9:00 is shown for each menu item offered on that day. Figure 9B shows an example of a display screen for the predicted number of meals consumed 110 at 10:00 on the same day. In the example display screen of Figure 9B, in addition to the predicted number of meals consumed 110 for each menu item at 10:00, the difference from the predicted number of meals consumed 110 at 9:00 is shown in parentheses. Since the number of meals consumed can be predicted at any given time, the difference from the previous prediction can also be displayed, as shown in Figure 9B. As shown in Figures 9A and 9B, factors that may cause a difference in the predicted number of meals consumed (110) include, for example, a higher-than-usual number of people coming to Building 1 between 9:00 and 10:00. By displaying the predicted number of meals consumed (110) to cafeteria staff and others performing preparation work, it can be used as reference information to adjust the amount of food prepared.

[0051] In the above explanation, the meal count prediction unit 14 predicted the number of meals consumed for each menu item on the target day and outputted the predicted number of meals consumed 110. However, the meal count prediction unit 14 also functions as a demand forecasting unit that predicts the demand for items other than meals, and can predict the demand (number of items to be served, etc.) for each menu item and output the predicted demand.

[0052] In the description of the first embodiment above, an example was shown in which the meal count prediction device 10 is implemented by an information processing device installed in Building 1. However, the invention is not limited to this, and the meal count prediction device 10 may also be implemented by a cloud server (or a virtual server hosted on a computer or cloud server) connected to various devices, equipment, terminals, etc. in Building 1 via a network such as the Internet. The cloud server also includes a CPU, memory, secondary storage device, etc., similar to the information processing device described above. The floor-specific menu preference scores 105 are stored in the secondary storage device of the cloud server or in a storage device separate from the cloud server, and the cloud server executes the floor cumulative weight calculation program, menu feature extraction program, and meal count prediction program in memory to realize the configurations of the floor cumulative weight calculation unit 11, the menu feature extraction unit 12, and the meal count prediction unit 14. The cloud server may also be remotely operated from an input / output device connected via the network. Therefore, unlike the information processing device described above, the cloud server does not need to have input means or output means.

[0053] When the meal count prediction device 10 is implemented using a cloud server, elevator weight data 101 is input to the cloud server via the network from the car weight sensor 4 or the information processing device 304 connected to the car weight sensor 4. Menu data 103 is input to the cloud server via the network from information processing terminals owned by cafeteria operators or employees. Meanwhile, the predicted meal count 110 output by the meal count prediction unit 14 is input to an information processing terminal via the network and displayed on the information processing terminal's display, or displayed on other displays or monitors via the information processing terminal. The implementation example of the meal count prediction device is the same for the meal count prediction device in other embodiments described later.

[0054] As explained above, the meal consumption prediction device in the first embodiment makes it possible to predict the number of meals consumed in cafeterias, etc., without the need to install gate devices or IC card equipment for building entry and exit management in the building, by using the floor cumulative weight calculated from elevator weight data. Furthermore, by using floor-specific menu preference scores that show the preference trends of employees on each floor, the meal consumption prediction device can predict the number of meals consumed for each menu on the target day, reflecting the preferences of employees on each floor. Moreover, by predicting the floor cumulative weight at a future time using the floor cumulative weight at an early time on the target day, the meal consumption prediction device can predict the number of meals consumed for each menu at any future time using this predicted value, and this makes it possible to reflect the predicted number of meals consumed in the preparation work for each menu. [Examples]

[0055] In the first embodiment, an example was described in which menu preferences for each floor, predetermined by the cafeteria operator or staff, and stored in the menu preference storage unit, are used by the meal count prediction unit to predict the number of meals consumed. However, even if menu preferences for each floor predetermined by the cafeteria operator or staff are used in the initial stage, it is also possible to update and revise the menu preferences for each floor using the cafeteria's operational record, that is, the consumption record of various menus offered in the cafeteria. Therefore, in the second embodiment, an example of a meal count prediction device with such a configuration will be described. Note that in the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be described.

[0056] Figure 10 shows an example of the configuration of the meal count prediction device of the second embodiment, and the same reference numerals are used for components identical to those of the meal count prediction device 10 shown in Figure 2. In Figure 10, the meal count prediction device 20 includes a floor cumulative weight calculation unit 11, a menu feature extraction unit 12, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, similar to the meal count prediction device 10 shown in Figure 2, and further includes a learning unit 15 for updating the floor-specific menu preference 105.

[0057] Figure 11 is a diagram showing an example of a flowchart of the learning and updating process of floor-specific menu preference scores 105 by the learning unit 15 in the meal consumption prediction device 20 of the second embodiment. Figure 11 shows an example of learning and updating floor-specific menu preference scores 105 based on new data for one day, which is the learning target day. The operation of the learning unit 15 will be explained using Figure 11. In Figure 11, the learning unit 15 acquires (inputs) the floor cumulative weight 102 for a predetermined period on the learning target day from the floor cumulative weight calculation unit 11 (S2001). The predetermined period is, for example, the period from 8:00, the opening time of building 1, to 12:00, the opening time of cafeteria 5. The learning unit 15 acquires (inputs) menu feature quantities 104 related to the menus served on the learning target day from the menu feature quantity extraction unit 12 (S2002). The learning unit 15 obtains (inputs) the predicted number of meals consumed 110 (in this case, the predicted number of meals consumed on all floors) for the learning target day from the meal consumption prediction unit 14 (S2003). Furthermore, the learning unit 15 obtains (inputs) the actual number of meals consumed 106 for the learning target day (S2004). The actual number of meals consumed is the actual number of meals consumed for each menu item in cafeteria 5 on the learning target day. The actual number of meals consumed 106 may be obtained, for example, from payment data (POS data, etc.) in cafeteria 5, or from information processing terminals owned by cafeteria operators or employees. The learning unit 15 calculates the floor-specific menu preference update value 107 for each menu feature using the floor cumulative weight 102, menu features 104, predicted number of meals consumed 110, and actual number of meals consumed 106 for a predetermined period on the learning target day (S2005). The learning unit 15 updates the floor-specific menu preference 105 stored in the floor-specific menu preference storage unit 13 using the calculated floor-specific menu preference update value 107 (S2006).

[0058] Figure 12 is a table showing an example of floor-specific menu preference update values ​​107 calculated by the learning unit 15, including the floor cumulative weight 102, menu features 104, predicted number of meals consumed 110, actual number of meals consumed 106, the predicted difference (the difference between the predicted number of meals consumed 110 and the actual number of meals consumed 106), and the menu features 104 used for menu features, all of which are input to the learning unit 15. Figure 12 shows an example where the food category is used as the menu feature 104. In addition, to facilitate understanding of the contents of the table, Figure 12 also shows the date of the learning unit and the menu name, as well as the menu features 104 and floors in their pre-quantified state. As described above, the learning unit 15 acquires (inputs) the floor cumulative weight 102, menu features 104, predicted number of meals consumed 110, and actual number of meals consumed 106 for a predetermined period of the learning unit, and calculates floor-specific menu preference update values ​​107 for each menu feature. In Figure 12, the floor cumulative weight 102, menu features 104, and predicted number of meals consumed 110 (predicted number of meals consumed across all floors) for the specified period are the same as in the example shown in Figure 7, including the date and menu name. The actual number of meals consumed 106 and the predicted difference are as described above.

[0059] Here, similar to the example shown in Figure 7, the menu feature 104 (dish category in the example in Figure 12) is k, the floor is n, and the floor cumulative weight 102 is W. n P(k) is the predicted number of meals consumed (predicted number of meals consumed on all floors) of 110, R(k) is the actual number of meals consumed of 106, and Δp is the difference between the prediction and the actual number of meals consumed. k Floor-specific menu preference update value 107 ΔA k,n In that case, the predicted difference ΔP k Floor-specific menu preference update value: 107ΔA k,n The following equations (Equation 5) and (Equation 6) are used to calculate the value.

[0060]

number

[0061]

number

[0062] As shown in (Equation 5), the predicted difference is the value (difference) obtained by subtracting the predicted number of meals consumed (110) from the actual number of meals consumed (106). For example, Figure 12 shows that for chicken cutlet, whose menu feature 104 is in the "meat" category, the actual number of meals consumed (106) was 1.3 meals less than the predicted number of meals consumed (110). Therefore, in order to correct this prediction error, the learning unit 15 uses (Equation 6) to apportion the predicted difference across each floor using the floor cumulative weight 102, and calculates the floor-specific menu preference update value 107.

[0063] As described above, the learning unit 15 updates the floor-specific menu preference 105 stored in the floor-specific menu preference storage unit 13 using the calculated floor-specific menu preference update values ​​107. One method for updating the floor-specific menu preference 105 is to add each value of the floor-specific menu preference update value 107 to each corresponding value of the floor-specific menu preference 105 (for example, in the example in Figure 12, the floor-specific menu preference update value 107 for each floor where the dish category is "meat," which is a menu feature quantity 104, is similarly added to the floor-specific menu preference 105 for each floor where the dish category is "meat") and replace each value of the floor-specific menu preference 105 with the resulting value. In addition to this method, other methods include adding the floor-specific menu preference update value 107 to the floor-specific menu preference 105 and using the average of the values ​​of the original floor-specific menu preference 105 as the new value for floor-specific menu preference 105, or similarly, using an arbitrary percentage (for example, around 30%) of the difference between the values ​​added to the floor-specific menu preference update value 107 and the values ​​of the original floor-specific menu preference 105 as the new value for floor-specific menu preference 105. Thus, any method of updating the floor-specific menu preference 105 is acceptable, and any update method should be appropriately adopted (decided) and set in the learning unit 15 in advance.

[0064] The learning and updating process of the floor-specific menu preference scores 105 by the learning unit 15 may be executed at any time (date and time) as long as it does not interfere with the meal count prediction by the meal count prediction unit 14. For example, as shown in the examples in Figures 11 and 12, if the floor-specific menu preference scores 105 are learned and updated based on new data for one day, the learning and updating process may be executed after the cafeteria 5 has closed for business on the learning day, using the floor cumulative weight 102, menu features 104, predicted meal count 110, and actual meal count 106 for that day, or it may be executed on a day after that, for example, when the cafeteria 5 is closed. Furthermore, the learning day is not limited to one day, but may be any multiple days, or any period such as one week, one month, six months, or one year. If the learning day is multiple days or a period, the learning and updating process may be executed at any time (date and time) after those days or periods have elapsed, using data from multiple days or periods. Furthermore, the learning target dates can include calendar information such as months and seasons. This makes it possible to predict the number of meals consumed, reflecting menu preferences that may change according to the season. In addition, even for items other than food, it becomes possible to predict demand, such as predicting that autumn clothes will sell well at the change of season, such as in September, or that wrapping supplies will sell well before events such as Christmas, for example.

[0065] In the description of the second embodiment above, as an example of learning and updating the floor-specific menu preference score 105 by the learning unit 15, the learning unit 15 outputs a floor-specific menu preference score update value 107 using the above-mentioned (Equation 5) and (Equation 6), and updates the floor-specific menu preference score 105. However, the learning and updating method is not limited to this, and various learning methods may be used. For example, the learning unit 15 may learn past data of the cafeteria 5 (floor cumulative weight 102, menu features 104, predicted number of meals consumed 110, actual number of meals consumed 106, etc.) using various machine learning methods such as logistic regression, support vector machines, and neural networks, and output a floor-specific menu preference score update value 107 or floor-specific menu preference score 105.

[0066] As explained above, the meal consumption prediction device in the second embodiment, by newly incorporating a learning unit, makes it possible to update the menu preference levels for each floor as needed. As a result, in addition to the same effects as in the first embodiment, the preference levels for each menu feature of staff members on each floor will always be values ​​that reflect the actual situation, making it possible to improve the accuracy of meal consumption prediction. [Examples]

[0067] In the first embodiment, when the system is configured to use food categories as menu features, an example was described in which the menu feature extraction unit determines menu features by selecting food categories associated with words extracted from each menu name. In the third embodiment, similar to the first embodiment, when the system is configured to use food categories as menu features, an example of a specific configuration and method for the menu feature extraction unit to determine menu features will be described. Note that in the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be described.

[0068] Figure 13 shows an example of the configuration of the third embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those of the meal count prediction device 10 shown in Figure 2. In Figure 13, the meal count prediction device 30 includes a floor cumulative weight calculation unit 11, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, similar to the meal count prediction device 10 shown in Figure 2, as well as a menu feature extraction unit 31 similar to the menu feature extraction unit 12 shown in Figure 2. The menu feature extraction unit 31 also includes a feature word classification data creation unit 32, a feature word classification data storage unit 33, and a feature word matching unit 34.

[0069] Figure 14 shows an example of a flowchart of the feature term classification data creation process by the feature term classification data creation unit 32 in the menu feature extraction unit 31 of the third embodiment. Here, feature terms are, for example, terms that represent individual categories within a cuisine category. Examples of feature terms include terms related to ingredients such as "chicken," "pork," "beef," and "vegetable," terms related to cooking methods such as "fry," "grill," "boil," and "steam," as well as terms related to seasoning methods and terms related to food genres (Italian, Chinese, Japanese, etc.). Multiple terms (for example, several dozen to several hundred) related to ingredients and cooking methods are pre-set as feature terms in the feature term classification data creation unit 32. Note that the terms related to ingredients and cooking methods can be arbitrarily determined, including representative, general, or special terms.

[0070] In Figure 14, the feature term classification data creation unit 32 acquires data on the menu names of each menu item previously provided by the cafeteria 5 (S3001). Preferably, the menu name data acquired here is text data showing all menu names for the past several months to several years. The menu name data is acquired from an information processing terminal or the like owned by the cafeteria operator or staff.

[0071] Next, the feature word classification data creation unit 32 extracts menu words such as nouns, verbs, and adjectives from the acquired menu name data (S3002). One method for extracting menu words is to decompose the text related to each menu name contained in the menu name data into parts of speech using morphological analysis and extract only nouns, verbs, or adjectives. Using such a method, for example, the text of the menu name "Grilled Chicken Tomato Sauce" is decomposed into parts of speech (elements) such as "Grill," "Chicken," "Tomato," and "Sauce," and each is extracted as a menu word. This menu word extraction step makes it possible to efficiently analyze past menu name data.

[0072] Next, the feature term classification data creation unit 32 classifies each extracted menu word against pre-set feature terms and registers them as related terms (S3003). Specifically, the feature term classification data creation unit 32 determines which of the pre-set feature terms each extracted menu word belongs to. For example, regarding the menu words "grill," "chicken," "tomato," and "sauce" in the above example, the feature term classification data creation unit 32 determines that "grill" should be classified under the feature term "grill" and registers it as a related term for the feature term "grill." Similarly, the feature term classification data creation unit 32 determines and registers "chicken" as a related term for the feature term "chicken" and "tomato" as a related term for the feature term "vegetable." In this case, for example, for the menu word "sauce," if there is a feature term to which it should be classified (corresponding), it will be registered as a related term for that feature term, but if there is no corresponding feature term, it will not be registered. This feature keyword classification step allows related keywords corresponding to menus previously offered by Cafeteria 5 to be classified and registered under each feature keyword.

[0073] Next, the feature term classification data creation unit 32 calculates (counts) the number of occurrences of each menu word registered as a related term for each feature term in the acquired menu name data (the number of each menu word included in the menu name data). Then, for each feature term, the feature term classification data creation unit 32 calculates the number of occurrences of one or more menu words registered as related terms for that feature term and sums them up to calculate the number of occurrences of that feature term in the menu name data (S3004). This feature term occurrence count calculation step makes it possible to quantitatively grasp and evaluate the frequency with which each pre-set feature term is used in menu names.

[0074] Next, the feature term classification data creation unit 32 registers multiple feature terms (for example, about 10 to 100) and their related terms as feature term classification data, in descending order of frequency of occurrence, based on the calculated number of occurrences of each feature term, and creates feature term classification data 111 (S3005). At this time, the feature term classification data creation unit 32 registers feature terms in the feature term classification data 111 based on criteria such as an occurrence count of 10 or more. Such criteria can be arbitrarily set in advance. There are no restrictions on the number of feature terms registered in the feature term classification data 111; it can be any number. The feature term classification data creation unit 32 stores the created feature term classification data 111 in the feature term classification data storage unit 33.

[0075] Figure 15 is a table showing an example of the characteristic phrase classification data 111 created as described above. In the example in Figure 15, characteristic phrases such as "fry," "pork," and "chicken," and their related phrases (for example, menu words such as "katsu," "age," and "fry" corresponding to "fry") are registered in the characteristic phrase classification data 111 in order of the number of occurrences of each characteristic phrase. By executing the characteristic phrase classification data creation process shown in Figure 14, it is possible to create characteristic phrase classification data 111 that is in line with the menu history that Cafeteria 5 has provided in the past.

[0076] In the above explanation, an example was shown in which the feature term classification data creation unit 32 of the menu feature extraction unit 31 creates the feature term classification data 111. However, for example, the feature term matching unit 34 may create the feature term classification data 111, or other information processing devices other than the meal count prediction device may create the feature term classification data 111. When the feature term classification data 111 is created by another information processing device, the feature term classification data 111 is input to the meal count prediction device 30 from an external source and stored in the feature term classification data storage unit 33. Furthermore, when the feature term matching unit 34 or other information processing devices create the feature term classification data 111, the feature term classification data creation unit 32 is unnecessary.

[0077] Using the feature term classification data 111 created as described above, the feature term matching unit 34 determines the menu features 104 for each menu item from the menu data 103 for the target date. Specifically, the feature term matching unit 34 reads the feature term classification data 111 stored in the feature term classification data storage unit 33 and matches the text data of each menu name in the input menu data 103 with the related terms for each feature term in the feature term classification data 111. That is, it reads each feature term and its corresponding related terms from the feature term classification data 111 and matches and determines whether any of the related terms are included in the text data of the menu name in the menu data 103. For example, if the menu name is the text data "Kurobuta Tonkatsu", the feature term matching unit 34 determines that the text data "Kurobuta Tonkatsu" contains the term "katsu", which is one of the related terms corresponding to the feature term "fry". Similarly, the feature word matching unit 34 determines that the text data "Kurobuta Tonkatsu" contains the words "buta" and "ton," which are related words corresponding to the feature word "pork." On the other hand, the feature word matching unit 34 determines that the text data "Kurobuta Tonkatsu" does not contain any of the related words corresponding to the feature word "chicken." In this way, the feature word matching unit 34 performs a match between all feature words registered in the feature word classification data 111 and the text data of each menu name in the menu data 103, and outputs the matching score 112, which is the matching result, as the menu feature quantity 104.

[0078] Figure 16 is a table showing an example of the matching score 112 as a menu feature 104, using the menu name "Kurobuta Tonkatsu" as an example. It includes the text data of the menu name in the menu data 103 that the feature word matching unit 34 matches, the feature words registered in the feature word classification data 111 and their corresponding related words, and the matching score 112 as a menu feature 104. In Figure 16, the matching score 112 shows the result of matching and determining for each feature word whether the text data of each menu name contains the related words corresponding to each feature word. In the example of Figure 16, the matching results are shown as binary values ​​("True", "False"). That is, as described above, the text data of the menu name "Kurobuta Tonkatsu" contains the related words for the feature words "fry" and "pork", but does not contain the related words for the feature word "chicken". Therefore, the matching result for the feature words "fry" and "pork" is "True", and the matching result for the feature word "chicken" is "False". The feature word matching unit 34 outputs a binary value (i.e., "1,1,0,...") representing the matching result for all feature words registered in the feature word classification data 111 as the matching score 112 for each menu in the menu data 103. This allows the feature word matching unit 34 to output the matching results quantitatively.

[0079] Furthermore, the matching score 112 may be, as described above, not only a binary value representing the matching results for all feature terms registered in the feature term classification data 111, but also the number of related terms included in the text data of the menu name for each feature term. In this case, in the example of the menu name "Kurobuta Tonkatsu" above, the feature term "fry" is "1", the feature term "pork" is "2", and the feature term "chicken" is "0", and the number corresponding to all feature terms (i.e., the number of related terms included in the menu name "1, 2, 0, ...") becomes the matching score 112.

[0080] As explained above, the meal count prediction device in the third embodiment provides the same effects as the first embodiment, but because the menu feature extraction unit includes a feature word classification data creation unit and a feature word matching unit, it is possible to create feature word classification data in which related words corresponding to menus previously provided by Cafeteria 5 are registered, and feature words that have a high frequency of appearance in past menus are registered. Furthermore, the meal count prediction device in the third embodiment makes it possible to make more accurate matches (output of a matching score) by matching each menu name to be provided on the prediction target day with such feature word classification data. [Examples]

[0081] In the third embodiment, an example was described in which the menu feature extraction unit creates feature word classification data and compares it with the text data of each menu name. In the fourth embodiment, similarly, when the system is configured to use the dish category as the menu feature, an example of a specific configuration and method for the menu feature extraction unit to analyze the image of each menu using image analysis technology to determine the dish category will be described. In the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be explained.

[0082] Figure 17 shows an example of the configuration of the fourth embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those of the meal count prediction device 10 shown in Figure 2. In Figure 17, the meal count prediction device 40 includes a floor cumulative weight calculation unit 11, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, similar to the meal count prediction device 10 shown in Figure 2, as well as a menu feature extraction unit 41 similar to the menu feature extraction unit 12 shown in Figure 2. The menu feature extraction unit 41 also includes an image analysis model storage unit 42 and a menu image analysis unit 43. The menu image analysis unit 43 uses the image analysis model 113 stored in the image analysis model storage unit 42 to analyze the menu images of each menu item offered by the cafeteria 5 on the prediction target day, and outputs the analysis score 114, which is the analysis result, as menu features 104.

[0083] Figure 18 is a table showing an example of menu images in menu data 103 analyzed by the menu image analysis unit 43 of the fourth embodiment, and an analysis score 114 as menu feature quantities 104. Figure 18 shows an example of analyzing the menu image of a menu item called "Beef Steak". As shown in Figure 5, the menu data 103 includes menu images. The menu image analysis unit 43 analyzes the menu images of each menu item in the input menu data 103 using the image analysis model 113 read from the image analysis model storage unit 42. In the image analysis by the menu image analysis unit 43, for example, object detection in the image is performed using the image analysis model 113, and the detected objects are identified. In this embodiment, various known models such as CNN, R-CNN, SSD, YOLO, HOG, etc. can be used as the image analysis model (method or algorithm) for performing such object detection.

[0084] By using this image analysis model 113, in the case of the menu image shown in Figure 18, the menu image analysis unit 43 detects two objects in the image. In Figure 18, the regions where each object is detected are shown as bounding boxes. The menu image analysis unit 43 identifies the two detected objects as "beef" and "potato," respectively. In Figure 18, the analysis score 114 shows the identification result of the objects detected in each menu image, and in the example of Figure 18, the identification result is shown as a binary value ("True," "False"). For example, in the example of Figure 18, as described above, the two objects detected in the menu image are identified as "beef" and "potato," so the identification result for "beef" and "potato" in the food category in the analysis score 114 is "True." On the other hand, although not shown, there are other food categories such as "pork" and "chicken," and the identification result for all of them is "False." The menu image analysis unit 43 outputs a binary value showing the identification result for all food categories as the analysis score 114.

[0085] Furthermore, the menu image analysis unit 43 may output the identification results for all pre-set feature words or all feature words registered in the feature word classification data 111 as an analysis score 114, similar to the third embodiment. In addition, as the analysis score 114, the percentage indicating the accuracy or degree of agreement (hereinafter referred to as "certainty") of the identified object name output by the image analysis model along with the object name such as "beef" when identifying an object in each menu image (e.g., "beef": 98.8%) may be used. In this case, the certainty of categories that were not identified (not detected) will be a small positive value (e.g., "chicken": 5.2%). The certainty value for all food categories becomes the analysis score 114.

[0086] As explained above, the meal count prediction device in the fourth embodiment offers the same advantages as the first embodiment, but because the menu feature extraction unit is equipped with a menu image analysis unit, it becomes possible to identify dish categories and characteristic phrases using only menu images, thereby improving the efficiency of menu feature extraction. [Examples]

[0087] In the third and fourth embodiments, examples of the various configurations of the menu feature extraction unit and methods for determining menu features were described when the dish category is set to be used as the menu feature. In the fifth embodiment, a specific configuration of the menu feature extraction unit and an example of a method for determining menu features will be described when the nutrition category is set to be used as the menu feature. In the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be explained.

[0088] Figure 19 shows an example of the configuration of the fifth embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those in the meal count prediction device 10 shown in Figure 2. In Figure 19, the meal count prediction device 40 includes a floor cumulative weight calculation unit 11, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, similar to the meal count prediction device 10 shown in Figure 2, as well as a menu feature extraction unit 51 similar to the menu feature extraction unit 12 shown in Figure 2. The menu feature extraction unit 51 also includes a nutrition category storage unit 52 and a nutrition category classification unit 53.

[0089] In this embodiment, the nutrition category 115 represents categories related to nutrients or components and their amounts, such as high protein, low carbohydrate, and high salt. Several dozen to several hundred individual categories in the nutrition category 115 are predetermined and stored in the nutrition category storage unit 52. In this embodiment, the menu data 103 includes data on nutritional components (nutrients (components) and values ​​(amounts)) for each menu, in addition to the date and menu name. The nutrition category classification unit 53 determines the menu features 104 for each menu by determining whether each menu can be classified into the individual categories read from the nutrition category storage unit 52 (applicable or not) based on the nutritional components of each menu in the input menu data 103.

[0090] Figure 20 is a table showing an example of the nutritional components (nutrient names and values) in the menu data 103 classified by the nutritional category classification unit 53 of the fifth embodiment, and the classification score 116 (nutrition category 115 and classification result) as menu features 104, using the menu item "Karaage Teishoku" (fried chicken set meal) as an example. In Figure 20, the nutritional components in the menu data 103 show the names of multiple nutrients included in "Karaage Teishoku" and their respective values. As described above, the nutritional category classification unit 53 refers to these nutritional components and determines whether "Karaage Teishoku" can be classified into the individual categories read from the nutritional category storage unit 52, and then classifies it. For example, based on the values ​​of protein, salt, and vitamin E in the nutritional components, the nutritional category classification unit 53 determines that it can be classified into the nutritional categories "high protein," "high salt," and "high vitamin E," and classifies it accordingly. In this way, the nutrition category classification unit 53 determines whether each menu item can be classified into all categories read from the nutrition category storage unit 52, classifies them, and outputs the classification score 116, which is the classification result, as the menu feature quantity 104.

[0091] The classification score 116 represents the classification result for all nutritional categories 115 for each menu item, and in the example in Figure 20, the classification result is shown as a binary value ("True", "False"). That is, as described above, the "Karaage Teishoku" (fried chicken set meal) is classified into nutritional categories such as "high protein", "high in salt", and "high in vitamin E", so the classification result for these is "True". On the other hand, although not shown in the figure, for example, based on the carbohydrate value in the nutritional components, it is determined not to be classified as a "low carbohydrate" nutritional category, so the classification result for "low carbohydrate" is "False". The nutritional category classification unit 53 outputs a binary value representing the classification result for all nutritional categories 115 as the classification score 116 for each menu item. This allows the nutritional category classification unit 53 to output the classification result quantitatively.

[0092] As explained above, the meal count prediction device in the fifth embodiment provides the same effects as the first embodiment, and because the menu feature extraction unit is equipped with a nutrition category classification unit, even when the system is set to use nutrition categories as menu features, it can determine whether each menu item can be classified into a nutrition category and output the classification result as a menu feature. [Examples]

[0093] In the sixth embodiment, we will describe an example of a meal consumption prediction device that predicts the number of meals consumed for each menu item at a future date and time (for example, 12:00 two days from now), and uses these predicted values ​​to calculate and recommend the recommended amount of ingredients to purchase. In the following description, we will omit explanations that overlap with the first embodiment and explain the differences.

[0094] Figure 21 shows an example of the configuration of the sixth embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those in the meal count prediction device 10 shown in Figure 2. In Figure 21, the meal count prediction device 60, like the meal count prediction device 10 shown in Figure 2, includes a floor cumulative weight calculation unit 11, a menu feature quantity extraction unit 12, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, and further includes a menu ingredient data storage unit 61 and a recommended purchase quantity calculation unit (also called a recommended purchase quantity output unit) 62.

[0095] In this embodiment, the meal count prediction unit 14 predicts the number of meals consumed for each menu item at the target date and time, using a future date and time (for example, 12:00 two days later) as the target date and time. As explained in the first embodiment, in order to predict the number of meals consumed at the target date and time, it is necessary to predict the floor cumulative weight 102 at that date and time. Therefore, in this embodiment, as an example, the function f explained in relation to (Equation 4) in the first embodiment is used. n (T) is used to predict the floor cumulative weight 102 for the target date and time. As described in the first embodiment, function f n (T) can be determined, for example, by averaging the time progression of the floor total weight 102 for each floor on the same date in the past (or the relevant date), so the f of the date and time to be predicted n(12:00) is uniquely determined.

[0096] Furthermore, the prediction of the floor cumulative weight 102 for the target date and time is performed using the function f n In addition to using (T), for example, the actual floor weight 102 at 12:00 the day before or on the day of the prediction (if the target date and time is 12:00 two days from now, then 2-3 days before) can be used as the predicted value, or the average of the actual floor weight 102 at 12:00 on the same day of the week as the target date and time over the past month (or any period such as six months or a year) can be used as the predicted value, or if there is no historical data for a newly constructed building, etc., if actual floor weight data from other buildings with a similar number of employees can be obtained, that can be used.

[0097] The meal count prediction unit 14 predicts the number of meals consumed using the predicted floor cumulative weight 102 for each floor for the target date and time, which has been obtained as described above. In addition, the meal count prediction unit 14 obtains menu features 104 and floor-specific menu preference scores 105, other than the predicted floor cumulative weight 102, in the same manner as in the first embodiment, and calculates and outputs the predicted number of meals consumed 110.

[0098] Next, we will explain the calculation of recommended purchase quantities and the recommendation process. First, the menu ingredient data storage unit 61 stores the menu ingredient data 117 that is input from an external source. The menu ingredient data is data (name of ingredient, quantity (amount per serving), etc.) about the ingredients necessary for cooking each menu item to be served in the cafeteria 5 on the predicted date. The menu ingredient data 117, like the menu data 103, is pre-input (registered) by the cafeteria operator or staff.

[0099] Figure 22 is a diagram showing an example of a flowchart of the calculation process for the recommended purchase quantity 118 by the recommended purchase quantity calculation unit 62 of the sixth embodiment. The operation of the recommended purchase quantity calculation unit 62 will be explained using Figure 22. In Figure 22, the recommended purchase quantity calculation unit 62 acquires (inputs) the predicted number of meals consumed 110 for each menu item for the predicted date and time, which has been calculated and output by the meal consumption prediction unit 14 as described above (S4001). The recommended purchase quantity calculation unit 62 reads the menu ingredient data 117 from the menu ingredient data storage unit 61 (S4002). The recommended purchase quantity calculation unit 62 calculates and outputs the recommended purchase quantity 118 using the predicted number of meals consumed 110 for each menu item and the menu ingredient data 117 (S4003). For the recommended purchase quantity 118, the recommended purchase quantity 118 for each ingredient can be calculated by multiplying the quantity of each ingredient for each menu item by the predicted number of meals consumed 110 (predicted number of meals multiplier). Furthermore, if there is inventory of one or more ingredients, the recommended purchase quantity calculation unit 62 may, in the recommended purchase quantity calculation step, take into account the inventory quantity of each ingredient and calculate and output the recommended purchase quantity 118 for the ingredients that are insufficient in inventory. In this case, the recommended purchase quantity 118 can be calculated by first calculating the purchase quantity of each ingredient and then subtracting the inventory quantity for each ingredient from that.

[0100] In this way, the recommended purchase quantity 118 calculated and output by the recommended purchase quantity calculation unit 62 is displayed on a display on the information processing device operating as a meal count prediction device 60, similar to the predicted meal count 110, or on displays or monitors installed in rooms such as kitchens where cafeteria staff perform various tasks, and the number of items to purchase for the predicted day is recommended to the cafeteria operators and staff. It is also possible that the food procurement work is carried out by cafeteria operators and staff working at a different location than Building 1. In that case, the recommended purchase quantity 118 is displayed on displays or monitors installed at the location where the food procurement work is carried out.

[0101] As explained above, the meal count prediction device in the sixth embodiment, in addition to the same effects as the first embodiment, is equipped with a recommended purchase quantity calculation unit, which allows for the calculation of recommended purchase quantities of ingredients that reflect the predicted meal count for the target date and time, and can be recommended to cafeteria operators and staff, thereby enabling the optimization of purchase quantities in the cafeteria. [Examples]

[0102] In the seventh embodiment, an example of a meal count prediction device is described that uses the predicted number of meals consumed for each menu item on the day of prediction to calculate and recommend revised meal counts to cafeteria staff before or during meal preparation work. In the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be described.

[0103] Figure 23 shows an example of the configuration of the seventh embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those of the meal count prediction device 10 shown in Figure 2. In Figure 23, the meal count prediction device 70, like the meal count prediction device 10 shown in Figure 2, includes a floor cumulative weight calculation unit 11, a menu feature quantity extraction unit 12, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, and further includes a preparation quantity difference calculation unit (also called a preparation quantity difference output unit) 71. The meal count prediction unit 14, like the first embodiment, calculates and outputs the predicted number of meals consumed 110 for each menu item at the target date and time (i.e., 12:00 on the day).

[0104] Figure 24 is a diagram showing an example of a flowchart of the process for creating a revised preparation quantity 120 by the preparation quantity difference calculation unit 71 of the seventh embodiment. The operation of the preparation quantity difference calculation unit 71 will be explained using Figure 24. In Figure 24, the preparation quantity difference calculation unit 71 acquires (inputs) the predicted number of meals consumed 110 for each menu item, which has been calculated and output by the meal consumption prediction unit 14 (S5001). The preparation quantity difference calculation unit 71 acquires the planned preparation quantity 119 for each menu item, which is input from an external source (S5002). The planned preparation quantity 119 is a value predetermined by the cafeteria operator or staff, etc., for the number of meals to be prepared for each menu item to be served in the cafeteria 5 on the day of the prediction target, and is pre-input (registered) by the cafeteria operator or staff, etc., similar to the menu data 103. Note that, as in the sixth embodiment, if the predicted number of meals consumed for each menu item has been calculated in advance for the purpose of purchasing ingredients, this predicted number of meals consumed may be used as the planned preparation quantity 119. In this case, the predicted number of meals consumed may be entered by the meal consumption prediction unit 14, similar to the predicted number of meals consumed 110 on the day of the prediction, or it may be entered by the cafeteria operator or staff.

[0105] The preparation quantity difference calculation unit 71 calculates the difference between the planned preparation quantity of 119 and the predicted number of meals consumed of 110 for each menu item (S5003). Here, if the difference is denoted as Δr, it is calculated as Δr = planned preparation quantity - predicted number of meals consumed. The preparation quantity difference calculation unit 71 determines for each menu item whether the calculated Δr is "0" or greater (S5004). If the Δr of any menu item is greater than "0", then for that menu item, the planned preparation quantity of 119 is greater than the predicted number of meals consumed of 110, so it is assumed that there will be a surplus of that menu item. On the other hand, if the Δr of any menu item is less than "0" (i.e., a negative value), then for that menu item, the planned preparation quantity of 119 is less than the predicted number of meals consumed of 110, so it is assumed that there will be a shortage of that menu item. Therefore, the preparation quantity difference calculation unit 71 determines whether Δr is greater than the excess allowable value if Δr for each menu is "0" or greater (S5005), and whether |Δr| (the absolute value of Δr) is greater than the shortage allowable value if Δr for each menu is less than "0" (S5006).

[0106] The excess tolerance value is a positive number representing the degree of excess that Cafeteria 5 considers acceptable even if there is leftover menu items. On the other hand, the shortage tolerance value is a positive number representing the degree of shortage that Cafeteria 5 considers acceptable even if there is a shortage of menu items. The excess tolerance value and shortage tolerance value may be the same for all menu items, may be different for each menu item, may be the same regardless of the menu or date, may be different only for specific menu items (for example, special menus offered only on special days such as Christmas), or both tolerance values ​​may be the same. The excess tolerance value and shortage tolerance value are predetermined by the cafeteria operator or staff and set in the preparation quantity difference calculation unit 71, or they may be input externally, similar to the planned preparation quantity 119. Figure 24 shows an example where the excess tolerance value and shortage tolerance value are set as common values ​​for each menu item.

[0107] If the calculation unit 71 determines that the Δr (or its absolute value) for each menu item is greater than the excess or deficit tolerance, it creates and outputs a revised preparation quantity 120 based on the Δr for each menu item (S5007). For example, if the Δr for any menu item is "20" and the excess tolerance is "10", the calculation unit 71 creates and outputs a revised preparation quantity 120 that reduces the Δr for that menu item by "10" or more so that the Δr for that menu item becomes "10" or less. Similarly, if the Δr for any menu item is "-20" and the deficit tolerance is "10", the calculation unit 71 creates and outputs a revised preparation quantity 120 that increases the Δr for that menu item by "10" or more so that the Δr for that menu item becomes "-10" or less.

[0108] The revised preparation quantity 120, created and output by the preparation quantity difference calculation unit 71 in this manner, is displayed on a display on the information processing device operating as a meal count prediction device 70, or on a display or monitor installed in a kitchen or other room where cafeteria staff perform preparation work, similar to the predicted meal count 110, and is recommended to cafeteria operators and staff. The recommendation of the revised preparation quantity may be made multiple times, both before and after the start of preparation work. This makes it possible to improve the accuracy of the revised preparation quantity 120 by reflecting the floor cumulative weight 102 according to the predicted meal count time.

[0109] As explained above, the meal count prediction device in the seventh embodiment, in addition to the same effects as the first embodiment, is equipped with a preparation quantity difference calculation unit. This unit creates a revised preparation quantity plan that corresponds to the predicted meal count, acceptable surplus, and acceptable shortage based on the attendance status of workers on the day, and can recommend it to cafeteria operators and staff, thereby enabling the optimization of the preparation quantity for each menu item in the cafeteria and reducing food waste. [Examples]

[0110] In the eighth embodiment, an example of a meal count prediction device that notifies staff of the surplus or shortage situation for each menu item when the predicted number of meals consumed for each menu item on the day of prediction deviates from the actual consumption situation in the cafeteria on that day (i.e., the prediction is wrong). In the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be explained.

[0111] Figure 25 shows an example of the configuration of the eighth embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those in the meal count prediction device 10 shown in Figure 2. In Figure 25, the meal count prediction device 80 includes a floor cumulative weight calculation unit 11, a menu feature extraction unit 12, a floor-specific menu preference storage unit 13, and a meal count prediction unit 14, similar to the meal count prediction device 10 shown in Figure 2. Furthermore, it includes a final meal count prediction unit 81 for the day and a meal surplus / deficit notification unit 82. The meal count prediction unit 14 calculates and outputs the predicted meal count 110 for each menu item at the target date and time (i.e., 12:00 on the day), similar to the first embodiment.

[0112] Figure 26 is a diagram showing an example of a flowchart of the process for predicting the final number of meals consumed on a given day by the final meal consumption prediction unit 81 of the eighth embodiment. The operation of the final meal consumption prediction unit 81 will be explained using Figure 26. Note that the process for predicting the final number of meals consumed on a given day shown in Figure 26 is executed after the opening time of the cafeteria 5. In Figure 26, the final meal consumption prediction unit 81 acquires (inputs) the current time T1 (S6001). The current time T1 may be acquired from time information inside the information processing device operating as the meal consumption prediction device 80, or it may be input from an external source. The final meal consumption prediction unit 81 acquires (inputs) the actual number of meals consumed for each menu item from the opening time of the cafeteria 5 on the day (12:00) to the current time T1 (hereinafter referred to as the real-time number of meals consumed on the day) 121 (S6002). The real-time number of meals consumed on the day can be obtained, for example, from payment settlement data (POS data, etc.) in the payment settlement system or meal ticket issuance system installed in the cafeteria 5. The final number of meals consumed on the day prediction unit 81 uses the acquired current time T1 and the real-time number of meals consumed on the day 121 to predict and output the final number of meals consumed on the day (hereinafter referred to as the final number of meals consumed on the day) 122 for each menu item at the closing time T2 of cafeteria 5 (for example, 13:00) (S6003). The final number of meals consumed on the day 122 can be predicted, for example, from the average of the past daily actual number of meals consumed at cafeteria 5 (which may be the total actual number of meals for all menu items, or the actual number of meals for each menu feature) over time (hereinafter referred to as the average number of meals consumed trend).

[0113] Figure 27 is a graph showing an example of the trend in the average number of meals consumed for a menu item whose food category, menu feature 104, is "meat." In Figure 27, the actual number of meals consumed increases proportionally with time from the opening time of Cafeteria 5, and after 12:30 the rate of increase decreases (slightly increases) to reach the final number of meals consumed. The actual average number of meals consumed per hour does not follow such a neat proportional relationship as shown in Figure 27, but for example, by obtaining a function of order of magnitude or higher using linear fitting with the average number of meals consumed per hour, it is possible to obtain the trend in the average number of meals consumed as shown in Figure 27.

[0114] For example, for a menu item on the day whose menu feature 104 is "meat," if the real-time consumption count 121 at 12:10 is lower than the actual consumption count at 12:10 in the average consumption trend shown above, the final consumption count prediction unit 81 for the day adjusts the function of the average consumption trend to match the increasing trend of the real-time consumption count 121 and calculates (predicts) the final consumption count 122 for the day. The same applies if the real-time consumption count 121 at 12:10 is higher than the actual consumption count at 12:10 in the average consumption trend shown above. Note that the prediction time (timing) for the final consumption count 122 for the day by the final consumption count prediction unit 81 is not limited to 12:10 as described above and can be any time. However, it is preferable that it be a time when the real-time consumption count 121 for the day, which allows for understanding the day's eating trends, can be obtained. Furthermore, the prediction of the final number of meals consumed on the day (122) is not limited to just one prediction; it may be made multiple times. For example, the prediction of the final number of meals consumed on the day (122) may be made at 10-minute intervals from 12:10 to 12:30.

[0115] Figure 28 is a diagram showing an example of a flowchart of the food consumption surplus / deficit notification process by the food consumption surplus / deficit notification unit 82 of the eighth embodiment. The operation of the food consumption surplus / deficit notification unit 82 will be explained using Figure 28. In Figure 28, the food consumption surplus / deficit notification unit 82 obtains (inputs) the predicted number of meals consumed 110 for each menu item for the prediction target date and time (12:00 on the day) from the food consumption prediction unit 14 (S7001). The food consumption surplus / deficit notification unit 82 obtains (inputs) the final number of meals consumed on the day for each menu item, which is output by the final number of meals consumed on the day prediction unit 81. The food consumption surplus / deficit notification unit 82 calculates the difference between the predicted number of meals consumed 110 and the final number of meals consumed on the day 122 for each menu item (S7003). If the difference is Δm, it is calculated as Δm = predicted number of meals consumed 110 - final number of meals consumed on the day 122. The food surplus / shortage notification unit 82 determines whether the calculated Δm for each menu item is "0" or greater (S7004). If the Δm for any menu item is greater than "0", it is assumed that the predicted number of servings 110 for that menu item is greater than the final number of servings 122 for the day, so that menu item is expected to be in surplus. On the other hand, if the Δm for any menu item is less than "0" (i.e., a negative value), it is assumed that the predicted number of servings 110 for that menu item is less than the final number of servings 122 for the day, so that menu item is expected to be in short supply. Therefore, if the Δm for each menu item is "0" or greater, the food surplus / shortage notification unit 82 determines whether Δm is greater than the surplus notification threshold (S7005), and if the Δm for each menu item is less than "0", it determines whether |Δm| (the absolute value of Δm) is greater than the shortage notification threshold (S7006).

[0116] The surplus notification threshold and the shortage notification threshold are both positive numbers that indicate the threshold (minimum value) for the surplus or shortage number that necessitates a food intake surplus or shortage notification 123. The surplus notification threshold and the shortage notification threshold may be a common value for each menu item, a different value for each menu item, the same value regardless of the menu or date, a different value only for specific menu items, or the same value for both thresholds. The surplus notification threshold and the shortage notification threshold are determined in advance by the cafeteria operator or staff and set in the food intake surplus / shortage notification unit 82, or input from an external source. Figure 28 shows an example where the surplus notification threshold and the shortage notification threshold are set as common values ​​for each menu item.

[0117] If the Δm (or its absolute value) for each menu item is determined to be equal to or greater than the excess notification threshold or the shortage notification threshold, the food intake surplus / shortage notification unit 82 determines that a food intake surplus / shortage notification 123 is necessary and outputs (sends) the food intake surplus / shortage notification 123 (S7007). For example, if the Δm of any menu item (here, "Karaage Teishoku") is equal to or greater than the excess notification threshold, the food intake surplus / shortage notification unit 82 sends a food intake surplus / shortage notification 123 for that menu item, for example, "Karaage Teishoku is likely to be left unsold, so please come and eat it." Similarly, if the |Δm| of any menu item is equal to or greater than the shortage notification threshold, the food intake surplus / shortage notification unit 82 sends a food intake surplus / shortage notification 123 for that menu item, for example, "Karaage Teishoku is expected to be sold out, so please come early." On the other hand, if Δm (or its absolute value) for each menu item is less than the surplus notification threshold or the shortage notification threshold, the surplus or shortage for that menu item is considered to be within the acceptable range (not problematic) for the cafeteria 5, and the surplus / shortage notification unit 82 terminates processing without outputting (sending) a surplus / shortage notification 123.

[0118] In this way, the food intake surplus / deficit notifications 123 (each message) transmitted from the food intake surplus / deficit notification unit 82 are received and displayed on notification units 84 of notification terminals 83, such as displays and monitors installed near the entrance of the cafeteria 5 or the ticket vending terminal, signage terminals installed on each floor of building 1, and company PCs, tablets, smartphones, etc., used by individual employees, and the employees are notified. (If the notification terminal 83 is a company PC, tablet, smartphone, etc., the employee application (software) installed on it acts as the notification unit 84 to receive the food intake surplus / deficit notifications 123 and display them on the display.) Furthermore, the food surplus / shortage notification 123 is not limited to a single instance, but may be issued multiple times during the cafeteria 5's operating hours. For example, as in the example above, if the final number of meals consumed for the day prediction unit 81 makes multiple predictions for the final number of meals consumed for the day 122, the food surplus / shortage notification unit 82 may determine whether a food surplus / shortage notification 123 is necessary and issue the notification each time the final number of meals consumed for the day prediction unit 81 outputs the final number of meals consumed for the day 122. The timing of the food surplus / shortage notification 123's issuance depends on the timing of the final number of meals consumed for the day prediction unit 81 makes its prediction for the final number of meals consumed for the day 122. However, in order to curb menu surplus (unsold) and warn of menu shortages (sold out), it is preferable that the food surplus / shortage notification 123 (the first food surplus / shortage notification 123 if issued multiple times) is issued as early as possible after the cafeteria 5 opens for business.

[0119] As explained above, the meal count prediction device of the eighth embodiment provides the same effects as the first embodiment, plus a final meal count prediction unit for the day and a meal surplus / shortage notification unit. This allows the device to predict the final meal count for the day based on the real-time meal count for the day, determine whether or not to notify staff of meal surplus / shortages based on this predicted value, and notify staff of unsold or sold-out menu items and messages to encourage customers to visit. In particular, it helps to encourage the consumption of menu items that are likely to be left unsold, thereby reducing food waste. [Examples]

[0120] In the first embodiment, an example was described in which the meal count prediction unit acquires the floor cumulative weight, menu features, and floor-specific menu preference scores over a predetermined period and predicts the number of meals consumed for each menu item at the target date and time. However, it is also possible for the meal count prediction unit to consider other data in addition to these to predict the number of meals consumed. For example, weather may affect the eating trends for each menu item in the dining hall, so in the ninth embodiment, an example of a meal count prediction device that also takes weather data into account will be described. Note that in the following description, explanations that overlap with the first embodiment will be omitted, and only the differences will be described.

[0121] Figure 29 shows an example of the configuration of the ninth embodiment of the meal count prediction device, and the same reference numerals are used for components identical to those of the meal count prediction device 10 shown in Figure 2. In Figure 29, the meal count prediction device 90 includes a floor cumulative weight calculation unit 11, a menu feature quantity extraction unit 12, and a floor-specific menu preference storage unit 13, similar to the meal count prediction device 10 shown in Figure 2, and a meal count prediction unit 91 similar to the meal count prediction unit 14 shown in Figure 2. The meal count prediction unit 91 receives input of the floor cumulative weight 102, menu feature quantity 104, and floor-specific menu preference 105 for a predetermined period, as well as weather data 124. The weather data 124 is, for example, data including the weather code, temperature (°C), and humidity (%), and is weather forecast data for the day to be predicted. Weather forecast data can be obtained from websites such as those of the Japan Meteorological Agency and the Japan Weather Association, and the necessary data from the weather forecast data is input to the food consumption prediction device 90 from outside as weather data 124.

[0122] The weather data 124 is used in the meal count prediction unit 91 to calculate the predicted number of meals consumed 125 as a weight for the preference of each menu feature 104 (in the example of Figure 6, the food category) in the floor-specific menu preference 105. The weight of each menu feature for the weather data 124 is predetermined and set in the meal count prediction unit 91 according to the possible values ​​of the weather data 124. For example, if the weather code is a value indicating rain or snow and the temperature is 10℃ or lower, the weight for "udon" and "ramen" among the food categories will be "1.2", and the weight for other food categories will be "1" or less. The meal count prediction unit 91 multiplies each preference by the weight corresponding to the content of the input weather data 124 from the predetermined weights to assign weights to each menu feature. The meal consumption prediction unit 91 uses the weighted floor-specific menu preference scores 105 in this manner to calculate and output the predicted number of meals consumed for each menu item (predicted number of meals consumed for the entire floor) 125 according to equation (3) described in the first embodiment.

[0123] As explained above, the meal count prediction device in the ninth embodiment offers the same advantages as the first embodiment. In addition, the meal count prediction unit also takes weather data into account when calculating the predicted meal count. This makes it possible to predict the number of meals consumed for each menu item according to the weather on the day of prediction, even when the weather affects the eating trends for each menu item in the cafeteria, thereby improving the accuracy of meal count prediction.

[0124] Although embodiments of the present invention have been described in detail above, the present invention is not limited to the embodiments described above, and various design modifications can be made without departing from the spirit of the invention as described in the claims. For example, each of the embodiments described above has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to having all of the described configurations. Furthermore, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Moreover, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations. [Explanation of Symbols]

[0125] 1... Building 2… Elevator 3...basket 4…Basket weight sensor 5…Dining room 6…Workplace 10, 20, 30, 40, 50, 60, 70, 80, 90... Meal consumption prediction device 11…Floor total weight calculation unit 12, 31, 41, 51... Menu Feature Extraction Unit 13…Floor-specific menu preference data storage section 14, 91... Consumption forecast department 15…Learning Department 32…Feature Term Classification Data Creation Department 33…Feature word classification data storage unit 34…Feature word matching unit 42…Image analysis model storage unit 43… Menu Image Analysis Department 52…Nutrition Category Storage Section 53…Nutrition Category Classification Department 61... Menu ingredient data storage section 62...Recommended Purchase Quantity Calculation Unit 71...Calculation unit for the difference in the number of items prepared 81...Department for predicting the final number of meals consumed on the day. 82... Department for Notifying the Number of Food Consumption Deficiencies or Overconsumption 83…Notification terminal 84…Notification department

Claims

1. A demand forecasting device for goods provided in a building, A floor attendance output unit outputs a floor attendance figure representing an indicator of the number of people present on each floor of the building. For each item on the menu, a menu feature extraction unit selects or determines menu features, Each of the aforementioned floors includes a floor-specific menu preference storage unit that stores floor-specific menu preference values ​​representing the preference level for each of the aforementioned menu features, The system includes a demand forecasting unit that outputs the predicted demand for each menu item on the target date based on the number of people on each floor for a predetermined period on the target date and the menu preference for each floor. Demand forecasting device.

2. A demand forecasting device according to claim 1, The aforementioned menu is a menu offered at a food and beverage establishment. The aforementioned predicted demand is the predicted number of meals consumed. Demand forecasting device.

3. A demand forecasting device according to claim 2, Each floor of the building is equipped with a floor weight output unit that outputs a floor weight representing the total weight of the occupants on that floor. The floor occupancy output unit outputs the floor occupancy based on the floor's total weight. Demand forecasting device.

4. A demand forecasting device according to claim 3, With an additional learning section, The aforementioned menu preference levels for each floor are predetermined and stored in the aforementioned menu preference level storage unit for each floor. The learning unit calculates a floor-specific menu preference update value based on the floor cumulative weight for the predetermined period on past learning target days, the menu features for each menu, the predicted number of meals consumed for each menu, and the actual number of meals consumed for each menu, and updates the floor-specific menu preference based on the floor-specific menu preference update value. Demand forecasting device.

5. A demand forecasting device according to claim 2, The menu feature extraction unit, The menu features are selected or determined using feature phrase classification data, which includes at least a predetermined set of feature phrases relating to ingredients or cooking methods, and a set of menu words extracted from the names of a set of menus served at the food and beverage establishment over a past period of months to years, wherein each menu word is classified into its respective feature phrase. Demand forecasting device.

6. A demand forecasting device according to claim 2, The system further includes a recommended purchase quantity output unit that outputs a recommended purchase quantity based on the predicted number of meals consumed on the predicted target date, which is a future date, and the ingredient data for each menu item. Demand forecasting device.

7. A demand forecasting device according to claim 2, The system further includes a preparation quantity difference output unit that outputs a preparation quantity revision proposal if the difference between the predetermined planned preparation quantity for each menu item and the predicted number of servings consumed is greater than a predetermined tolerance value. Demand forecasting device.

8. A demand forecasting device according to claim 2, The system further includes a food consumption surplus / deficit notification unit that, based on the actual number of meals consumed for each menu item from the start of business to a predetermined time on the day of the prediction target, predicts the final number of meals consumed for each menu item at the end of business hours of the food and beverage facility, and notifies the facility of a surplus or deficit if the difference between the final number of meals consumed and the predicted number of meals consumed is greater than or equal to a predetermined standard value. Demand forecasting device.

9. A demand forecasting device according to claim 2, The menu feature extraction unit, The menu features are selected or determined by analyzing the image of the menu and identifying one or more objects detected in the image. Demand forecasting device.

10. A demand forecasting device according to claim 2, The menu feature extraction unit, Based on a nutrition category that includes a predetermined number of such categories, which are categories relating to nutrients or components, and the nutritional components of the menu, the menu features are selected or determined. Demand forecasting device.

11. A demand forecasting device according to claim 2, The demand forecasting unit further predicts or outputs the number of meals consumed for each menu item on the target day based on weather data. Demand forecasting device.

12. A method for forecasting the demand for goods offered in a building, For each floor of the aforementioned building, the floor population, which represents an indicator of the number of people present on that floor, is output. For each item on the menu, select or determine menu features. For each of the aforementioned floors, a floor-specific menu preference score representing the preference score for each of the aforementioned menu features is stored in advance. Based on the number of people on each floor during a predetermined period on the target date and the menu preference for each floor, the predicted demand for each menu item on the target date is output. Demand forecasting methods.

13. A demand forecasting method according to claim 12, The aforementioned menu is a menu offered at a food and beverage establishment. The aforementioned predicted demand is the predicted number of meals consumed. Demand forecasting methods.

14. A demand forecasting method according to claim 13, For each floor of the aforementioned building, the total floor weight representing the total weight of occupants on that floor is output. The number of people on the floor is output based on the total weight of the floor. Demand forecasting methods.

15. A demand forecasting method according to claim 14, Based on the cumulative weight of the floors for the predetermined period on past learning target days, the menu features for each menu, the predicted number of meals consumed for each menu, and the actual number of meals consumed for each menu, a floor-specific menu preference update value is calculated, and the floor-specific menu preference is updated based on the floor-specific menu preference update value. Demand forecasting methods.

Images