Order quantity prediction system in restaurant

Abstract

To provide a system capable of predicting an order amount even under a specific condition that mainly a specific user repeatedly visits a restaurant and eats and drinks.SOLUTION: An order quantity prediction system in a cafeteria, wherein a past order history database stores a past order history in the cafeteria for each user in association with attribute information of the user and a time series, an external factor database stores arbitrary information excluding the order history in association with the time series, an order pattern model generation unit generates an order pattern model along the time series based on information stored in the past order history database and information stored in the external factor database, and an order quantity prediction calculation unit calculates an order prediction model of the entire cafeteria along the time series on a specific day, the order quantity update computation unit computes an updated order prediction model of the entire cafeteria along a time series on the specific day.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an order quantity prediction system for a restaurant. [Background technology]

[0002] In recent years, with the aim of achieving the SDGs, it has become extremely important for places that serve food and beverages, such as cafeterias and restaurants, to accurately predict daily order volumes in advance. For this reason, it is common for cafeteria management companies to have large databases. These databases typically analyze the sales volume of each menu item each day and store data over many years.

[0003] In recent years, systems have been considered that analyze store data such as the number of customers visiting the store, average customer spending, best-selling items, busy days, time periods, and comparisons with last year, to predict order volumes based on the number of customers visiting the next day or a year from now. Summary of the Invention

[0004] Conventional systems are models that predict the number and time of arrival of unspecified customers, making it extremely difficult to predict the demand for extremely large, concentrated orders in a limited period of time under the unique conditions in which specific users repeatedly visit a restaurant to eat and drink.

[0005] An object of the present disclosure is to provide a system that can predict order quantities even under unique conditions where a particular user repeatedly visits a restaurant and eats and drinks.

[0006] After extensive research, the present inventor has unexpectedly discovered that it is possible to predict the order volume at a restaurant by performing calculations using Bayesian statistics based on user attribute information and the user's order history. The present disclosure is based on this finding.

[0007] According to one embodiment of the present disclosure, there is provided an order quantity prediction system for a restaurant, comprising a past order history database, an external factor database, an order pattern model generation unit, an order quantity prediction calculation unit, and an order quantity update calculation unit, The past order history database stores the past order history of each user at the restaurant in association with the attribute information of the user and a time series of the order history, The external factor database stores any information other than the order history in association with a time series; the order pattern model generation unit generates, based on the information stored in the past order history database and the information stored in the external factor database, an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series; The order quantity prediction calculation unit (1) Calculating an order prediction model for each user along a time series for the specific day based on the order pattern model for each user, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each user. (2) calculating an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each attribute; and / or (3) calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; The order quantity update calculation unit (1) Calculating an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user using a Bayesian statistical method based on the actual order history of each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user. (2) calculating an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute based on the actual order history for each user on a specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute; and / or (3) updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant based on the time series on the specific day; An order quantity forecasting system is provided.

[0008] According to one embodiment of the present disclosure, a system can be provided that is capable of predicting order quantities even under unique conditions in which a specific user repeatedly visits a restaurant and eats and drinks there. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 illustrates an example of a functional block diagram of an order quantity forecasting system according to the present disclosure. [Figure 2] 1 shows an example of a processing flow of the order quantity prediction system of the present disclosure. [Figure 3] 10 shows a processing flow of an order pattern model generation unit in the order quantity prediction system of the present disclosure. [Figure 4] 10 shows a processing flow of an order quantity prediction calculation unit of the present disclosure. [Figure 5] 10 shows a processing flow of an order quantity update calculation unit of the present disclosure. [Figure 6] 1 shows a correction parameter calculation unit of the present disclosure. [Figure 7] An example of a conceptual diagram of Bayesian inference is shown below. [Figure 8] 1 shows a schematic diagram of the restaurant ordering system of the present disclosure. [Figure 9] FIG. 1 illustrates an example of a functional block diagram of an order receiving system according to the present disclosure. [Figure 10] FIG. 1 shows an example of a functional block diagram of a display system according to the present disclosure. [Figure 11] 1 illustrates an example of an order quantity prediction system and a restaurant ordering system according to the present disclosure. [Figure 12] 1 illustrates an example of information stored in a past order history database of the present disclosure. [Figure 13] 10 shows an example (box-and-whisker analysis) of the multiple regression analysis results in the order pattern model generation unit of the present disclosure. [Figure 14] 10 shows an example of a result of multiple regression analysis (Pearson correlation analysis) in the order pattern model generation unit of the present disclosure. [Figure 15] 1 shows an example of an order pattern model generated by an order pattern model generation unit of the present disclosure. [Figure 16] 1 shows an example of an order pattern model generated by an order pattern model generation unit of the present disclosure. [Figure 17] 1 illustrates an example of information stored in an order reservation database of the present disclosure. [Figure 18] 10 shows an example of an order prediction model for each user calculated by the order quantity prediction calculation unit of the present disclosure. [Figure 19] 10 shows an example of an order prediction model for each user calculated by the order quantity prediction calculation unit of the present disclosure. [Figure 20] 10 shows an example of an order prediction model for the entire dining hall calculated by the order quantity prediction calculation unit of the present disclosure. [Figure 21] 10 shows an example of an order prediction model for the entire dining hall calculated by the order quantity prediction calculation unit of the present disclosure. [Figure 22] 1 illustrates an example of information stored in a same-day order history database of the present disclosure. [Figure 23] 10 shows an example of an updated order prediction model for the entire dining hall calculated by the order quantity update calculation unit of the present disclosure. Specific Description of the Invention

[0010] [Conventional systems and social background] In cafeterias that serve large quantities of food in so-called "closed locations," such as company cafeterias or student cafeterias, where it is almost certain that a large number of specific members will visit and the members are always almost always the same people, data analysis focusing on individuals was not possible with conventional systems.

[0011] In other words, while conventional systems can predict the number of customers and the number of popular menu items in advance, they rarely or never take into account information about the attributes of each user. In "closed organizations" such as companies or schools, it is assumed that the members are almost always the same, and conventional systems were not designed to address the unique phenomena of cafeterias where such members visit regularly (e.g., almost daily). Therefore, conventional systems can be said to be static predictions that predict the number of customers and the number of menu items ordered in advance. In closed cafeterias where specific members eat daily, such predictions cannot be made based on individual members' past eating data. Furthermore, they are not suitable for predictions in "closed" cafeterias where a large number of members visit and eat within a limited time frame.

[0012] Conventional systems are primarily designed to be effective at predicting customer numbers at multiple general restaurants or specific restaurants (diners) in tourist destinations. This means that they can accurately forecast customer numbers by combining past customer data with information such as season, weather, and day of the week. They can also predict customer arrival times and the number of orders (orders) for popular menu items with relatively high accuracy. However, these conventional systems are not very effective in closed-location, high-volume restaurants within a closed organization. Time-based customer forecasts alone have the drawback of being unable to predict the demand for extremely large, concentrated orders within a limited time, which is necessary for closed-location, high-volume restaurants. Furthermore, they fail to take into account the unique characteristics of closed-location restaurants, where members of a closed organization repeatedly visit the restaurant and eat there every day.

[0013] A characteristic of bulk-supply cafeterias in "closed locations" within "closed organizations" is that they are frequented by the same type of people, i.e., members of that organization, on a regular basis (e.g., almost daily). Furthermore, the members who are always in the closed location tend to visit at concentrated times. Such bulk-supply cafeterias in "closed locations" typically experience peak customer numbers, for example, between 12:00 and 12:15, when lunch breaks begin. This means that the cafeterias are typically overwhelmed with customers. Conventional forecasting systems are not capable of accurately predicting the number of customers during such extremely concentrated periods. Therefore, they are not capable of predicting the number of people who will flood in between 12:00 and 12:15 every day, or the number of menu items that will be ordered there. Furthermore, they do not have the ability to predict the number of customers or the order quantities of each menu item for the remaining time until the cafeteria closes at 1:30, based on the situation up to 12:15. In these "closed place" cafeterias that supply large quantities of food, it is normal for roughly the same staff to visit and eat in large numbers at very limited times on a regular basis (for example, almost every day), but in such a normal situation, there was no system that enabled dynamic forecasting, such as constantly re-predicting future order quantities. However, while such dynamic order forecasting is exactly what is needed in "closed place" cafeterias that supply large quantities of food, conventional systems did not have the functionality to meet this need.

[0014] Therefore, the current situation is that a "closed-place" large-volume restaurant responds to short-term, concentrated customer visits and orders based solely on the intuition and human judgment of the person in charge (e.g., the head chef) of the restaurant, with the support of the restaurant's own system, based on past data and past memories up until the day before. The person in charge of a "closed-place" large-volume restaurant predicts order quantities based on past performance and human intuition, makes a menu order forecast, prepares the necessary ingredients, and commands his or her subordinate chefs to prepare ingredients and actually cook the menu items according to the predicted quantities. However, on the day the restaurant opens, once the restaurant opens, the person in charge almost always relies on intuition to determine how much the actual customer orders differ from the predicted quantities for each menu item, which is the manager's final decision. The manager then re-predicts the final state until closing based on human intuition, and adjusts the order quantities from that time onward. In other words, the reality is that operations are largely dependent on human intuition.

[0015] Furthermore, in specific "closed organizations" such as companies and schools, employee cafeterias and student cafeterias, which are "diners operated in closed locations," are characterized by the fact that employees and students who belong to those closed organizations habitually eat there almost every day, and the same people tend to eat there almost every day. On the other hand, conventional prediction systems, which are becoming popular in the food and beverage industry, are completely different in that they are typically used in restaurants where the same person almost never visits two days in a row. In other words, conventional systems are designed for people who visit almost only once. Therefore, conventional systems do not have the ability to predict the behavior of each individual or attribute based on data such as the attributes of each individual employee who eats, nor do they have the ability to predict the behavior of the entire cafeteria by accumulating predictions about the behavior of each individual or attribute based on individual data.

[0016] Furthermore, while conventional systems may be able to predict the number of customers and the number of orders for each menu item for an entire cafeteria in a "closed location" depending on the parameter and function settings, they do not record the behavior or food consumption data of individual staff members, nor do they have the functionality to make predictions for each individual staff member or their attributes, and they cannot solve problems in cafeterias where customers visit intensively during very limited, fixed periods every day.

[0017] The following model is currently common in closed-space cafeterias, such as company cafeterias or student cafeterias, where members of a specific closed organization eat daily. For example, the head chef predicts the menu order before the cafeteria opens at noon. He then prepares the initial cooking quantity based on this initial forecast. Guests begin arriving at noon, and by 12:15, a significant number of guests have arrived. At that point, human intuition is used to reassess the order quantity until the cafeteria closes around 1:30 PM, and the final cooking quantity is determined midway through the cafeteria's operation. These dynamic adjustments to order quantities often rely entirely on the head chef's intuition, based on past experience. Alternatively, traditional systems owned by cafeteria operators may use AI to predict the number of guests before the cafeteria opens and, based on past data, to predict how many of a particular menu item will be ordered. However, these traditional systems are not linked to order-taking systems (e.g., the cafeteria's cash register), nor do they have the ability to observe or record individual behavior. Therefore, for example, if a restaurant opens at 12:00 as described above, there is no function to re-predict the total order quantity, which is the sum of the number of customers and orders for a specific menu item up to 12:15 and the subsequent orders and the total order quantity until the restaurant closes, by looking at the individual customer arrival times and the order status of the menu items. In conventional systems, the only way to determine whether a deviation from the initially predicted quantity would occur was through human intuition, and there was no technology to correct this. To incorporate a technology for mid-course corrections, new technology different from conventional technology was absolutely necessary. Even if mid-course corrections were possible using existing technology, conventional technology lacks the functionality to link with the order reception system (e.g., the restaurant's cash register), making it impossible to perform mathematical and statistical analysis using real-time data on the behavior of specific individuals or groups with specific attributes. In other words, order quantity predictions would need to be constantly corrected using AI processing, but such conventional technology lacks the technology or functionality to provide such a basis.

[0018] This disclosure takes into consideration the particular nature of a large-volume cafeteria in a "closed place" within a "closed organization," and further considers the importance of comparing and recognizing the deviation between the actual values ​​up to that point and the prior forecast, by taking into account the predicted values ​​before the cafeteria opens, customer data up through the order reception system up to a certain elapsed time, for example 12:15, and real-time data obtained moment by moment from the cafeteria registers regarding individual employees and groups with specific attributes, and then predicting the arrival point and the intermediate order curve at that point. Conversely, without consideration of the "importance" of these things, the need for this disclosure would not be recognized at all.

[0019] The above "importance" will be explained in more detail. First, a large-volume cafeteria in a "closed location" is a place where a very large number of customers, for example, 1,000 or 2,000 people, especially members of a specific closed organization, eat daily. These members visit in large numbers, for example, at lunchtime, between 12:00 and 1:00 PM. Furthermore, they visit in large numbers over a very short, specific time period, around 12:00 PM to 12:15 PM. The premise of this disclosure is that such a large number of people visit in large numbers at a specific time. This concentration far exceeds that of a typical large restaurant. What makes it different from a regular restaurant is that customers who visit in a very short period of time are extremely concentrated, and it is normal for them to visit in overwhelming density, which is a unique characteristic that is discussed in this disclosure, and because of this unique characteristic, it is not enough to just make advance predictions; predictions of the number of customers and the number of orders for the time remaining until closing must be revised based on the number of customers and the number of orders that change from moment to moment through the order receiving system (for example, the restaurant's cash register) from the time the restaurant opens, as well as data on each individual employee.

[0020] Furthermore, these customers are often people who habitually visit the same type of employees every day, such as in a workplace like a company or factory, or a student cafeteria on a university campus. Therefore, since these are places where specific people repeatedly visit and eat, pre-opening predictions based on various past data can produce more accurate order forecasts than typical customer forecasting systems, compared to predictions for a general restaurant where an unspecified number of people visit and eat. However, conventional forecasting systems used by cafeteria operators do not have data on specific individuals or specific groups, and therefore do not have such capabilities.

[0021] To accommodate the overwhelming number of employees visiting, a "preparation program" is typically designed before opening, specifying how much of each menu item should be prepared in advance, how much ingredients (vegetables, meat, etc.) should be prepared in advance, how much should be fried at a certain time, and how much should be plated at a certain time. However, the unique feature of a "closed" high-volume cafeteria is that the total number of orders that must be prepared in advance can be enormous, ranging from hundreds to thousands. Customers visit at a time density far greater than that of a typical restaurant, and orders are placed in a completely different manner. Therefore, for example, if the actual order quantity differs from the pre-prepared forecast, it can have a significant financial impact on the cafeteria operator. If order forecasts for each menu item are not made wisely, discrepancies of hundreds of items, in extreme cases, could occur. This is the importance of the discrepancy between demand forecasts and actual results in a "closed" high-volume cafeteria. Furthermore, because the difference between the predicted amount and the actual amount is enormous, after the restaurant opens, the restaurant's operating status must be monitored and the order forecast adjusted according to the situation from moment to moment, a situation unique to restaurants that supply large quantities in a ``closed location.''

[0022] In the case of a large-volume cafeteria in a "closed location" within a "closed organization," the owner is typically the "closed organization" itself, i.e., a company or educational institution such as a university. The cafeteria operator, or "contract food service provider," is obligated to provide meals under a contract with the "closed organization." These cafeteria operators are called "contract food service providers." In other words, the company or educational institution acts as the contractor, outsourcing the provision of meals to the cafeteria operator under a contract under which the cafeteria operator must provide certain services. These contracts are often subsidized by the contractor in the form of a budgetary subsidy, and the cafeteria operator is obligated to provide the most comprehensive menu possible at the lowest possible price, ensuring that the employees and students of the client company or educational institution are satisfied with the food and beverages they provide daily. In other words, the cafeteria operator, as the contractor, is obligated to continue providing satisfying meals as efficiently as possible within a limited budget. Within relatively strict economic circumstances, that is, within a limited budget, it is necessary to provide a menu that satisfies customers as much as possible. Therefore, in a cafeteria that supplies a large amount of food in a "closed place," as mentioned above, a huge number of customers visit and eat intensively in a limited time, so it is not enough to think mainly about "safety measures" such as preparing a large number of meals in advance or that it is convenient to prepare more, but rather it is necessary to eliminate waste as much as possible and economic rationality is also very important, such as it is desirable for advance predictions to match the actual situation as much as possible.

[0023] Taking too many "safe measures" or overestimating the quantity required before the restaurant opens can result in wasted prepared ingredients or leftover finished dishes. Failure to accurately forecast results in significant discrepancies between the predicted and actual orders, resulting in significant "losses." In particular, finished dishes and ingredients prepared just before the final cooking stage (such as cooking or boiling) must be discarded if they remain left over when the restaurant closes. If this waste is significant, profitability declines, and the contractor's profits from the services under this contract—calculated by deducting the contractor's costs from the amount permitted to be received from the client (subsidies and grants)—are wasted by the cost of this waste. Historically, these contracts have typically been financially challenging for the contractor, who is the restaurant operator. High waste volumes result in significant reductions in profits. Based on actual data, waste exceeding 10% results in a significant deficit. For example, if 300 meals of a certain menu item are predicted and prepared before opening, but the actual order quantity is 270, 30 meals will have to be wasted, which would significantly reduce the profits of the cafeteria operator, who is the contractor, and result in a loss. Therefore, it is desirable for the difference between the predicted quantity and the actual order quantity to be as small as possible, and although this is purely a general operation, it is said that it is desirable for it to be less than 5% if possible. In other words, in the above example, if 300 meals are predicted, it is desirable for the actual order to be around 285 meals.

[0024] On the other hand, there is also the idea that it is sufficient to cook only the exact amount ordered. For example, it is possible to think that cooking only after an order is received is sufficient. However, even in this case, there are significant risks inherent in a large-scale cafeteria in a "closed location" within a "closed organization." As mentioned above, a large-scale cafeteria in a "closed location" within a "closed organization" is a place visited in large numbers by specific employees and students during limited time periods. It is a place that is used repeatedly, so much so that it is unheard of for a place visited by such a large number of people on a regular basis, almost every day. Furthermore, for employees and students, lunchtime is an important time to relax, a time for employee rest and rejuvenation, and a time of vital biological activity in which employees' bodies replenish with nutrients, hydration, and energy through meals. Cafeterias are a place where such important vital activities take place, where eating and drinking are performed. Furthermore, in recent years, cafeterias have gone far beyond being perceived as places where people simply eat; they are now places where employees or students can communicate, talk, relax, exchange information, and have a good time in a harmonious atmosphere, providing a valuable form of recreational time. Therefore, that time must be used extremely effectively, and it is desirable for it to be comfortable. Therefore, it is not very reasonable to just wait until an order is placed before cooking. There are times when making customers wait is acceptable and times when it is not, and this is deeply related to the cafeteria's operational policy and management philosophy, but in cafeterias that supply large quantities in a "closed place" within a "closed organization," it is common and desirable for the acceptable range of waiting time to be narrow.

[0025] On the other hand, there is also the idea that "it is better to serve meals as freshly prepared as possible." Even in a "closed-space" large-volume cafeteria, it is not ideal for all of the menu items to be pre-prepared and arranged on plates or other tableware. This is one of the important know-hows of cafeteria operators, and it lies in the combination of dishes: "(a) pre-prepared items, (b) items with very short cooking times, (c) items with a certain cooking time, and (d) items with long cooking times." This balance between (a) and (d) is deeply related to the cafeteria management philosophy of the client company, university, or other educational institution. Cafeteria operators, as contractors, can exercise a certain degree of originality based on their know-how, technology, and unique philosophy and thinking by adhering to the client's philosophy and ideas about what the cafeteria should be like. Therefore, it cannot be said that all of the food to be served in such a "closed-space" large-volume cafeteria must be prepared in advance as in (a). However, given the unique characteristics of these "enclosed" mass-serve cafeterias, as discussed above, it is clear that prompt service with minimal wait time, i.e., (a) and (b), is desirable. Furthermore, it is possible that some of the dishes described in (c) and (d) may be intentionally or strategically offered to add variety to the cafeteria's operations or to offer a fresh, unusual menu to the cafeteria's employees. However, the need to minimize wait times and provide meals without customer dissatisfaction is an important aspect of the unique characteristics of a "enclosed" mass-serve cafeteria within a "closed organization." Compared to the acceptable wait times at typical restaurants or the generally accepted wait times, it is generally desirable and common sense to provide food in even shorter times. For example, it is common sense that a typical "enclosed" mass-serve cafeteria requires a much shorter wait time than the time required for ordering, paying, waiting at the pickup location, and receiving the ordered food and drink at a typical fast-food hamburger restaurant.

[0026] Even more serious than the profitability issues facing cafeteria operators is the SDGs' mandate for "zero waste." As mentioned above, excessive waste puts pressure on cafeteria operators' profits, worsening their bottom lines and causing financial difficulties. However, even if the actual number of customers or orders differs significantly from the predicted figures, resulting in high waste volumes, does this problem simply go away if cafeteria operators simply accept lower profit margins? This is not the case. With current social norms and demands placing a strong emphasis on the SDGs, high waste is perceived as a serious problem, and is being taken even more seriously than before. Waste is not limited to cafeteria operators; it is beginning to have a significant impact on the "closed organizations"—companies and schools—that own enclosed cafeterias. In other words, the mere existence of high waste in the cafeterias within a company or school is increasingly being viewed as a serious SDG-related issue. In other words, this problem is not limited to cafeteria operators; the "closed organizations" themselves, such as the companies and schools that own cafeterias, tend to place a high priority on "zero waste." The more these "closed organizations" emphasize the SDGs and "zero waste," the more important it becomes for cafeteria operators to promote "zero waste" in their own cafeterias in line with these goals. In this context, if a cafeteria overestimates the amount of food they expect to eat, or if they adopt a policy of ample advance preparation, pre-cooking many dishes, or preparing too many ingredients in advance, and the actual number of orders is far less than the forecast, this can result in a significant amount of waste. To prevent this, increase convenience, and improve the quality of cafeteria service, it is desirable to make as accurate advance forecasts as possible. It is also important to be able to quickly and accurately correct forecasts based on actual customer traffic and order status from opening until closing time. The present inventors believe that a forecasting system with the functionality to meet these objectives is needed.

[0027] AI prediction systems that only make advance predictions do not have the functionality to actually open a restaurant, take in real-time customer data from the order reception system (for example, the restaurant's cash register), and continue to adjust and correct the predicted number of customers and menu orders until closing time in real time.Operations are initially started based solely on the targeted advance prediction, and if the number of customers or orders is lower than predicted, it is usually the head chef, or human intuition, who determines this, and the amounts predicted in advance using the prediction system are then revised again by humans, which is the general practice at present, and again, this is simply relying on human intuition.

[0028] In a cafeteria that provides food services in this "closed organization," users are people from a limited location, and the same person often has the habit of coming around a specific time, for example, during the one-hour lunch break. Therefore, if we can firmly grasp the data of each individual and know in advance the probability that a specific individual will come to the cafeteria at a specific time during the one-hour lunch break, we believe that it should be possible to fairly accurately predict the time of visit to the cafeteria for the entire number of customers belonging to a "closed organization" organization, such as a specific group of employees or students, or the number of customers for a specific group, as the sum of the probabilities of such employees. Currently, there are no functions in the world for predicting the number of customers at a "closed organization" or the sales of menu items that record the behavior of such specific individuals or specific groups over a long period of time, effectively utilizing that data, creating various models, and correcting them through learning.

[0029] In this way, one of the objectives of this disclosure is to simultaneously aim for zero waste and improved services.

[0030] Furthermore, the conventional technology does not have a function that links the cafeteria cash register with the calculation part that calculates the predicted numbers in advance, and based on this link, uses data recorded about each individual to constantly revise the predictions of the number of customers and menu orders for the remaining time, and only makes predictions before the cafeteria opens.

[0031] Other conventional systems are also capable of predicting the number of customers and best-selling items for a particular date, day of the week, season, etc., and predicting the order volume of a menu item using machine learning, etc. However, such conventional systems do not have the function of predicting the supply and demand of a large-volume cafeteria in a "closed location" within a "closed organization," based on real-time data received from an order receiving system (e.g., a cafeteria cash register) and the behavioral predictions of individual employees, and are therefore unable to predict the large number of customers and large orders at the cafeteria from moment to moment.

[0032] The previous system did not focus on each individual employee, nor did it have an AI function to predict and determine the number of menu items to be shipped on a moment-by-moment basis, such as predicting the order volume for the remaining cafeteria operating hours at 12:15. Rather, it was a system that only supported predictions of sales figures for the entire cafeteria in a specific season and on a specific date, based on past data.

[0033] Furthermore, the "menu sales forecasting systems" currently used by these cafeteria operators do not include data on individual employees or specific groups, as described above, and do not have a function to link with the cafeteria registers used by these employees for daily payments. Therefore, while sales forecasts are made in advance for the entire cafeteria, there is no function to measure and predict individual employees or specific groups, such as the probability of what they will eat and when. Furthermore, because the systems are not linked to the cafeteria registers, there is no function to keep track of data on what time a particular employee will arrive and what they will order. Therefore, the systems currently used by these cafeteria operators are based solely on the human judgment of the head chef or other personnel regarding the number of customers and menu orders for the entire cafeteria. The cafeteria registers constantly record the behavior of individual employees, and the system does not have a function to constantly correct any discrepancies with the forecast, or to constantly adjust the preparation quantities for each menu item for the entire cafeteria.

[0034] Furthermore, with the development of AI technology in recent years, systems that can predict what may happen in the future from moment to moment are becoming more sophisticated, even if there is insufficient prior data. Traditionally, menu delivery forecasts used by restaurant operators were based on past consumption data from many restaurants, and the relationship between weather, season, and other factors and order forecasts for various menu items was predicted using methods such as the Pearson correlation coefficient. However, this was largely based on traditional ideas in probability theory and not on "Bayesian inference," which is based on the idea of ​​calculating future probabilities based on ever-changing human behavior and actual events. One core aspect of AI computational processing that has been developing in recent years is based on "Bayesian inference," which recalculates the ever-changing probabilities of future events. Until now, in such "closed cafeterias" where "large quantities are supplied in a closed space," there has been no function that takes into account various factors and links with the cafeteria's cash registers, etc. to collect data in real time about the menu items ordered by customers from moment to moment, and based on that, creates a "prior prediction" as a prior distribution using AI from various data before opening, processes this using Bayesian statistical methods from moment to moment, and creates a new "order prediction" as a posterior distribution each time as a "Bayesian update."

[0035] In cafeterias, whose mission is to serve a large number of meals to many customers in a limited time, particularly in employee cafeterias, student cafeterias, and other "closed places" in so-called "closed organizations" such as companies and schools, where a large number of members of that "closed organization" come and eat at the same time, it is extremely important to predict the menu and the order quantities of each item in advance, and to predict the number of items to be served over time.Existing technology already exists, but there are AI-based calculation systems (including computer systems that perform calculations based on hypotheses) that support the advance prediction of the total menu quantity for cafeterias that serve a large amount of food in "closed places."However, it is not possible to analyze individual employees, and there is no system that has the function of constantly predicting and displaying appropriate values ​​for the number of items to be prepared and the actual number to be cooked for a given menu item, depending on the number of customers and the number of people eating and drinking, which changes from moment to moment along the time axis. There are currently various systems that correlate and predict the quantity and serving time of the entire menu to be served in a cafeteria, but these systems rely heavily on the intuition of the human chef, who constantly monitors the actual customer visits and orders, and constantly adjusts and determines the appropriate quantity. Furthermore, even if a system that analyzes the quantity of each menu item to be served in the entire cafeteria is based on existing technology, it cannot predict the time that each individual customer will arrive at the cafeteria based on data on each employee's attributes, preferences, and past behavior, nor does it have the functionality to predict which menu items to order based on preferences. Furthermore, even if there are current systems that predict the number of customers and order quantities for the entire cafeteria, they only predict the entire cafeteria, and do not incorporate predictions of the behavior of each individual employee.

[0036] [The order volume forecasting system disclosed herein] According to one embodiment of the present disclosure, there is provided an order quantity prediction system for a restaurant, comprising a past order history database, an external factor database, an order pattern model generation unit, an order quantity prediction calculation unit, and an order quantity update calculation unit, The past order history database stores the past order history of each user at the restaurant in association with the attribute information of the user and a time series of the order history, The external factor database stores any information other than the order history in association with a time series; the order pattern model generation unit generates, based on the information stored in the past order history database and the information stored in the external factor database, an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series; The order quantity prediction calculation unit (1) Calculating an order prediction model for each user along a time series for the specific day based on the order pattern model for each user, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each user. (2) calculating an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each attribute; and / or (3) calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; The order quantity update calculation unit (1) Calculating an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user using a Bayesian statistical method based on the actual order history of each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user. (2) calculating an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute based on the actual order history for each user on a specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute; and / or (3) updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant based on the time series on the specific day; An order quantity forecasting system is provided.

[0037] According to another embodiment of the present disclosure, the above order quantity forecasting system; an order receiving system that transmits actual order histories for each user on a specific day to the order quantity prediction system; a display system for displaying the restaurant-wide order prediction model and / or the restaurant-wide updated order prediction model; A restaurant ordering system is provided, comprising:

[0038] According to one embodiment of the present disclosure, even in a large-volume restaurant located in a "closed location" within a "closed organization," predictions are made based on data from individual employees, which significantly improves the accuracy of advance predictions of customer arrival times, order predictions for each time and menu item, and customer and order predictions for the entire restaurant, thereby advantageously significantly improving the amount of advance order forecasts. According to one embodiment of the present disclosure, it is advantageous in that order volumes can be predicted even in a restaurant with the unique characteristic that a large number of certain employees habitually eat and drink regularly (for example, almost daily) and further visit and eat intensively during a limited time period.

[0039] According to one embodiment of the present disclosure, by analyzing the real-time customer visit status of each employee, order status, and data on external factors such as weather, which can be obtained at the cafeteria register, in a time series manner, it becomes possible to predict the optimal menu selection, the number of dishes, the number of dishes to be cooked according to the time, and the number of ingredients required, etc., and present this information to the entire kitchen and to the chefs and staff, including the head chef. In particular, according to one embodiment of the present disclosure, in a cafeteria in a "closed place" such as an employee cafeteria or a student cafeteria, which is characterized by the habitual visits of the same staff members every day, the habits and preferences of each staff member that are unique to that closed place are taken into consideration, and in a very limited cafeteria where a large number of people eat and drink at once in a limited time that is unique to eating and drinking in such a closed place, the optimal number of dishes to be cooked for the entire cafeteria can be predicted from moment to moment based on the habits and data of each individual, and on data linked to each individual, such as the actual time of arrival of that individual on that day and the ordered menu, thereby preventing the cooking of wasted food, preventing waste due to overcooking, providing food in a timely manner, preventing food from going cold by preparing it in advance, and / or increasing customer satisfaction with the cafeteria.

[0040] According to one embodiment of the present disclosure, it is advantageous to be able to predict, based on past data, the approximate time each employee will arrive at the cafeteria, and similarly, to predict, based on past order history, the characteristics, types, and attributes of the menu items they are likely to order. Furthermore, according to one embodiment of the present disclosure, by using functions including box-and-whisker analysis and Pearson correlation coefficient calculation, it is advantageous to be able to improve the accuracy of predictions made in advance for each employee on a daily basis. According to the present disclosure, for example, based on the past history of an individual who tends to eat food A on sunny days but food B on cold days depending on external factors (e.g., weather data), it is possible to predict with a certain probability the dishes that will actually be ordered and the time of ordering them in the future under the influence of such external factors, and it is advantageous to be able to further improve the accuracy of this probability.

[0041] According to one embodiment of the present disclosure, even if orders are predicted in advance, once the restaurant actually opens, many customers arrive, and each customer orders their favorite dishes, the actual number of customers and order quantities for each menu item over time may differ from the prediction. In such cases, the predicted quantities can be compared with ever-changing data collected from the order reception system (e.g., the restaurant's register), i.e., the actual number of customers and orders received from the restaurant's register each time. This embodiment advantageously performs processing based on Bayesian statistics and allows for new predictions of events that will occur after that time. Therefore, according to the present disclosure, it is not necessary to stick to the order predictions made before the restaurant opens and continue preparing a specific menu until the end of opening hours. Instead, the head chef can accurately prepare menu items based on the ever-changing final quantities or estimated interim time requirements while viewing a display that displays newly corrected order predictions based on Bayesian updates, etc., that change from moment to moment. This is particularly advantageous in that it minimizes the discrepancy between the final order quantity at the time the restaurant closes and the final arrival point, thereby minimizing or reducing to zero the amount of waste generated by the restaurant.

[0042] According to one embodiment of the present disclosure, the system predicts the daily visitor times and menu items ordered for each individual based on data on the so-called "habits" and "preferences" of each individual, whether conscious or unconscious, obtained through long-term observation and recording, and utilizes data obtained in real time in conjunction with the restaurant's register, which is advantageous in that predictions for the entire restaurant can be made with higher accuracy than conventional prediction systems. In other words, the predictions used by many restaurant management companies in the past predict the number of customers and menu items for the restaurant as a whole, and do not have the function to make predictions for individual employees. However, according to the present disclosure, individual data can be collected from the linked order reception system (e.g., the restaurant's register), which is advantageous in that data such as individual habits can be generated, calculated, or recorded from various angles.

[0043] According to one embodiment of the present disclosure, it is advantageous in that it is possible to predict the behavior of each employee in the cafeteria with extremely high accuracy based on data on the habits of each employee. For example, if it is known that a certain employee has the habit of coming to the cafeteria around 12:15 and that this person is likely to order hot noodles on cold days, it is advantageous in that it is possible to predict the eating behavior of each employee in the cafeteria on that day by applying rules and external factors such as the weather and day of the week.

[0044] According to one embodiment of the present disclosure, it is possible to take into consideration the habits and preferences of each employee, and it is also advantageous in that it is possible to make full use of data recorded as employee habits, including preferences and nutritional imbalances such as whether the employee eats breakfast at around 10:00, whether they are accustomed to coming to the cafeteria around 11:30 just before it opens, whether they often come at around 11:45, or whether they are accustomed to coming after 12:30 when the peak is over, whether they have the habit of eating ramen once a week, whether they have the habit of eating salad every day, whether they eat a lot of carbohydrates, whether they eat meat or never eat fish, etc. Furthermore, according to one embodiment of the present disclosure, it is also possible to utilize data that corresponds to differences in behavior patterns according to various groups in a company. For example, within the same company, company regulations may stipulate that the lunch hours for Division P are from 11:30 to 12:30, for Division Q from 12:15 to 1:15, and for Division R from 12:30 to 1:30. This has the advantage of making it possible to utilize data that reflects the differences in behavior patterns of each group.

[0045] According to one embodiment of the present disclosure, since the above data is available for all employees belonging to the closed space and the above predictions are possible, it is advantageous in that the sum of the data can be used to predict with increased accuracy the number of customers and the number of orders for each menu item in the entire cafeteria at any given time. In other words, according to one embodiment of the present disclosure, it is advantageous in that it is possible to make predictions that are much more accurate and closer to reality than predictions made using existing technologies that calculate and predict as a whole without focusing on individuals. In other words, according to one embodiment of the present disclosure, it is possible to focus on employees of a closed organization and, based on individual data over a long period of time, apply correlations between different types of data or machine learning for each individual to determine the relationships between such data, thereby making it possible to make predictions for each individual even closer to reality and, as a sum, to improve the accuracy of predictions for the entire cafeteria.

[0046] According to a preferred embodiment of the present disclosure, the order volume update calculation unit performs the following (1) to (3): (1) Calculating an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user using a Bayesian statistical method based on the actual order history of each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user. (2) calculating an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute based on the actual order history for each user on a specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute; and / or (3) updating the order prediction model for the entire restaurant based on the actual order history of each user on a specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant based on the time series on the specific day; and (iii) perform at least two (preferably three) calculations selected from the group consisting of: In this way, the order quantity update calculation unit performs two (preferably three) calculations, and each calculation is weighted, after which an order prediction model for the entire restaurant on a specific day is calculated. Therefore, by changing the weightings for calculations using a Bayesian statistics-based method (e.g., Bayesian inference) focusing on individuals, calculations using a Bayesian statistics-based method focusing on attributes (gender, age, organization to which one belongs, etc.), and calculations using a Bayesian statistics-based method for the entire restaurant, this is particularly advantageous in that it makes it possible to calculate an order prediction model that is even more convenient for restaurants. For example, if only Factory P is closed at a certain company, by focusing only on the attributes of the employees working at Factory P and assigning a large weight to (β), it may be possible to quickly and accurately calculate an order prediction model for the entire cafeteria with less calculation effort.In the event of a major natural disaster or other calamity that far exceeds the scope of individuals or attributes, that is, an incident that has an extremely large impact on the operation of the entire cafeteria regardless of individuals or attributes, it is possible to assign an extremely large weight to (γ), which is particularly advantageous in that it greatly reduces the amount of calculation effort and allows for quick and accurate calculation of an order prediction model for the entire cafeteria.

[0047] According to one embodiment of the present disclosure, the value of the Pearson correlation coefficient with various data can be changed daily, and / or the accuracy of the box-and-whisker analysis can be updated daily. This advantageously allows the order pattern model used when performing function calculations based on this updated accuracy to be evolved using machine learning or the like. Furthermore, according to one embodiment of the present disclosure, this evolved order pattern model can be advantageously stored, for example, in an order pattern model database. That is, according to one embodiment of the present disclosure, the prediction accuracy for this closed restaurant can be improved daily as time passes and the amount of data learned increases, leading to further evolution.

[0048] According to one embodiment of the present disclosure, the system is characterized by being linked to an order receiving system (e.g., a cafeteria register), and has a function that can retrieve data on a daily basis regarding the time when a specific member of a "closed organization" such as a company or university, such as an office worker or student, actually visits the cafeteria on that day, places an order, and what they eat (what they ordered), each time that member actually passes through the cafeteria register. According to one embodiment of the present disclosure, this function can advantageously calculate a correction value that is calculated every moment using a Bayesian statistical method based on newly retrieved data every moment, such as the hourly order forecast quantity created before the cafeteria opens and the actual time when each member actually arrives at the cafeteria (order time) and the menu contents of what they ordered after the cafeteria opens. In other words, according to one embodiment of the present disclosure, in a restaurant that supplies large quantities in a "closed space," for example, it is possible to obtain data on who entered the restaurant and what they ordered up to that time, for example, at 12:15, and from this data on the specific times and menu orders for each employee, it is possible to constantly re-predict the final total for each menu item until closing time around 1:30 p.m. This has the advantage that the predicted values ​​before opening, as compared with the prior art, can be revised while the restaurant is open, resulting in highly accurate predicted values ​​being produced each time, which eliminates the need to prepare unnecessary ingredients and brings the restaurant closer to zero waste.

[0049] An embodiment of the order quantity prediction system of the present disclosure will be described in detail below with reference to the drawings. Note that the order quantity prediction system of the present disclosure is not limited to the embodiment shown in the drawings described below.

[0050] First, in many cases, the employees and students (users) who belong to a company, university, or the like are largely fixed. While people leave and new members join daily, monthly, or yearly, it is common for the majority of members to be fixed. Such places are considered to have fixed users, or, although there is some turnover, many users remain fixed for a certain period of time. In this sense, they are "closed organizations," and cafeterias within such organizations that primarily cater to a certain set of users can be called "closed organization cafeterias." The subject of this disclosure is primarily a "closed organization cafeteria" within such "closed organizations," and is intended to be places such as employee cafeterias and student cafeterias. These cafeterias are characterized by the fact that users of the organization gather in large numbers to eat during a limited time, such as lunchtime, and therefore are "cafeterias with a large supply of food in a closed organization," and the present disclosure is primarily intended to be an order quantity prediction system for such cafeterias.

[0051] Users who are members of so-called "closed organizations" regularly eat lunch and other meals in a "closed place" every day, mainly around lunchtime. A "closed place" is where such members work or study, and typically has a cafeteria to support these activities. The cafeteria typically has a kitchen staffed by chefs who provide meals primarily for the members to eat and drink. These meals are provided by a cafeteria operator or similar entity that has concluded a contract for outsourcing with the "closed organization" that operates the closed place, such as a company or university. In other words, the "operation of a cafeteria in a closed place" actually involves the company or university acting as the client, a selected cafeteria operator acting as the contractor, and the contractor providing meals to the contractor.

[0052] In this disclosure, a "closed organization" refers to an organization to which specific members (i.e., users) belong. Typical examples of closed organizations include companies, public institutions, high schools, and universities. Note that closed organizations are intended to encompass spaces (e.g., shopping malls) where it is common to use cards, such as membership cards, that store user attribute information. In this disclosure, a "closed place" refers to a place used by users who belong to the closed organization. In this disclosure, a "closed dining hall" refers to a dining hall where the majority of users are specific individuals. Typical examples of closed dining halls include, for example, dining halls within a company and dining halls at universities (e.g., student cafeterias). Furthermore, a "closed dining hall" also includes a dining hall (e.g., a dining hall located in a shopping mall, including a food court) where it is common to order a menu item by presenting a card, such as a membership card, that stores user attribute information.

[0053] In this disclosure, "user" is used interchangeably with "member" and the like.

[0054] "Attribute information" in this disclosure refers to any attribute information of a user. Examples of such attribute information include, but are not limited to, gender, age, date of birth, hometown, name of organization (e.g., company name, school name), user identification information in the organization (e.g., employee number, student number, etc.), job title (e.g., position), department (e.g., group name in the organization such as the name of the business division, factory, or faculty to which the user belongs), employment status (full-time, part-time, etc.), etc.

[0055] In the present disclosure, an order pattern model "per attribute," an order prediction model "per attribute," an updated order prediction model "per attribute," etc. refer to a model generated or calculated for each attribute. The "attribute" in "per attribute" is synonymous with the above-mentioned attribute information. For example, if the "attribute" is gender, an order pattern model etc. for each gender (male, female) may be generated or calculated; if the "attribute" is job position, an order pattern model etc. for each job position (e.g., general employee, manager, executive, etc.) may be generated or calculated; and if the "attribute" is department, an order pattern model etc. for each department (e.g., P Business Division, Q Business Division, etc.) may be generated or calculated.

[0056] [Configuration of a restaurant order quantity prediction system] An example of an order quantity prediction system according to the present disclosure will be described with reference to FIG.

[0057] The order volume prediction system 1 of the present disclosure includes at least a past order history database 101, an external factor database 102, an order pattern model generation unit 108, an order volume prediction calculation unit 109, and an order volume update calculation unit 110. The order volume prediction system 1 may further include an order reservation database 103, a current day order history database 104, an order pattern model database 105, an order prediction model database 106, an updated order prediction model database 107, and / or a correction parameter calculation unit 111. The order volume prediction system 1 may also include a communication unit 112, an input / output interface unit 113, etc., as appropriate.

[0058] (Past order history database) The past order history database 101 is a database that stores the past order history of each user at restaurants in a closed organization in association with attribute information of the users who belong to the organization, in chronological order.

[0059] The past order history database 101 may store, for example, the data shown in FIG. 12, which will be described later. In the example shown in FIG. 12, user attributes (for example, gender, age, job title, department, etc.) and past order information for each user (order date and time, ordered menu item, calorie and nutritional information for the menu item, etc.) are stored. The past order history database 101 may freely set the information to be stored in accordance with the policy, thinking, and philosophy of a "closed organization." Note that information related to user attributes does not have to be stored in the past order history database, and may be stored in another database, etc. In that case, the past order history stored in the past order history database and the information related to user attributes stored in the other database, etc. may be linked by an identification code, etc.

[0060] (External Factor Database) The external factor database 102 is a database that stores any information other than the order history in a time series. Examples of information stored in the external factor database include weather information, disaster information (e.g., earthquakes, typhoons, river flooding, tsunamis, etc.), economic information (e.g., currency exchange information, etc.), traffic information (e.g., traffic congestion information, delay information, etc.), pandemic information, crisis information (e.g., terrorism, civil unrest, missile attacks, war, etc.), labor dispute information (e.g., strikes, etc.), national or local government event information (e.g., a local marathon, an international competition, etc.), and event information within the organization (preferably the closed organization) (e.g., training for the P division only, shutdown of the Q factory, new employee training, training for women, management training, general manager meetings, training for general employees only, the date of the general shareholders' meeting, the first day of work for new employees, the scheduled date of the entrance ceremony, etc.), and the like. These may be stored alone or in any combination of two or more types. This information may be public information and may be set to be imported into the external factor database 102 at any frequency (for example, hourly or daily). According to one embodiment of the present disclosure, the external factor database stores at least one selected from the group consisting of weather information and event information in the organization. According to a more preferred embodiment of the present disclosure, the external factor database stores weather information.

[0061] (Order reservation database) The order reservation database 103 is a database that stores order reservation information for a specific day for each user. The order reservation database 103 can store a user's attribute information and the user's order reservation information for a specific day in association with each other. The order reservation database 103 can store order reservation information for a specific day transmitted from a user terminal used by the user, for example. The order reservation database 103 may store order reservation information for a single specific day, or may store order reservation information for multiple specific days. The order reservation database 103 may store, for example, information exemplified in FIG. 17.

[0062] (Today's order history database) The same-day order history database 104 is a database that stores actual order histories for a specific day in association with user attribute information. The order reservation database 104 may store, for example, the information shown in FIG. 22.

[0063] (Order pattern model database) The order pattern model database 105 is a database that stores order pattern models for each user along a time series, order pattern models for each attribute along a time series, and / or order pattern models for the entire restaurant along a time series, which are generated by the order pattern model generation unit 108 described below.

[0064] The order pattern model database 105 may also store correction parameters calculated by a correction parameter calculation unit 111, which will be described later.

[0065] (Order prediction model database) The order prediction model database 106 is a database that stores order prediction models for each user along a time series on a specific day, order prediction models for each attribute along a time series on a specific day, and / or order prediction models for the entire restaurant along a time series on a specific day, calculated by the order quantity prediction calculation unit 109 described below.

[0066] (Updated order prediction model database) The updated order prediction model 107 is a database that stores an updated order prediction model for each user along a time series on a specific day, an updated order prediction model for each attribute along a time series on a specific day, and / or an updated order prediction model for the entire restaurant along a time series on a specific day, all calculated by the order quantity update calculation unit 110 described below.

[0067] (Order pattern model generation unit) The order pattern model generation unit 108 can generate an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series, based on the information stored in the past order history database 101 and the information stored in the external factor database 102.

[0068] The order pattern model generation unit 108 can generate the order pattern model by, for example, regression analysis (e.g., multiple regression analysis), etc. The order pattern model can be generated using calculations using calculation software, etc., machine learning (e.g., deep learning, etc.), etc.

[0069] (Order volume prediction calculation section) The order quantity prediction calculation unit 109 can calculate an order prediction model for each user along a time series on a specific day, based on the order pattern model for each user, the menu for the specific day, and information related to the specific day stored in the external factor database 102. Furthermore, the order quantity prediction calculation unit 109 can calculate an order prediction model for the entire restaurant along a time series on a specific day, based on the sum of the order prediction models for each user.

[0070] The order quantity prediction calculation unit 109 can also calculate an order prediction model for each attribute along a time series on a specific day, based on the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database 102. Furthermore, the order quantity prediction calculation unit 109 can calculate an order prediction model for the entire restaurant along a time series on a specific day, based on the sum of the order prediction models for each attribute.

[0071] In addition, the order quantity prediction calculation unit 109 can calculate an order prediction model for the entire restaurant along a time series on a specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database 102.

[0072] The order quantity prediction calculation unit 109 may further calculate each of the order prediction models based on the order reservation information for each user stored in the order reservation database 103, as necessary.

[0073] The order quantity prediction calculation unit 109 may further calculate each of the order prediction models based on correction parameters calculated by a correction parameter calculation unit 111, which will be described later, as necessary.

[0074] The order quantity prediction calculation unit 109 can calculate each of the order prediction models by, for example, machine learning (for example, deep learning) or the like.

[0075] (Order quantity update calculation section) The order quantity update calculation unit 110 can calculate an updated order prediction model for each user along a time series, which is obtained by updating the order prediction model for each user using Bayesian statistics techniques based on the actual order history of each user on a specific day. Furthermore, the order quantity update calculation unit 110 can calculate an updated order prediction model for the entire restaurant along a time series on a specific day, based on the updated order prediction model for each user.

[0076] The order quantity update calculation unit 110 can also calculate an updated order prediction model for each attribute along a time series, which is obtained by updating the order prediction model for each attribute using Bayesian statistical techniques based on the actual order history of each user on a specific day. Furthermore, the order quantity update calculation unit 110 can calculate an updated order prediction model for the entire restaurant along a time series on a specific day, based on the updated order prediction model for each attribute.

[0077] In addition, the order quantity update calculation unit 110 can calculate an updated order prediction model for the entire restaurant along a time series on a specific day, which is obtained by updating the order prediction model for the entire restaurant using Bayesian statistical techniques based on the actual order history of each user on a specific day.

[0078] The "Bayesian statistical method" in the present disclosure is not particularly limited as long as it is a method based on Bayesian statistics. Examples of Bayesian statistical methods include Bayesian inference; state space models (e.g., Kalman filter, extended Kalman filter, non-Gaussian filter, particle filter, Bayesian structural time series model); time-series recurrent neural network (RNN) models (e.g., simple RNN, LSTM, GRU, encoder-decoder model, bidirectional recurrent model); TFT (Temporal Fusion Transformer) model; Prophet model; Bayesian dynamic linear model; Bass model; stochastic volatility model; Markov switching model; chaos analysis (e.g., logistic map); and other time series prediction models based on Bayesian statistics. Bayesian statistical methods are currently being researched daily around the world, and new methods are being developed at a rapid pace, with various methods being published daily. These methods are also intended to be encompassed by the present disclosure. According to one embodiment of the present disclosure, the Bayesian statistical method is Bayesian inference.

[0079] If necessary, the order quantity update calculation unit 110 may further calculate each updated order prediction model based on the information stored in the external factor database 102 and each order pattern model stored in the order pattern model database 105.

[0080] The order volume update calculation unit 110 can calculate the above-mentioned post-update order prediction models by, for example, machine learning (for example, deep learning) or the like.

[0081] (Correction parameter calculation section) The correction parameter calculation unit 111 can calculate correction parameters to be used for correcting the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant, based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on a specific day.

[0082] The correction parameter calculation unit 111 can calculate the correction parameters by, for example, machine learning (for example, deep learning) or the like.

[0083] (Communications Department) The communication unit 112 can transmit and receive data and the like to and from each other by wireless communication and / or wired communication via the communication network NW.

[0084] (Input / output interface section) The input / output interface unit 113 is connected to a built-in and / or externally connected input unit 114 and / or output unit 115, and can control the input unit 114 and / or output unit 115 (note that the input unit 114 and the output unit 115 are omitted in FIG. 1). Examples of the input unit 114 include a keyboard, a mouse, a microphone, etc. Examples of the output unit 115 include a display, a monitor, a speaker, etc.

[0085] (others) The order quantity prediction system 1 may include functional blocks other than the above-described functional blocks as needed. The order quantity prediction system 1 may also include, for example, a database that stores menus for specific days.

[0086] 1 shows an example in which each of the above-described functional blocks is provided in one computing device (computer). In this case, the single computing device provided with each of the above-described functional blocks may constitute the order quantity prediction system 1 in one embodiment of the present disclosure. Furthermore, each of the above-described functional blocks may be separated and present in two or more computing devices, and these two or more computing devices may be connected to each other so that they can communicate with each other via a communication network NW or the like. In this case, the two or more computing devices may constitute the order quantity prediction system 1 in one embodiment of the present disclosure.

[0087] The order quantity prediction system 1 may be a stand-alone system or a cloud system.

[0088] The order quantity prediction system 1 may be able to mutually send and receive data, etc. with the order receiving system 2, the display system 3 and / or the user terminal 4 via wireless communication and / or wired communication via the communication network NW (Figure 8).

[0089] [Processing flow of restaurant order quantity prediction system] An example of the processing flow of the order volume prediction system will be described with reference to FIG.

[0090] The information stored in the past order history database (past order history) and the information stored in the external factor database (external factor information) are input to the order pattern model generation unit (S101, S102). Note that in the example of Fig. 2, the information relating to the past order history is input first, followed by the external factor data, but these may be input to the order pattern model generation unit in any order, or the information relating to the past order history and the external factor data may be input to the order pattern model generation unit simultaneously.

[0091] The order pattern model generation unit generates an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series (S103).

[0092] The order quantity prediction calculation unit receives the generated order pattern model for each user along a time series, the order pattern model for each attribute along a time series, and / or the order pattern model for the entire restaurant along a time series, as well as the menu information for the specific day and the order reservation information (S104, S105). Note that this information may be input in any order or simultaneously.

[0093] Based on the input information, the order quantity prediction calculation unit calculates an order prediction model for each user along a time series on the specific day, an order prediction model for each attribute along a time series on the specific day, and / or an order prediction model for the entire restaurant along a time series on the specific day (S106).

[0094] On a specific day, the order quantity update calculation unit receives the calculated order prediction model for each user along a time series for the specific day, the order prediction model for each attribute along a time series for the specific day, and / or the order prediction model for the entire restaurant along a time series for the specific day, as well as the actual order history for each user on the specific day, and, as necessary, the order pattern model for each user along a time series, the order pattern model for each attribute along a time series, and / or the order pattern model for the entire restaurant along a time series (S107).Based on this input information, the order quantity update calculation unit calculates the updated order prediction model for each user along a time series, the updated order prediction model for each attribute along a time series, and / or the updated order prediction model for the entire restaurant along a time series using a Bayesian statistical method (S108).

[0095] If necessary, the updated order prediction model for all dining halls and the actual order volume for all dining halls on a specific day are input to a correction parameter calculation unit. Based on the input information, the correction parameter calculation unit may calculate correction parameters (not shown) used to correct the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for all dining halls. The calculated correction parameters may be sent to and stored in, for example, an order pattern model database (not shown).

[0096] (Processing flow of the order pattern model generation unit) An example of the processing flow of the order pattern model generation unit will be described in more detail with reference to FIG.

[0097] The order pattern model generation unit 108 receives input of information relating to past order history stored in the past order history database and information (external factor data) stored in the external factor database (S201, S202). Note that in the example of Fig. 3, the external factor data is input after the information relating to past order history is input, but these may be input to the order pattern model generation unit 108 in any order, or the information relating to past order history and the external factor data may be input to the order pattern model generation unit 108 at the same time.

[0098] The order pattern model generation unit 108 performs multiple regression analysis processing based on the input information to generate an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series (S203, S204). Note that the results obtained by the multiple regression analysis processing may be transmitted to and stored in any storage unit, database, etc. provided in the order quantity prediction system, as needed. Furthermore, each of the generated order pattern models may be transmitted to and stored in an order pattern model database provided in the order quantity prediction system, as needed.

[0099] (Processing flow of the order volume prediction calculation unit) An example of the processing flow of the order volume prediction calculation unit will be described in more detail with reference to FIG.

[0100] The order volume prediction calculation unit 109 receives as input each of the order pattern models generated by the order pattern model generation unit 108 or stored in the order pattern model database, the menu for the specific day, and, as necessary, order reservation information and external factor information (S301 to S304). Note that, although the order pattern model is input first in the example of Fig. 4, these may be input to the order volume prediction calculation unit 109 in any order, or all of these may be input to the order volume prediction calculation unit 109 at the same time.

[0101] Based on the input information, the order quantity prediction calculation unit 109 calculates an order prediction model for each user along a time series for the specific day, an order prediction model for each attribute along a time series for the specific day, and / or an order prediction model for the entire restaurant along a time series for the specific day (S305). That is, when the order quantity prediction calculation unit 109 has calculated an order prediction model for each user along a time series for the specific day, it calculates an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the calculated order prediction models for each user (S305). Furthermore, when the order quantity prediction calculation unit 109 has calculated an order prediction model for each attribute along a time series for the specific day, it calculates an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the calculated order prediction models for each attribute (S305). The calculated order prediction models may be transmitted to and stored in an order prediction model database included in the order quantity prediction system, as necessary.

[0102] The order quantity prediction calculation unit 109 may repeatedly perform the calculations of each order prediction model as necessary and at any frequency until the specific date arrives.

[0103] (Processing flow of the order volume update calculation unit) An example of the processing flow of the order volume update calculation unit will be described in more detail with reference to FIG.

[0104] The order quantity update calculation unit 110 inputs the order prediction models calculated as above, the actual order history of each user on the specific day, and, as necessary, the order pattern models (S401 to S403). Note that, in the example of Fig. 5, the order pattern model is input after the order history for the day, but these may be input to the order quantity update calculation unit 110 in any order, or all of them may be input to the order quantity update calculation unit 110 at the same time.

[0105] The order quantity update calculation unit 110 calculates, based on the input information, an updated order prediction model for each user along a time series, an updated order prediction model for each attribute along a time series, and / or an updated order prediction model for the entire restaurant along a time series, using a Bayesian statistical method (S404). That is, when the order prediction model for each user is input, the order quantity update calculation unit 110 calculates an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user according to a Bayesian statistical method based on the actual order history of each user on a specific day, and calculates an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user (S404). Furthermore, when the order prediction model for each attribute is input, the order quantity update calculation unit 110 calculates an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute using a Bayesian statistical method based on the actual order history of each user on the specific day, and calculates an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute (S404).When the order prediction model for the entire restaurant is input, the order quantity update calculation unit 110 calculates an updated order prediction model for the entire restaurant along a time series for the specific day by updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method (S404).

[0106] On a particular day, the order history for that day is updated every time a user places an order at a restaurant, so the order quantity update calculation unit 110 can calculate the updated order prediction models at any frequency. The calculated updated order prediction models may be transmitted to and stored in an updated order prediction model database provided in the order quantity prediction system, as necessary.

[0107] (Processing flow of the correction parameter calculation unit) An example of the processing flow of the correction parameter calculation unit will be described in more detail with reference to FIG.

[0108] At any timing (for example, after the restaurant closes on a specific day), the correction parameter calculation unit 110 inputs the updated order prediction model for the specific day (preferably, the latest updated order prediction model for the specific day) and the actual order history for each user on the specific day (preferably, the actual order history for each user on the specific day for all restaurants) (S501-S502). Note that in the example of Fig. 6, the order history is input after the updated order prediction model, but these may be input to the correction parameter calculation unit 110 in any order, or may all be input to the correction parameter calculation unit 110 at the same time.

[0109] The correction parameter calculation unit 110 can calculate the correction parameters based on the input information, for example, by machine learning (deep learning, etc.) The calculated correction parameters may be transmitted to and stored in any storage unit or database (for example, an order pattern model database) provided in the order quantity prediction system, as necessary.

[0110] [Dining Order System] According to one embodiment of the present disclosure, the order quantity forecasting system; an order receiving system that transmits actual order histories for each user on a specific day to the order quantity prediction system; a display system for displaying the restaurant-wide order prediction model and / or the restaurant-wide updated order prediction model; A restaurant ordering system is provided, comprising:

[0111] An example of a restaurant ordering system will be outlined with reference to Fig. 8. The restaurant ordering system 11 may be able to transmit and receive data and the like to and from the order quantity prediction system 1, the order receiving system 2, the display system 3, and / or the user terminal 4 by wireless communication and / or wired communication via the communication network NW.

[0112] The restaurant ordering system may be a standalone system or a cloud system.

[0113] (Order reception system) An example of functional blocks of the order receiving system will be described with reference to Fig. 9. The order receiving system 2 may include a processing unit 201, a payment unit 202, a user information identification unit 203, an order receiving unit 204, a display unit 205, a storage unit 206, a communication unit 207, and / or an input / output interface unit 208. The order receiving system 2 may include other functional blocks as necessary.

[0114] The processing unit 201 can perform various processes and calculations.

[0115] The payment unit 202 is not particularly limited as long as it is configured to enable payment. For example, it may be configured to enable cash payment, credit card payment, or two-dimensional code payment. It may also be configured to enable payment using an employee ID card, student ID card, membership card, or the like.

[0116] The user information identification unit 203 can identify the user information of the person who placed the order. For example, the user information may be identified by an identification number (e.g., employee number, student number, etc.) entered by the user or staff, or by information read by any linked reader (e.g., two-dimensional code reader, camera, etc.). For example, an employee ID card or student ID card may be read by the reader.

[0117] The order receiving unit 204 can receive the contents of an order placed by a user. The ordered menu items can be manually input by a staff member, or a code (e.g., a barcode or two-dimensional code) attached to a menu or the like can be read and input using a barcode reader or a camera. The code can also be attached to a food plate or other container. When inputting the order contents by reading a code, the user who is placing the order can have the code on the menu or the like read using a barcode reader or a camera without having a staff member present.

[0118] The display unit 205 may be, for example, a display, a monitor, a speaker, or the like.

[0119] The storage unit 206 can store various types of information and data.

[0120] The communication unit 207 can transmit and receive data and the like via wireless communication and / or wired communication via the communication network NW.

[0121] The input / output interface unit 208 is connected to an internal and / or externally connected input unit 209 and / or output unit 210, and can control the input unit 209 and / or output unit 210 (note that the input unit 209 and the output unit 210 are omitted in FIG. 9). Examples of the input unit 209 include a keyboard, a mouse, a microphone, etc. Examples of the output unit 210 include a display, a monitor, a speaker, etc.

[0122] Each functional block of the order receiving system 2 may be provided in one device, or one or more arbitrary functional blocks may be provided in two or more devices.

[0123] According to one embodiment of the present disclosure, the order receiving system is a restaurant cash register system (preferably a restaurant POS cash register system).

[0124] The order receiving system may be a stand-alone system or a cloud system.

[0125] (Display System) An example of functional blocks of the display system will be described with reference to Fig. 10. The display system 3 may include a processing unit 301, a display unit 302, a storage unit 303, a communication unit 304, and / or an input / output interface unit 305. The display system 3 may include other functional blocks as necessary.

[0126] The processing unit 301 can perform various processes and calculations.

[0127] The display unit 302 may be, for example, a display, a monitor, a speaker, or the like.

[0128] The storage unit 303 can store various types of information and data.

[0129] The communication unit 304 can transmit and receive data and the like via wireless communication and / or wired communication via the communication network NW.

[0130] The input / output interface unit 305 is connected to an internal and / or externally connected input unit 306 and / or output unit 307, and can control the input unit 306 and / or output unit 307 (note that the input unit 306 and the output unit 307 are omitted in FIG. 10). Examples of the input unit 306 include a keyboard, a mouse, a microphone, etc. Examples of the output unit 307 include a display, a monitor, a speaker, etc.

[0131] The order receiving system may be a stand-alone system or a cloud system.

[0132] (user terminal) The user terminal is not particularly limited and may be a smartphone, a personal computer, a tablet, etc.

[0133] [Method for predicting restaurant order volume] According to another embodiment of the present disclosure, there is provided a method for predicting order volume at a restaurant, the method comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for each user and / or an updated order prediction model for each attribute along a time series, the updated order prediction model for each user and / or the updated order prediction model for each attribute along a time series, by updating the order prediction model for each user and / or the order prediction model for each attribute according to a Bayesian statistical method based on an actual order history for each user on a specific day; calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user and / or the updated order prediction model for each attribute; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, A method is provided in which the external factor database stores any information other than the order history in association with a time series.

[0134] According to another embodiment of the present disclosure, there is provided a method for predicting order volume at a restaurant, the method comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history for each user on the specific day using a Bayesian statistical method; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, A method is provided in which the external factor database stores any information other than the order history in association with a time series.

[0135] According to another embodiment of the present disclosure, there is provided a method for predicting order volume at a restaurant, the method comprising: generating a time-series ordering pattern model for the entire restaurant based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, A method is provided in which the external factor database stores any information other than the order history in association with a time series.

[0136] According to another preferred embodiment of the present disclosure, the method further comprises the step of: calculating correction parameters to be used for correcting the ordering pattern model for each user, the ordering pattern model for each attribute, and / or the ordering pattern model for the entire restaurant, based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on the specific day. Further includes:

[0137] [Programs that predict order volume in restaurants, etc.] According to another embodiment of the present disclosure, there is provided a program for causing a computer device to execute the above method.

[0138] According to another embodiment of the present disclosure, there is provided a recording medium readable by a computer device, on which the above program is recorded.

[0139] According to another embodiment of the present disclosure, there is provided a computing device having the above program recorded in an internal storage unit.

[0140] Hereinafter, more specific aspects of the order quantity prediction system and cafeteria ordering system of the present disclosure will be described with reference to Fig. 11 etc. Fig. 11 shows an example of an employee cafeteria within a company.

[0141] The order quantity prediction system 1A is capable of mutually transmitting and receiving data, etc., with an order receiving system 2A, a display system 3A, and a user terminal 4A (not shown) used by user A via a network NW (not shown).

[0142] Past order history database 101A stores the past order history of each user at a restaurant, detailing the actions each user took at the restaurant. Past order history database 101A may obtain this order history, for example, from linked order reception system 2A, such as information about the time and minute at which each user placed an order and what each user ordered, and store this information in association with each user's attributes (e.g., gender, age, department, job title, etc.). In other words, past order history database 101A can store information in chronological order about the reservations and orders made by all users at the restaurant, the eating behaviors, and the times of their eating over a certain period of time (e.g., one week, one month, one year, or longer). Past order history database 101A stores, for example, the information shown in FIG. 12.

[0143] The external factor database 102A receives and stores external factor information such as weather information in chronological order from organizations that hold weather information, such as the Japan Meteorological Agency. The external factor information may be input at any frequency.

[0144] The order pattern model generation unit 108A performs multiple regression analysis based on the past order history for each user along a time series stored in the past order history database 101A and external factor information such as weather information along a time series stored in the external factor database 102A, to generate an order pattern model for each user along a time series, an order pattern model for each attribute along a time series (e.g., gender, age group, department, position), and / or an order pattern model for the entire restaurant (S601-S603). The multiple regression analysis may be performed using spreadsheet software or the like, or may be machine learning (e.g., deep learning). Examples of multiple regression analysis include box-and-whisker analysis and Pearson correlation analysis. The results of the multiple regression analysis may be stored in any database or storage unit provided in the order quantity reservation system 1A.

[0145] 13 and 14 show examples of the results of the multiple regression analysis performed by the order pattern model generation unit 108A. In FIG. 13, it can be seen that the daily miso soup sells well around 12:10, while udon and soba noodles sell well around 12:40. In addition, in FIG. 13, it can be seen that Chinese noodles sell well around 12:40 during the lunch hour and around 10:10 in the morning. In this way, through box-and-whisker analysis, it is possible to read the sales of specific menu items over time and the order volume for each menu item during normal business hours.

[0146] Figure 14 shows an example of analyzing the relationship between external factor information such as weather, temperature, and humidity and the order volume for each menu item using Pearson correlation analysis, obtained from external organizations such as the Japan Meteorological Agency. In Figure 14, the correlation between the order volume for each menu item and weather and other meteorological information is calculated. It can be evaluated that 0 indicates no correlation at all, a positive value close to 1 indicates a high probability of a positive correlation, and a negative value close to 1 indicates a high probability of a negative correlation. By analyzing various combinations like these, it is possible to evaluate the relationship between various external factors and the order volume for each menu item.

[0147] 15 and 16 show examples of order pattern models for each user generated by the order pattern model generation unit 108A. The example in FIG. 15 is an order pattern model for Mr. A, a manager. The probability distribution model calculated indicates that Mr. A tends to come to the restaurant around 11:45 on good days and order a set meal, but tends to come later (for example, around 1:15 p.m.) and order noodles such as ramen or soba on bad days. On the other hand, the example in FIG. 16 is an order pattern model for Ms. B, an office worker. The probability distribution model calculated indicates that Mr. B usually comes to the restaurant around 12:15 and tends to order pasta, but tends to order salad when he comes later (for example, around 12:45 p.m.). The generated order pattern model is sent to and stored in the order pattern model database 105A (S609).

[0148] The past order history database 101A also stores order history after the specific date X as needed. Furthermore, the external factor database 102A stores external factor information as needed. Therefore, the order pattern model generation unit 108A may retrieve data from these databases at any time, and may also retrieve order pattern models (for each user, for each attribute, and / or for the entire restaurant) stored in the order pattern model database 105A to regenerate an order pattern model, and this process can be repeated each time. This further improves the accuracy of the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant.

[0149] User A can use his / her own terminal (for example, a smartphone) or other IT tools such as a PC to reserve a menu for a specific date X in advance by viewing the menu (meal plan) posted on the website published by the restaurant operator on a daily basis. User A's reservation information is sent to and stored in order reservation database 103A (S505). An example of information stored in order reservation database 103A is shown in FIG. 17.

[0150] The order reservation information for each user on the specific day X stored in the order reservation database 103A, the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant stored in the order pattern model database 105A, and external factor information such as weather information related to the specific day X (in this case, the weather forecast for the specific day X) stored in the external factor database are transmitted to the order quantity prediction calculation unit 109A (S606-S608). Based on the transmitted information and the menu for the specific day X, the order quantity prediction calculation unit 109A calculates an order prediction model for each user on the specific day X by machine learning (for example, deep learning) (S609). This calculation may be a multiple regression analysis. The menu for the specific day X may be input each time and transmitted to the order quantity prediction calculation unit 109A of the order quantity prediction system 1A, or may be stored in any database or storage unit included in the order quantity prediction system 1A or in an external database not included in the order quantity prediction system 1A and transmitted to the order quantity prediction calculation unit 109A (not shown). The calculated order prediction model for the specific day X for each user, the order prediction model for the specific day X for each attribute, and / or the order prediction model for the entire restaurant on the specific day X are transmitted to and stored in the order prediction model database 106A. The order quantity prediction calculation unit 109A may be preset to generate the order prediction model at any timing (e.g., one day or one week before the specific day X). The order quantity prediction calculation unit 109A may calculate the order prediction model for each of a plurality of specific days (e.g., two days, one week, one month, three months, or more).

[0151] 18 and 19 show examples of order prediction models (order prediction models for a specific day X for each user) generated by the order quantity prediction calculation unit 109A. For example, in the case of Mr. A, the order quantity prediction calculation unit 109A predicts that there is a high probability that he will come to eat lunch around 1:30 p.m. on the specific day X because he is likely to be attending a general manager's meeting as a manager on the specific day X, and that there is a high probability that he will order something light to eat, such as soba noodles. The order prediction model (probability distribution model) shown in FIG. 18 is calculated as Mr. A's prior distribution for the specific day X. Meanwhile, in the case of Mr. B, the order quantity prediction calculation unit 109A predicts that there is a high probability that he will come around 12:30 p.m. as usual because he does not have any special meetings or the like scheduled on the specific day X. The order prediction model (probability distribution model) shown in FIG. 19 is calculated as Mr. B's prior distribution for the specific day X.

[0152] The order quantity prediction calculation unit 109A may calculate an order prediction model for the entire restaurant on the specific day X based on the calculated order prediction model for each user on the specific day X and / or the order prediction model for each attribute on the specific day X (S609). That is, the order quantity prediction calculation unit 109A can calculate an order prediction model for the entire restaurant on the specific day X based on the sum of the order prediction models for all users belonging to an organization on the specific day X and / or the sum of the order prediction models for each attribute of users belonging to the organization on the specific day X. Such calculations may be performed using spreadsheet software, machine learning, or the like. An example of the calculated order prediction model for the entire restaurant on the specific day X is shown in FIG. 20. The calculated order prediction model for the entire restaurant on the specific day X is sent to the order prediction model database 106A and stored therein (S610).

[0153] In addition, when there are two or more of the order prediction model (α) for the entire restaurant calculated based on the sum of the order prediction models for each user on the specific day X, the order prediction model (β) for the entire restaurant calculated based on the sum of the order prediction models for each attribute on the specific day X, and the order prediction model (γ) for the entire restaurant calculated based on the order pattern model for the entire restaurant, each may be weighted and then the order prediction model for the entire restaurant on the specific day X may be calculated. The calculation including such weighting may also be performed by the order quantity prediction calculation unit 109A (for example, by machine learning such as deep learning). The parameters used for weighting may be calculated by machine learning (for example, deep learning) based on the external factor information for the specific day X stored in the external factor database 102A, the order prediction models (α, β, γ) for the entire restaurant on the specific day X calculated by each calculation method, and the actual order quantity for the entire restaurant on the specific day X. In addition, the parameters used for weighting may be calculated by the correction parameter calculation unit 111A described later. Alternatively, the parameters used for the weighting may be set in advance based on the past knowledge and experience of restaurant operating company C, for example.

[0154] Because the information stored in the order reservation database 103A, the external factor database 102A, etc. is updated as needed until the specific day X, the order quantity prediction calculation unit 109A may calculate an order prediction model for each user on the specific day X, an order prediction model for each attribute on the specific day X, and / or an order prediction model for the entire restaurant on the specific day X at any timing (for example, at a specific time every day until the specific day X). Each order prediction model recalculated in this manner may be transmitted to and stored in the order prediction model database 106A. For example, one week before the specific day X, sunny weather is forecast for the specific day X, and the order prediction models for the specific day X are calculated by the order quantity prediction calculation unit 109A based on this information and stored in the order prediction model database 106A. However, two days before the specific day X, heavy rain is forecast for the specific day X, and the order prediction models for the specific day X are recalculated by the order quantity prediction calculation unit 109A based on this information and stored in the order prediction model database 106A (FIG. 21).

[0155] On the specific day X, the latest order prediction model for the entire restaurant on the specific day X stored in the order prediction model database 106A is transmitted to and displayed on the display system 3A (S611). Before opening time on the specific day X, a chef (e.g., head chef) of restaurant operating company C begins preparing meals based on the order prediction model displayed on the display system 3A.

[0156] On the specific date X, user A can actually order the reserved menu item using, for example, the order reception system 2A installed in the restaurant (S612). Meanwhile, user B, who has not made a reservation, can refer to the menu for that day on the specific date X (for example, by referring to a menu table (menu) presented on a display or bulletin board in the restaurant, or in some cases, an actual or mock-up menu), and select a menu item by operating the order reception system 2A (S613). Unlike user A, user B does not make a reservation or other special action before visiting the restaurant. For example, user B is assumed to be a user who casually visits the restaurant around lunchtime. After visiting the restaurant and selecting the menu item he or she wants to eat, user B must make payment. At the time of payment, user B can make the payment by holding his or her own terminal (for example, a smartphone) over a card reader attached to or built into the order reception system 2A. Payment can also be made without using the terminal, for example, by using an IC card (for example, an employee ID card). 11 shows the order receiving system 2A as a single system, the order receiving system 2A may be configured with multiple IT devices, such as a combination of a large display and a cash register system, or a combination of a display, a card reader, and a PC. Note that the order receiving system 2A in FIG. 11 may also be a restaurant cash register payment system that allows not only ordering menu items but also payment for those menu items.

[0157] This settled data, i.e., the actual order history for the specific day X, including the attributes of each of User A and User B, the time and minute, and the number of each menu item ordered, is transmitted from the order receiving system 2A to the order quantity reservation system 1A and stored in the current day order history database 104A (S614). This data may be transmitted from the order receiving system 2A to the past order history database 101A and stored therein (not shown). Alternatively, the data may be transmitted from the current day order history database 104A to the past order history database 101A at any time (for example, after business hours on the specific day X have ended) and stored therein (not shown). An example of information stored in the current day order history database 104A is shown in FIG. 22.

[0158] The actual order history (today's order history) may be transmitted to the current day order history database 104A and / or the past order history database 101A at any frequency (for example, at any time), which can be set appropriately in cooperation with the order quantity prediction system 1A and / or the order receiving system 2A. It is preferable that the actual order history be transmitted in close to real time.

[0159] The order history for that day stored in the current day order history database 104A is transmitted to the order quantity update calculation unit 110A at any timing (for example, every minute, every 5 minutes, every 10 minutes, or every 30 minutes) (S615). Furthermore, the latest order prediction models for each user, each attribute, and / or the entire restaurant on a specific day X stored in the order prediction model database 106A are transmitted to the order quantity update calculation unit 110A (S616). The transmission frequency of the order prediction models may be the same as the transmission frequency of the order history for that day. The order quantity update calculation unit 110A calculates an updated order prediction model for each user (i.e., a posterior distribution) based on the transmitted order prediction model (prior distribution) and the transmitted current day order history using Bayesian inference (S617). The order quantity update calculation unit 110A may calculate the updated order prediction model using Bayesian inference, as needed, while referring to the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant, which are stored in the order pattern model database 105A (for example, using these as likelihood functions). That is, the order quantity update calculation unit 110A can use the order prediction model as a prior distribution, perform Bayesian updating using the order history for the day and, as needed, the order pattern model, and calculate the updated order prediction model as a posterior distribution. Such calculations may be performed using machine learning (for example, deep learning).

[0160] The order quantity update calculation unit 110A can calculate an updated order prediction model for the entire restaurant on the specific day X based on the updated order prediction model for each user (S617). For example, the order quantity update calculation unit 110A can apply the updated order prediction model to a group of users who have already placed orders before the calculation is performed, and the order prediction model to a group of users who have not yet placed an order at the restaurant, and calculate an updated order prediction model for the entire restaurant on the specific day X based on the sum of these. Such calculations may be performed using spreadsheet software, machine learning, or the like.

[0161] Furthermore, the order quantity update calculation unit 110A can calculate an updated order prediction model for the entire restaurant on the specific day X based on the updated updated order prediction model for each attribute (S617). For example, the order quantity update calculation unit 110A uses the order prediction model for each attribute on the specific day X as a prior distribution, calculates an updated order prediction model for each attribute on the specific day X using Bayesian inference based on the current day order history transmitted from the current day order history database 104A, and then calculates an updated order prediction model for the entire restaurant based on the calculated updated order prediction model for each attribute. Such calculations may be performed using spreadsheet software, machine learning, etc.

[0162] Furthermore, the order quantity update calculation unit 110A may calculate an updated order prediction model for all dining halls on the specific day X based on the order prediction model for all dining halls on the specific day X (S617). For example, the order quantity update calculation unit 110A may use the order prediction model for all dining halls on the specific day X as a prior distribution and calculate an updated order prediction model for all dining halls on the specific day X by Bayesian inference based on the current day order history transmitted from the current day order history database 104A. Such calculation may be performed by machine learning or the like. An example of the calculated updated order prediction model for all dining halls on the specific day X is shown in FIG. 23.

[0163] The calculated updated order prediction models for each user, each attribute, and / or the entire restaurant on specific day X are sent to and stored in updated order prediction model database 107A (S618). The updated order prediction model for the entire restaurant calculated in this way is also sent to display system 3A and displayed (S619). A chef (e.g., head chef) of restaurant operating company C can adjust the amount of food to be cooked based on the updated order prediction model displayed on display system 3A.

[0164] On specific day X, the current day order history database 104A is updated each time a user visits the restaurant. Therefore, the order quantity update calculation unit 110A may use the latest updated order prediction model stored in the updated order prediction model database 107A as a prior distribution (S620) and, based on the constantly updated current day order history input in the current day order history database 110A (S615), recalculate updated order prediction models for each user, attribute, and / or the entire restaurant (S617). The updated order prediction models recalculated in this manner are then transmitted to the updated order pattern model database 107A for storage (S618) and transmitted to the display system 3A for display (S619). For example, if user D comes to the restaurant much earlier than usual (e.g., he usually comes to the restaurant around 12:30, but on specific day X he comes to the restaurant around 11:30) and orders "pork cutlet," which he does not usually order, based on this order history for that day, user D's order prediction model and / or the order prediction model (prior distribution) of user D's attributes are updated based on Bayesian inference, and an updated order prediction model for user D and / or an updated order prediction model (posterior distribution) of user D's attributes are calculated, and an updated order prediction model for the entire restaurant is also calculated.

[0165] A chef (e.g., head chef) of cafeteria operating company C can adjust the amount of ingredients to be prepared and the amount to be cooked moment by moment by referring to the updated order prediction model (i.e., the predicted order amount over time) that is updated moment by moment and displayed on display system 3A.

[0166] After the restaurant closes on the specific day X, the latest updated order prediction models for each user, each attribute, and / or the entire restaurant stored in the updated order prediction model database 107A may be transmitted to and stored in any database or storage unit (e.g., the order pattern model database 105A) included in the order quantity prediction system 1A (not shown). The updated order prediction models thus transmitted and stored may then be transmitted to the order quantity prediction calculation unit 109A, together with the order pattern models stored in the order pattern model database 105A, as needed (S608), to recalculate the order prediction models (S609). Alternatively, after the restaurant closes on the specific day X, the latest updated order prediction models for each user, each attribute, and / or the entire restaurant stored in the updated order prediction model database 107A may then be transmitted to the order quantity prediction calculation unit 109A, together with the order pattern models stored in the order pattern model database 105A, as needed (S608), to recalculate the order prediction models (S609).

[0167] Furthermore, after the restaurant's business hours on the specific day X, the latest updated order prediction models for each user, each attribute, and / or the entire restaurant stored in the updated order prediction model database 107A and information on the actual order volume for the entire restaurant on the specific day X (such information may be stored, for example, in the same-day order history database 104A) may be transmitted to the correction parameter calculation unit 111A (S621), and a correction parameter may be calculated by machine learning (for example, deep learning) based on the transmitted information (S622). This correction parameter represents, as a parameter, the gap between the updated order prediction models for each user, each attribute, and / or the entire restaurant on the specific day X and the actual order volume on the specific day X. The calculated correction parameter may be transmitted to and stored in any storage unit or database (for example, the order pattern model database 105A) provided in the order prediction system 1A (S623). The correction parameters stored in this manner are transmitted to the order pattern model generation unit 108A (not shown) together with information stored in the past order history database 101A (S601) and / or information stored in the external factor database 102A (S602), and may be used to generate order pattern models for each user, for each attribute, and / or for the entire restaurant (S603). The correction parameters stored in this manner are also transmitted to the order quantity prediction calculation unit 109A (not shown) together with information stored in the order reservation database 103A (S606), information stored in the external factor database 102A (S607), and / or each order pattern model stored in the order pattern model database (S608), and may be used to generate order prediction models for each user, for each attribute, and / or for the entire restaurant (S609). Generating or calculating the order pattern model and / or the order prediction model using the correction parameters calculated by the correction parameter calculation unit in this manner advantageously increases the accuracy of the order pattern model and / or the order prediction model.

[0168] The present disclosure encompasses the following: [1] An order quantity prediction system for a restaurant, comprising a past order history database, an external factor database, an order pattern model generation unit, an order quantity prediction calculation unit, and an order quantity update calculation unit, The past order history database stores the past order history of each user at the restaurant in association with the attribute information of the user and a time series of the order history, The external factor database stores any information other than the order history in association with a time series; the order pattern model generation unit generates, based on the information stored in the past order history database and the information stored in the external factor database, an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series; The order quantity prediction calculation unit (1) Calculating an order prediction model for each user along a time series for the specific day based on the order pattern model for each user, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each user. (2) calculating an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each attribute; and / or (3) calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; The order quantity update calculation unit (1) Calculating an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user using a Bayesian statistical method based on the actual order history of each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user. (2) calculating an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute based on the actual order history for each user on a specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute; and / or (3) updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant based on the time series on the specific day; Order quantity forecasting system. [2] The order quantity prediction system according to [1], wherein the external factor database stores at least one selected from the group consisting of weather information and event information in an organization. [3] The order quantity prediction system described in [1] or [2], wherein the order quantity prediction calculation unit further calculates an order prediction model for each user, an order pattern model for each attribute, and / or an order pattern model for the entire restaurant based on the order reservation information for each user on the specific day. [4] The system according to any one of [1] to [3], wherein the Bayesian statistical method is Bayesian inference. [5] The order quantity prediction system according to [3], further comprising an order reservation database that stores order reservation information for each user. [6] The order quantity prediction system according to any one of [1] to [5], further comprising an order pattern model database that stores the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant. [7] The order quantity prediction system according to any one of [1] to [6], further comprising an order prediction model database that stores the order prediction model for each user, the order prediction model for each attribute, and / or the order prediction model for the entire restaurant. [8] The order quantity prediction system according to any one of [1] to [7], further comprising an updated order prediction model database that stores the updated order prediction model for each user, the updated order prediction model for each attribute, and / or the updated order prediction model for the entire restaurant. [9] The order quantity prediction system further comprises a correction parameter calculation unit; the correction parameter calculation unit calculates correction parameters to be used for correcting the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant, based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on the specific day; An order quantity prediction system according to any one of [1] to [8].

[10] An order quantity prediction system according to any one of [1] to [9]; an order receiving system that transmits actual order histories for each user on a specific day to the order quantity prediction system; a display system for displaying the restaurant-wide order prediction model and / or the restaurant-wide updated order prediction model; A restaurant ordering system that includes:

[11] A method for predicting order volume at a restaurant, comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for each user and / or an updated order prediction model for each attribute along a time series, the updated order prediction model for each user and / or the updated order prediction model for each attribute along a time series, by updating the order prediction model for each user and / or the order prediction model for each attribute according to a Bayesian statistical method based on an actual order history for each user on a specific day; calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user and / or the updated order prediction model for each attribute; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, The method, wherein the external factor database stores any information other than the order history in association with a time series.

[12] A method for predicting order volume at a restaurant, comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history for each user on the specific day using a Bayesian statistical method; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, The method, wherein the external factor database stores any information other than the order history in association with a time series.

[13] A method for predicting order volume at a restaurant, comprising: generating a time-series ordering pattern model for the entire restaurant based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method; Including, The past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series of the user, The method, wherein the external factor database stores any information other than the order history in association with a time series.

[14] A step of calculating correction parameters to be used for correcting the ordering pattern model for each user, the ordering pattern model for each attribute, and / or the ordering pattern model for the entire restaurant based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on the specific day. The method according to any one of

[11] to

[13] , further comprising: [Explanation of symbols]

[0169] 1. 1A Order Quantity Forecasting System 101, 101A Past order history database 102, 102A External Factor Database 103, 103A Order Reservation Database 104, 104A Current day order history database 105, 105A Order Pattern Model Database 106, 106A Order Forecast Model Database 107, 107A Updated Order Forecast Model Database 108, 108A Order pattern model generation unit 109, 109A Order quantity prediction calculation unit 110, 110A Order quantity update calculation unit 111, 111A Correction parameter calculation section 112 Communications Department 113 Input / Output Interface Section 2. 2A Order Reception System 201 Processing section 202 Payment Department 203 User information identification unit 204 Order Reception Department 205 Display section 206 Memory section 207 Communications Department 208 Input / Output Interface Section 3. 3A Display System 301 Processing section 302 Display section 303 Storage section 304 Communications Department 305 Input / Output Interface Section 4. User terminal 11. Restaurant Ordering System NW communication network

Claims

1. An order quantity prediction system for a restaurant, comprising a past order history database, an external factor database, an order pattern model generation unit, an order quantity prediction calculation unit, and an order quantity update calculation unit, The past order history database stores the past order history of each user at the restaurant in association with the user's attribute information and a time series; The external factor database stores any information other than the order history in association with a time series; the order pattern model generation unit generates, based on the information stored in the past order history database and the information stored in the external factor database, an order pattern model for each user along a time series, an order pattern model for each attribute along a time series, and / or an order pattern model for the entire restaurant along a time series; The order quantity prediction calculation unit (1) Calculating an order prediction model for each user along a time series for the specific day based on the order pattern model for each user, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each user. (2) calculating an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database, and calculating an order prediction model for the entire restaurant along a time series for the specific day based on the sum of the order prediction models for each attribute; and / or (3) calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; The order quantity update calculation unit (1) Calculating an updated order prediction model for each user along a time series obtained by updating the order prediction model for each user using a Bayesian statistical method based on the actual order history of each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user. (2) calculating an updated order prediction model for each attribute along a time series obtained by updating the order prediction model for each attribute using a Bayesian statistical method based on the actual order history for each user on a specific day, and calculating an updated order prediction model for the entire restaurant along a time series for the specific day based on the updated order prediction model for each attribute; and / or (3) updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method, and calculating an updated order prediction model for the entire restaurant based on the time series on the specific day; Order quantity forecasting system.

2. 2. The order quantity forecasting system according to claim 1, wherein the external factor database stores at least one selected from the group consisting of weather information and event information in an organization.

3. 2. The order quantity prediction system according to claim 1, wherein the order quantity prediction calculation unit further calculates an order prediction model for each user, an order pattern model for each attribute, and / or an order pattern model for the entire restaurant based on the order reservation information for each user on the specific day.

4. The system of claim 1 , wherein the Bayesian statistical method is Bayesian inference.

5. The order quantity prediction system according to claim 3 , further comprising an order reservation database that stores order reservation information for each user.

6. The order quantity prediction system according to claim 1 , further comprising an order pattern model database that stores the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for all restaurants.

7. The order quantity prediction system according to claim 1 , further comprising an order prediction model database that stores the order prediction model for each user, the order prediction model for each attribute, and / or the order prediction model for all restaurants.

8. The order quantity prediction system according to claim 1 , further comprising an updated order prediction model database that stores the updated order prediction model for each user, the updated order prediction model for each attribute, and / or the updated order prediction model for all dining rooms.

9. the order quantity prediction system further comprises a correction parameter calculation unit, the correction parameter calculation unit calculates correction parameters to be used for correcting the order pattern model for each user, the order pattern model for each attribute, and / or the order pattern model for the entire restaurant, based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on the specific day; The order quantity forecasting system according to claim 1 .

10. An order quantity prediction system according to any one of claims 1 to 9; an order receiving system that transmits actual order history for each user on a specific day to the order quantity prediction system; a display system for displaying the restaurant-wide order prediction model and / or the restaurant-wide updated order prediction model; A restaurant ordering system that includes:

11. 1. A method for predicting order volume at a restaurant, comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for each user and / or an updated order prediction model for each attribute along a time series, the updated order prediction model for each user and / or the updated order prediction model for each attribute along a time series, by updating the order prediction model for each user and / or the order prediction model for each attribute based on an actual order history for each user on a specific day using a Bayesian statistical method; calculating an updated order prediction model for the entire restaurant along a time series on the specific day based on the updated order prediction model for each user and / or the updated order prediction model for each attribute; Including, the past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series; The method, wherein the external factor database stores any information other than the order history in association with a time series.

12. 1. A method for predicting order volume at a restaurant, comprising: generating an order pattern model for each user along a time series and / or an order pattern model for each attribute along a time series based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for each user along a time series for the specific day and / or an order prediction model for each attribute along a time series for the specific day based on the order pattern model for each user and / or the order pattern model for each attribute, the menu for the specific day, and information related to the specific day stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series on the specific day based on the sum of the order prediction model for each user and / or the order prediction model for each attribute; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history for each user on the specific day using a Bayesian statistical method; Including, the past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series; The method, wherein the external factor database stores any information other than the order history in association with a time series.

13. 1. A method for predicting order volume at a restaurant, comprising: generating a time-series ordering pattern model for the entire restaurant based on the information stored in the past order history database and the information stored in the external factor database; calculating an order prediction model for the entire restaurant along a time series for the specific day based on the order pattern model for the entire restaurant, the menu for the specific day, and information related to the specific day stored in the external factor database; a step of calculating an updated order prediction model for the entire restaurant along a time series for the specific day, the updated order prediction model being calculated by updating the order prediction model for the entire restaurant based on the actual order history of each user on the specific day using a Bayesian statistical method; Including, the past order history database stores the past order history of each user at the restaurant in association with attribute information of the user and a time series; The method, wherein the external factor database stores any information other than the order history in association with a time series.

14. calculating correction parameters to be used for correcting the ordering pattern model for each user, the ordering pattern model for each attribute, and / or the ordering pattern model for the entire restaurant based on the updated order prediction model for the entire restaurant and the actual order volume for the entire restaurant on the specific day; The method of any one of claims 11 to 13, further comprising:

Images