Planning device, planning method, and program

Abstract

To provide a plan creation device, a plan creation method and a program that improve accuracy for creating plans.SOLUTION: In an image optimization system, a plan creation device 1 comprises: a prediction section 122 which executes a predictive algorithm using a model that inputs a predictive data set including at least one type of information from period specifying information, period attribute information, and quantity information relating to a quantity of work targets in a period and outputs a man-hour predictive value, so that one or more man-hour predictive values corresponding to one or more future periods are obtained; and a creation section 124 which creates a work plan relating to work performed by one or more staffs during the one or more future periods based on the obtained one or more man-hour predictive values and one or more pieces of staff information corresponding to the one or more staffs performing the work.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a planning device, a planning method, and a program, and more particularly to a planning device, a planning method, and a program that create a work plan based on information such as man-hours required for work. [Background technology]

[0002] Patent document 1 describes a work plan creation device that creates a work plan based on process information that stores values ​​related to the required man-hours for a work process and candidate work locations, and location information that stores values ​​related to which work locations a product should pass through when moving it between work locations. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6261920 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the background art, since the required man-hours are not predicted, it is not easy to improve the accuracy of the plan creation.

[0005] An object of the present disclosure is to provide a planning device, a planning method, and a program that can improve the accuracy of planning. [Means for solving the problem]

[0006] A plan creation device according to one aspect of the present disclosure includes: A planning device that creates a work plan for a future period including a plurality of days from the next day onward, The prediction unit executes a prediction algorithm using a model to predict future period Corresponding to Work The estimated man-hours value is obtained. futureThe model is a model that uses a prediction dataset as input and outputs a predicted value of the man-hours. The prediction dataset is For each of the plurality of days, date information, day attribute information, and arrival amount information or visitor number information The creation unit includes Tako The predicted number and the above work People To members Regarding Based on personnel information, for the work object in the future period Create a work plan 。

[0007] A planning method according to one aspect of the present disclosure is executed by one or more processors. Create a work plan for a future period that includes multiple days beyond the next day The planning method includes a prediction step and a creation step. In the prediction step, a prediction algorithm using a model is executed to predict future period Corresponding in time Work The estimated man-hours value is obtained. future The model is a model that uses a prediction dataset as input and outputs a predicted value of the man-hours. The prediction dataset is For each of the plurality of days, date information, day attribute information, and arrival amount information or visitor number information In the creating step, Tako The predicted number and the above work People To members Regarding Based on personnel information, for the work object in the future period Create a work plan 。

[0008] A program according to one embodiment of the present disclosure includes: Create a work plan for a future period that includes multiple days A program for causing one or more processors to execute the planning method. The plan creation method includes a prediction step and a creation step. In the prediction step, a predicted man-hour value corresponding to a future period is obtained by executing a prediction algorithm using a model. The predicted man-hour value is the result of predicting the work man-hours for the future period. The model is a model that takes a prediction dataset as input and outputs a predicted man-hour value. The prediction dataset includes date information, day attribute information, and arrival amount information or visitor number information for each of the multiple days. In the creation step, a work plan for the work target for the future period is created based on the predicted man-hour value obtained in the prediction step and personnel information on the personnel who will perform the work. [Effects of the Invention]

[0009] The planning device, planning method, and program disclosed herein have the effect of improving the accuracy of planning. [Brief explanation of the drawings]

[0010] [Figure 1]FIG. 1 is a block diagram of a plan optimization system including a plan creation device according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a block diagram for explaining one-stage prediction by a generating unit and a predicting unit that configure the plan creating device. [Figure 3] FIG. 3 is a block diagram illustrating two-stage prediction by the generation unit and the prediction unit. [Figure 4] FIG. 4 is a block diagram illustrating three-stage prediction by the generation unit and the prediction unit of the embodiment. [Figure 5] FIG. 5 is a flowchart for explaining the operation of the plan creation device. [Figure 6] FIG. 6 is a flowchart for explaining the operation of the difference detection device constituting the plan optimization system. [Figure 7] FIG. 7 is a data structure diagram of work performance information used for 1 to 3 stage prediction by the generation unit and the prediction unit of the same. [Figure 8] FIG. 8 is a data structure diagram of datasets for one-stage prediction by the generation unit and the prediction unit (the first to fifth datasets for learning used by the generation unit, and the sixth and seventh datasets for prediction used by the prediction unit). [Figure 9] FIG. 9A is a data structure diagram showing the state in which the predicted man-hour values ​​have been added to the prediction dataset of the same data, and FIG. 9B is a data structure diagram showing the state in which the actual man-hour values ​​and the like have been added to make it for learning. [Figure 10] FIG. 10 is a data structure diagram of work schedule information constituting a work plan created based on the results of the above-mentioned one to three stage predictions. DETAILED DESCRIPTION OF THE INVENTION

[0011] The configurations described in the following embodiments are merely examples of the present disclosure. The present disclosure is not limited to the following embodiments, and various modifications are possible depending on the design, etc., as long as the effects of the present disclosure can be achieved.

[0012] (1) Planning optimization system As shown in FIG. 1, a plan optimization system 100 according to an embodiment of the present disclosure includes a plan creation device 1, a difference detection device 2, a camera 3, and an LPS (Local Positioning System) 4.

[0013] The plan creation device 1 is communicably connected to each of the difference detection device 2, camera 3, and LPS 4 via a network 200. The network 200 is, for example, a LAN (Local Area Network), the Internet, a telephone line network, etc. Each of the plan creation device 1, difference detection device 2, camera 3, and LPS 4 has a communication module (not shown) that enables communication via the network 200.

[0014] The plan creation device 1 creates a work plan (described later). The plan creation device 1 has a processor (CPU, MPU, GPU, etc.) and memory (semiconductor memory, disk, etc.). Various data and programs are stored in the memory, and the functions of the plan creation device 1 are realized by the processor executing the programs using the various data. In the following, the processor and memory that realize the various functions may be referred to as a "computer."

[0015] The difference detection device 2 observes, via the camera 3 and LPS 4, the work being performed based on the work plan created by the plan creation device 1, and detects differences between the work planned in the work plan and the work that was actually performed based on the observation results. The difference detection device 2 has a processor and memory. Various data and programs are stored in the memory, and the functions of the difference detection device 2 are realized when the processor uses the data to execute the programs.

[0016] The camera 3 is installed at the location where the work is to be performed, and captures images of the object of the work (hereinafter referred to as the work object), the personnel performing the work (store clerk, worker, etc.), and the like.

[0017] The LPS 4 is installed at the location where work is to be performed and detects the location of personnel (or items) at that location. The LPS 4 is composed of, for example, multiple beacons and multiple scanners (neither of which is shown). The beacons are carried by personnel (or attached to items) and emit signals containing personnel identification information that identifies the personnel (or item identification information that identifies the item). The scanners receive signals from the beacons and detect the location of the personnel (or item) based on the reception strength of the signals, the personnel identification information contained therein, and the location information of the scanner itself, etc.

[0018] The work is, for example, work related to sales. Sales is, for example, the sale of goods (food, daily necessities, etc.), but it may also be the sale of services (food and drink services, accommodation services, etc.). Work related to sales is shared among multiple departments, such as cashiers, stocking, and processing, but the work does not have to be shared.

[0019] Alternatively, the work may be, for example, a work related to logistics (see a modified example of the plan creation device). Work related to logistics is shared among multiple departments, such as collection and sorting, but the work does not have to be shared.

[0020] Alternatively, the work may be, for example, work related to the manufacture of an article, and the type of work is not important. Note that the work related to the manufacture of an article may be broken down into, for example, two or more steps, and the two or more steps may be performed sequentially. However, some of the two or more steps may be performed in parallel.

[0021] The place where the work is carried out is, for example, a store that serves as a sales base, but it may also be a warehouse that serves as a logistics base.

[0022] The work target may be, for example, a customer visiting the store or an item to be sold to the customer (item arriving at the store), but it may also be an item handled in logistics, or a manufactured item or part thereof.

[0023] A store may be, for example, a retail outlet that sells goods (such as a supermarket that sells food or a home improvement store that sells daily necessities), but it may also be a retail outlet that sells services (such as a restaurant that provides food and beverage services or a lodging facility that provides lodging services), or a factory that manufactures goods.

[0024] The personnel performing the work may be, for example, a store clerk, but may also be a warehouse worker or a factory production staff member. The personnel may belong to one of the above-mentioned multiple departments and perform work in that department. The personnel's affiliation may be variable, for example, the cashier department during one period and the stocking department during another period, but may also be fixed.

[0025] (2) Planning device (2-1) Overview 1 , the plan creation device 1 according to the embodiment of the present disclosure includes a reception unit 11, a processing unit 12, and an output unit 13. However, the plan creation device 1 may include only the processing unit 12, and the reception unit 11 and the output unit 13 may be included in a terminal device (for example, a tablet terminal, a smartphone, etc.) separate from the plan creation device 1. Furthermore, the reception unit 11 may be separate from the output unit 13.

[0026] The processing unit 12 includes a generating unit 121, a predicting unit 122, a setting unit 123, a creating unit 124, and an updating unit 125. However, the plan creation device 1 (processing unit 12) does not necessarily have to include at least one of the generating unit 121 and the setting unit 123.

[0027] The receiving unit 11 receives various types of information, such as standard man-hours (to be described later) and difference information (to be described later).

[0028] The reception by the reception unit 11 includes, for example, reception of information (e.g., standard labor hours) input via an input device such as a touch panel or a keyboard, and reception of information (e.g., difference information) handed over from other elements (e.g., difference detection device 2) that constitute the plan optimization system 100.

[0029] Furthermore, the reception in this embodiment may also include the reception of information transmitted from an external device, the reception of information read from a recording medium, and the like.

[0030] The processing unit 12 performs various processes. The various processes are, for example, the processes of the generation unit 121, the prediction unit 122, the setting unit 123, the creation unit 124, and the update unit 125. The processing unit 12 also performs various decisions that will be explained in the flowcharts. Note that other processes will be explained as appropriate.

[0031] The output unit 13 outputs various information, such as a work plan (work plan by time period and department) described later, a predicted man-hour value (a predicted man-hour value by time period and department, and a predicted total man-hour value) described later, and a predicted incoming quantity value described later.

[0032] The output of the output unit 13 is, for example, a display on a display, but may also be audio output from a speaker, printout by a printer, transmission to another device, recording on a recording medium, or the like.

[0033] The output unit 13 may convert information processed by a computer, such as a man-hour prediction value, into information that can be visually recognized by a person, and output it (for example, by displaying it on a display or printing it out on a printer).

[0034] (2-2) Man-hours forecast The prediction unit 122 constituting the processing unit 12 executes a prediction algorithm (PA1 to PA5: described later) using models (PM1 to PM5: described later) to obtain one or more predicted man-hour values ​​(PV2: described later) corresponding to one or more future periods.

[0035] A period in this embodiment is, for example, one day, or multiple time periods that make up one day (midnight, 1 a.m., 11 p.m., etc.). However, time periods are not limited to one-hour units, and may be, for example, 15-minute units, 30-minute units, or two-hour units. Furthermore, a period may be, for example, one week, one month, etc., and its length (period length) is not important. Furthermore, periods may be divided into, for example, morning / afternoon, early morning / daytime / nighttime, etc.

[0036] When the period is two or more, the two or more periods are usually continuous but may be discontinuous. Also, the two or more periods are usually the same length but may be different lengths.

[0037] The model in this embodiment is a model that receives a prediction dataset (DP: described later) as input and outputs a predicted man-hour value.

[0038] (2-2-1) Prediction Dataset The prediction dataset in this embodiment includes a set of period identification information (e.g., time slot identification information TI: described later), period attribute information (e.g., day attribute information AI: described later), and quantity information (e.g., predicted number of visitors PV1, predicted amount of arrival PV3: described later). However, the prediction dataset may include, for example, only either the period identification information or the period attribute information. The prediction dataset may include at least one type of information from among the period identification information, period attribute information, and quantity information. In general, the prediction dataset may include one piece of information associated with the information to be predicted.

[0039] Period identification information is information that identifies a period. The period may be a future period or a past period. Period identification information is, for example, a date or time slot identification information. Time slot identification information is a set of a start time and an end time, but it may also be a set of a start time and a duration. Furthermore, if the duration is a fixed value, the time slot identification information may be only the start time. Note that the time is usually expressed in year, month, day, and hour, minute, but the hour, minute, or minute may be omitted. For example, the start time and end time may each be "year, month, day" if the period is one day, or "year, month, day, hour" if the period is one hour.

[0040] Period attribute information is information relating to the attributes of a period. When the period is a day, the period attribute information is day attribute information. Day attribute information is, for example, the day of the week (Monday to Sunday), the day category (weekday / Saturday, Sunday, or holiday), a sale day flag, etc., but it can also be a holiday flag, a non-business day flag, or a New Year's holiday flag. The sale day flag is a flag indicating whether the day in question is a sale day or not. The holiday flag is a flag indicating whether the day in question is a holiday or not. The non-business day flag is a flag indicating whether the day in question is a non-business day or not. The New Year's holiday flag is a flag indicating whether the day in question falls within the New Year's holiday period.

[0041] When the period is a time period, the period attribute information is, for example, time period attribute information. However, even if the period is a time period, the period may include the above-mentioned day attribute information. The time period attribute information is, for example, a sale time period flag, a shortened hours flag, etc. The sale time period flag is a flag indicating whether the time period is a sale time period or not. The shortened hours flag is a flag indicating whether the time period is a time period during which the store is closed as a result of shortened business hours.

[0042] The quantity information is information about the quantity of work to be done in a period. The period may be attribute-specific information that specifies the attributes of a period that lags with respect to the period in question, such as the previous day or the next day.

[0043] As described above, the work target in this embodiment is, for example, a customer (a sales partner) or an item (a sales target, a distribution target). However, the work target may also be information (a target for administrative processing in a company, a target for application processing in a government office, etc.), a facility (a target for cleaning, a target for equipment replenishment, etc.), or a vehicle (a target for vehicle inspection, a target for car washing, a target for refueling, etc.). In the case of sales, quantity information is, for example, visitor count information regarding the number of customers visiting the store during a period, or information regarding the amount of goods arriving at the store during a period (incoming quantity information). In the case of logistics, quantity information is, for example, information regarding the amount of goods handled during a period (handling quantity information). In the case of manufacturing, quantity information may be, for example, information regarding the amount of goods shipped during a period (shipping quantity information).

[0044] However, the period attribute information may not only be the above-mentioned categories and flags, but may also be information relating to the number of days, such as how many days before or after a particular event (for example, Christmas).

[0045] In the case of sales, the quantity information is, for example, information on the number of visitors. The information on the number of visitors is usually a predicted number of visitors, but it may also be an actual number of visitors (the actual number of visitors: an automatically calculated or manually entered number). The predicted number of visitors is the value of the predicted number of visitors, and the actual number of visitors is the value of the actual number of visitors.

[0046] The predicted number of visitors is, for example, a value of the number of visitors predicted by a person or a machine, and is obtained, for example, from an external device (such as a store server). However, the predicted number of visitors may also be a value obtained using a customer number prediction model (see other variations). For example, the prediction unit 122 may predict the number of visitors using the customer number prediction model and obtain the predicted number of visitors.

[0047] The actual number of visitors is used as a substitute for the predicted number of visitors when the predicted number of visitors is not obtained. The actual number of visitors is a value automatically tallied, for example, using a sensor (not shown) installed at the entrance / exit of the store, or a camera 3 and PLS 4. However, the actual number of visitors may also be a value entered by a person (manually entered). The actual number of visitors may also be obtained, for example, from an external device.

[0048] In particular, when the sale is of goods (product sales), the quantity information may be information on the amount of stock received. The amount of stock received is usually a predicted amount of stock received, but it may also be an actual amount of stock received (AV2: described later). The predicted amount of stock received is the value of the predicted amount of stock received, and the actual amount of stock received is the value of the actual amount of stock received.

[0049] The actual amount of goods received is generally determined at the time of arrival, but may also be determined at the time of ordering. The actual amount of goods received may include such a determined amount of goods received (amount scheduled to arrive) that has not yet arrived.

[0050] In the case of physical distribution, the quantity information is, for example, handling volume information. The handling volume information is usually a past value of the handling volume, but may also be a predicted value of the handling volume.

[0051] The predicted man-hours are the results of predicting the man-hours for work over the period described above. Work man-hours are the man-hours required to complete a task. Man-hours are a numerical value that indicates the amount of work, and are expressed as the product of time and the number of people. The unit of man-hours is, for example, "man-hours," but time units (days, hours, minutes, etc.) can also be used as is.

[0052] In this embodiment, the amount of work (work volume) is expressed in man-hours. That is, the work volumes of the various tasks described above are all converted into man-hours. Furthermore, the total man-hours is obtained by adding up the man-hours corresponding to the multiple tasks. This makes it possible to compare the work volumes of different types of tasks and calculate the total work volume. The conversion from work volume to man-hours may be performed manually or using a predetermined algorithm. The predetermined algorithm may be, for example, an algorithm that uses standard man-hours. An algorithm that uses standard man-hours converts one or more work volumes into one or more man-hours using a set standard man-hour.

[0053] (2-2-2) Prediction Algorithm The prediction algorithm corresponds to the learning algorithm of machine learning executed by the generation unit 121, which will be described later, and is an algorithm for obtaining a predicted man-hour value using a model generated by the generation unit 121 through machine learning.

[0054] However, the prediction algorithm may be an algorithm that acquires a predicted man-hour value using a model generated by a method other than machine learning (for example, a linear prediction model based on statistical data).

[0055] Note that the machine learning suitable for generating a man-hour prediction model is supervised learning, but unsupervised learning, reinforcement learning, etc. are also acceptable. For supervised learning, regression (continuous value prediction) type is suitable, but classification (discrete value prediction) type is also acceptable.

[0056] The machine learning in the embodiment is LightGBM (Gradient Boosting Machine), which is an algorithm that combines a decision tree and gradient boosting, and has both regression and classification functions.

[0057] However, the machine learning may be a decision tree or other algorithms (for example, neural networks, clustering, support vector machines, deep learning, random forests, etc.) Furthermore, the machine learning may be an ensemble method that combines multiple of these algorithms.

[0058] (2-3) Setting standard man-hours The setting unit 123 sets the standard man-hours for the work. The set standard man-hours are stored in, for example, the memory of the plan creation device 1, but the storage location is not important.

[0059] Standard man-hours are the standard man-hours for a task. For example, standard man-hours are the man-hours required to complete a task (net work time) plus slack time such as travel time to the work site, preparation time before work, and rest time between tasks. However, adding slack time is not required, and standard man-hours may be the man-hours required to complete a task.

[0060] The standard man-hours are set, for example, manually. More specifically, a person inputs a numerical value indicating the standard man-hours via an input device such as a touch panel. In the plan creation device 1, the receiving unit 11 receives the input numerical value, and the setting unit 123 sets the received numerical value as the standard man-hours.

[0061] However, the standard man-hours may be set automatically. More specifically, the setting unit 123 sets the standard man-hours by executing an algorithm that takes the above-mentioned quantity information, personnel information, and the like as input and outputs the standard man-hours.

[0062] Personnel information is information about personnel. Personnel information includes personnel identification information. Personnel identification information is information that identifies personnel. Personnel identification information is, for example, name and address, mobile phone number, email address, etc., but may also be an ID associated with name and address, etc. Note that personnel information is obtained, for example, from an external device (such as a store server), but some of it (such as vacation information, which will be described later) may be manually entered by personnel.

[0063] Furthermore, personnel information typically also includes work performance information and vacation information. Work performance information is information related to the work performance of personnel. Work performance information is, for example, a collection of period-specific information or a collection of pairs of period-specific information and department-specific information.

[0064] Vacation information is information about the employee's planned vacation. Vacation information is, for example, a collection of period-specific information. The period-specific information that makes up vacation information is usually a date, but it can also be morning / afternoon, time period, etc.

[0065] Specifically, vacation information may be, for example, a set of planned vacation dates. Alternatively, vacation information may be calendar information in which planned vacation dates are marked with a vacation flag indicating the planned vacation date (i.e., part of work schedule information, which will be described later).

[0066] Skill information is information about the work skills of a person. For example, skill information is a collection of department-specific information that identifies the departments that the person can be in charge of. Skill information may also include information about, for example, years of work experience, position, qualifications, etc.

[0067] The personnel information may also include work schedule information. The work schedule information is information about the work schedule of the personnel. The work schedule information is, for example, a set of period-specific information and department-specific information (see FIG. 10). However, in the case of a single department, the work schedule information may also be a set of period-specific information (not shown).

[0068] Specifically, work schedule information is, for example, a set of pairs of the scheduled work date, start time, and department name. However, in the case of regular work hours, the start time may not be necessary. Also, in the case of a single department, the department name may not be necessary. Alternatively, work schedule information may be a set of start time and department name, or a start time and a work flag indicating the work day, attached to the scheduled work date among the multiple dates that make up the calendar information.

[0069] The work schedule information is a portion of a work plan, which will be described later, that corresponds to one piece of personnel identification information. In other words, the work schedule information is acquired from the work plan after the creation unit 124, which will be described later, creates the work plan. However, the initial value of the work schedule information is information based on manual input by the personnel, and the initial value may be updated later in the work plan created by the creation unit 124.

[0070] The setting unit 123 in this embodiment also uses department information to set standard man-hours for each department. Department information is information about multiple departments that share work. Department information is, for example, a set of pairs of department identification information and priority information. Department identification information is information that identifies a department. Department identification information is, for example, a department name such as "cash register" or "stocking," but it may also be an ID associated with the department name. Priority information is information regarding the priority between departments. Priority information is, for example, a number indicating the priority (such as "1," "2," etc.).

[0071] A set of pairs of department-specific information and priority information is, for example, "(cash register, 1), (stocking, 2), ...". Alternatively, the department information may be an arrangement of multiple department-specific information in order of priority (for example, "cash register, stocking, ...").

[0072] However, the priority information is not essential, and the department information may be, for example, a collection of multiple pieces of department-specific information corresponding to multiple departments.

[0073] Furthermore, multiple department identification information may be hierarchical (store, department within store, team within department, etc.). For example, the top department may be a "store," the lower department may be a "department" within the store, and the lower department may be a "team" within the department.

[0074] The setting unit 123 may also use sales information related to sales obtained through the work to set the standard man-hours taking costs into consideration, for example.

[0075] In the case of manufacturing an article, the standard man-hours may be set for each process.

[0076] (2-4) Planning The creation unit 124 creates a work plan based on one or more man-hour prediction values ​​acquired by the prediction unit 122 and one or more pieces of personnel information corresponding to one or more personnel who will perform the work. The number of personnel who will perform the work is usually two or more, but may be one. Therefore, the number of pieces of personnel information used to create a work plan is usually two or more, but may be one. In the following description, the number of personnel is two or more, and the number of pieces of personnel information is two or more. The created work plan is stored, for example, in the memory of the plan creation device 1, but the storage location is not important.

[0077] A work plan is a plan for work to be performed by two or more personnel in one or more future time periods. In this embodiment, the work plan is, for example, a set of two or more pieces of scheduled work information corresponding to two or more pieces of personnel identification information, as shown in Fig. 10. As described above, the scheduled work information is a set of pairs of period identification information and department identification information.

[0078] Details will be explained using a specific example, but for example, the work schedule information corresponding to personnel identification information "aa" is "{(2021,2,21,8),AA},{(2021,2,21,9),AA},···{(2021,2,21,14),AA},{(2021,2,22,9),BB}···".

[0079] The work plan may be composed of, for example, a shift plan and a personnel plan. A shift plan is a plan regarding the work schedules (shifts) of multiple personnel. A shift plan is information indicating which personnel will work during which period (for example, which day), and is composed of, for example, a set of personnel identification information and period identification information (for example, day identification information). A personnel plan is a plan regarding the allocation of personnel scheduled to work to each department. A personnel plan is, for example, a set of personnel identification information, department identification information, and period identification information (for example, time period identification information).

[0080] However, the above is merely an example, and the data structure of the work plan is not important.

[0081] The creation unit 124 in this embodiment creates a work plan for each time period and each department by using department information as well. Note that the work plan may be created taking into consideration constraints other than personnel (skills) and departments (priority levels).

[0082] The creation unit 124 in this embodiment creates a work plan using a creation algorithm that takes as input a creation dataset including, for example, one or more man-hour prediction values ​​and two or more pieces of personnel information, and outputs a work plan.

[0083] However, the creation dataset may include one or more workloads and standard workloads set by the setting unit 123, instead of one or more predicted workload values. In this case, the creation unit 124 converts one or more workloads into one or more workloads based on the standard workloads, for example, by using the above-mentioned "algorithm using standard workloads." Then, the creation unit 124 may create a work plan based on one or more workloads and two or more pieces of personnel information using the creation algorithm.

[0084] (2-5) Update of standard man-hours The update unit 125 updates the standard man-hours set by the setting unit 123 based on the difference information from the difference detection device 2 .

[0085] In more detail, for example, the standard man-hours set by the setting unit 123 are stored in the memory of the plan creation device 1. The receiving unit 11 receives the difference information output from the difference detection device 2, and the updating unit 125 updates the standard man-hours in the memory so as to reduce the difference indicated by the received difference information.

[0086] Specifically, for example, difference information for the current period is stored in the memory of the difference detection device 2. When transitioning from the current period to the next period, the update unit 125 increments or decrements the current standard man-hours in the memory. Then, when transitioning from the next period to the period after that, the update unit 125 acquires difference information from the difference detection device 2 and compares the acquired difference information with the difference information in the memory. If the comparison shows that the difference has decreased, the update unit 125 retains the standard man-hours after the increment or decrement. On the other hand, if the difference has increased, the update unit 125 restores the standard man-hours after the increment or decrement to the original standard man-hours.

[0087] By repeating the above process, the standard man-hours are updated so that the difference between the planned man-hours and the actual man-hours is reduced.

[0088] (2-6) Model generation The generation unit 121 generates a model by executing a machine learning learning algorithm (LA: described later) using two or more learning datasets (DT1 to DT4: described later). The generated model is stored in, for example, the memory of the plan creation device 1, but the storage location is not important.

[0089] In detail, for each of two or more training datasets, the generation unit 121 executes a machine learning learning algorithm using a part (first part) of the training dataset as input data and another part (second part) of the training dataset as training data.

[0090] The first part is a part common to the prediction dataset, and is, for example, a set of period identification information, period attribute information, and quantity information. The second part is a part corresponding to the output of the prediction model, and is, for example, the actual man-hour value. As described above, the first part, which is a part common to the prediction dataset, may include at least one type of information among period identification information, period attribute information, and quantity information. The first part may include one piece of information associated with the information to be predicted.

[0091] This allows the relationship between the first and second parts to be learned, and a prediction model is generated that outputs a predicted value of the second part for the unknown first part.

[0092] In addition, executing a machine learning learning algorithm using two or more training datasets may involve, for example, generating a predictive model by inputting the input values ​​(teacher data) of the training datasets into the input layer of the predictive model and inputting the output values ​​of the training datasets into the output layer of the predictive model.

[0093] The training dataset includes, for example, a set of period specifying information, period attribute information, and quantity information, and an actual man-hour value (AV1: described below). However, the training dataset only needs to include, for example, at least one type of information from among period specifying information, period attribute information, and quantity information, and an actual man-hour value. In general, the training dataset only needs to include information on the prediction target and one piece of information associated with the information on the prediction target.

[0094] The actual man-hour value is the actual man-hours of work performed during a period. The actual man-hour value is obtained, for example, from two or more pieces of work performance information (see FIG. 7, described later) corresponding to two or more personnel. The actual man-hour value corresponds to the second part described above.

[0095] In this embodiment, the work is work related to sales, which may be sales of goods or sales of services.

[0096] The target here is visitors to the place where the sale is taking place. The quantity information includes information on the number of visitors. The information on the number of visitors is information on the number of visitors during a period.

[0097] The visitor number information includes a visitor number prediction value (PV1: described later). The visitor number prediction value is the result of predicting the number of visitors for a period.

[0098] The period attribute information includes a sale period flag. The sale period flag is a flag that indicates whether or not a period belongs to a sale period. The sale period is, for example, a sale day, a sale time period, etc. The sale period flag is, for example, a sale day flag, a sale time period flag, etc.

[0099] The period attribute information further includes weather information. The weather information is information about the weather in the location where the work is performed during the period. The weather information is, for example, an actual measured value, but may also be a predicted value. The weather information may be obtained, for example, from a weather company's server.

[0100] The period may be, for example, each of one or more time slots belonging to one day. In this case, the setting unit 123 sets, for example, a unified standard man-hour for the work, but may also set standard man-hours for each time slot. The predicted number of visitors is a predicted value of the number of visitors on a day that includes one or more time slots. The predicted man-hour value includes a predicted man-hour value for each time slot.

[0101] Work may be shared among two or more departments. The two or more departments may be organized hierarchically. Two or more hierarchical departments may be, for example, multiple stores belonging to a single company, multiple departments belonging to a single store, or multiple teams belonging to a single department. A department may be, for example, various sales areas such as a food section or a clothing section, a kitchen, or a back room. A team may be, for example, the fresh fish section or prepared food section in a food section.

[0102] In this case, each of the two or more departments has one or more personnel who perform the work. The work of each department is usually performed in parallel, but may be performed sequentially. The setting unit 123 sets standard man-hours for each time period and department. The predicted man-hour values ​​include predicted man-hour values ​​for each time period and department.

[0103] The processing unit 12 may determine whether an input data set is for prediction or learning, and transfer the data set to an algorithm corresponding to the determination result. This allows a group of data sets including two types of data sets, one for learning and one for prediction, to be used to generate a model and perform prediction using the model.

[0104] (2-7) Variations of the model origin In this modification, the model is generated by an external device and is transferred from the external device to the planning device 1. Alternatively, the model may be stored in advance in the memory of the planning device 1, and its origin does not matter.

[0105] (2-8) Model details: One-step prediction As shown in FIG. 2, the model in this embodiment is a first man-hour prediction model PM1 that receives a prediction data set DP as input and outputs man-hour prediction values ​​PV2 by time period and department.

[0106] The prediction dataset DP includes, for example, a set of time slot identification information TI, day attribute information AI, and predicted number of visitors PV1 for each of one or more future time slots. However, the prediction dataset DP only needs to include at least one type of information from among time slot identification information TI, day attribute information AI, and predicted number of visitors PV1. The time slot identification information TI is information that identifies the time slot. The day attribute information AI is information about the attributes of a day to which the time slot belongs. The predicted number of visitors PV1 is a predicted value for the number of visitors on a day to which the time slot belongs.

[0107] The prediction data set DP is usually associated with department-specific information, which is information that identifies a department.

[0108] However, there may be prediction datasets that are not associated with department identification information. Furthermore, the predicted number of visitors constituting the prediction dataset DP may be, for example, a predicted value of the number of visitors to a department identified by the department identification information associated with the prediction dataset DP, or a predicted value of the number of visitors to a department higher than the department identified by the department identification information.

[0109] Specifically, for example, when the department specified by the department identification information is the "clothing department," the predicted number of visitors may be, for example, the predicted number of visitors to the "clothing department," or the predicted number of visitors to a "sales floor" higher than the "clothing department." Furthermore, when the department specified by the department identification information is the "fresh fish section," the predicted number of visitors may be, for example, the predicted number of visitors to the "fresh fish section," or the predicted number of visitors to a "food section" higher than the "fresh fish section," or the predicted number of visitors to an even higher-level "sales floor" or the "store" as a whole. These matters also apply to the predicted number of visitors that constitute the training dataset DT.

[0110] The prediction unit 122 executes a first labor-hour prediction algorithm PA1 using a first labor-hour prediction model PM1 with a prediction data set DP corresponding to each of one or more department-specific information as input, thereby obtaining one or more labor-hour prediction values ​​for each department corresponding to one or more time periods in the future.

[0111] In this manner, in this embodiment, one-stage prediction is performed to predict man-hours by time period and department directly from the prediction dataset DP.

[0112] According to this embodiment, the accuracy of man-hour prediction by time period and department can be further improved by one-stage prediction using the first man-hour prediction model PM1.

[0113] In this embodiment, the training dataset is a first training dataset DT1. The first training dataset DT1 includes, for each of one or more past time periods, a set of time period identification information TI, day attribute information AI, and a predicted number of visitors PV1, as well as a time period and department-specific actual man-hours value AV1. However, it is sufficient that the training dataset DT1 includes at least one type of information from among the time period identification information TI, day attribute information AI, and the predicted number of visitors PV1, as well as the time period and department-specific actual man-hours value AV1.

[0114] The man-hour performance value AV1 by time period and department is the actual man-hours by department for the work in the time period. The first learning dataset DT1 is associated with department identification information.

[0115] The generation unit 121 generates a first labor-hour prediction model PM1 for each of two or more pieces of department-specific information by executing a machine learning learning algorithm LA using a first part of a first learning dataset DT1 corresponding to the department-specific information as input data and a second part as training data.

[0116] In this way, in this embodiment, the plan creation device 1 itself can generate the first man-hour prediction model PM1 by machine learning using the first learning dataset (input data and teacher data: the same applies hereinafter).

[0117] (2-9) Model Variation: Multi-Step Prediction The model may be a model that performs multi-stage prediction. In the following modified examples 1 and 2 of the model, models that perform two-stage prediction and three-stage prediction will be described.

[0118] In a modified example, the sales are sales of goods. The quantity information further includes a predicted arrival amount PV3 in addition to the predicted number of visitors PV1. The predicted arrival amount PV3 is the result of predicting the arrival amount of goods at the location where the work is performed during the period.

[0119] In the modified example, the predicted arrival quantity value PV3 is a predicted arrival quantity value for one day that includes one or more time slots. However, the predicted arrival quantity value PV3 may also be a predicted value for one or more time slots. Furthermore, the quantity information constituting the learning dataset DT1 may include an actual arrival quantity value AV2 instead of the predicted arrival quantity value PV3.

[0120] (2-9-1) Model Variation 1: Two-stage Prediction In this first modification, as shown in FIG. 3, the model includes an incoming quantity forecasting model PM4 and a second man-hour forecasting model PM2.

[0121] The incoming quantity prediction model PM4 is a model that takes the prediction dataset DP as input and outputs the incoming quantity predicted value PV3. The prediction dataset DP in this modification 1 includes, for each of one or more future time periods, a set of, for example, time period identification information TI, day attribute information AI, and predicted number of visitors PV1. However, the predicted number of visitors PV1 does not have to be included in the prediction dataset DP. The prediction dataset DP only needs to include, for example, at least one type of information from time period identification information TI, day attribute information AI, and predicted number of visitors PV1. The time period identification information TI is information that identifies the time period. The day attribute information AI is information regarding the attributes of a day to which the time period belongs.

[0122] The second man-hour prediction model PM2 is a model that takes as input the prediction dataset DP and the incoming quantity forecast value PV3, which is the output of the incoming quantity prediction model PM4, and outputs the man-hour forecast value PV2 by time period and department.

[0123] The time zone and department-specific man-hours predicted value PV2 is a result of predicting the work man-hours for each of two or more future time zones by department.

[0124] The prediction data set DP is associated with department identification information, which is information that identifies a department.

[0125] The prediction unit 122 acquires a predicted incoming quantity value PV3 for each of two or more pieces of department identification information by executing an incoming quantity prediction algorithm PA4 using an incoming quantity prediction model PM4 with the prediction data set DP associated with the department identification information as input.The prediction unit 122 then acquires one or more predicted man-hour values ​​for each department corresponding to one or more future time periods by executing a second man-hour prediction algorithm PA2 using a second man-hour prediction model PM2 with the acquired incoming quantity prediction value PV3 and the prediction data set DP associated with the department identification information as input.

[0126] In this way, in the first modification regarding the model, a two-stage prediction is performed in which the incoming quantity is predicted from the prediction data set DP, and the incoming quantity prediction value PV3 is used to predict the man-hours by time period and department.

[0127] According to the first modification of the model, the accuracy of man-hour prediction by time period and department can be improved by two-stage prediction using the incoming quantity prediction model PM4 and the second man-hour prediction model PM2. In addition, the incoming quantity prediction value PV3 obtained in the man-hour prediction process can be used to optimize the work plan.

[0128] In addition, in Modification 1 related to the model, the training data sets include a second training data set DT2 and a third training data set DT3.

[0129] The second learning dataset DT2 includes, for each of one or more past time periods, for example, a set of time period identification information TI, day attribute information AI, and predicted number of customers PV1, and an actual amount of goods received AV2. However, the second learning dataset DT2 only needs to include one or more types of information from time period identification information TI, day attribute information AI, and predicted number of customers PV1, and the actual amount of goods received AV2. The actual amount of goods received AV2 is the actual amount of goods received on a day to which the one or more time periods belong.

[0130] The third training data set DT3 includes the second training data set DT2 and a time-slot and department-specific actual man-hour value AV1 for each of one or more past time slots. The time-slot and department-specific actual man-hour value AV1 is the actual man-hours for work by department in that time slot.

[0131] Each of the second training data set DT2 and the third training data set DT3 is associated with the department specifying information.

[0132] The generation unit 121 generates an incoming quantity prediction model PM4 for each of two or more department identification information by executing a machine learning learning algorithm LA1 using the first part of the second learning dataset DT2 corresponding to the department identification information as input data and the second part as training data.

[0133] In addition, for each of two or more pieces of department identification information, the generation unit 121 generates a second labor-hour prediction model PM2 by executing a machine learning learning algorithm LA2 using the first part of the third learning dataset DT3 corresponding to the department identification information and the arrival quantity prediction value that is the output of the arrival quantity prediction model PM4 as input data and the second part as training data.

[0134] According to variant example 1 regarding the model, the planning device 1 itself can generate the incoming quantity prediction model PM4 by machine learning on the second training data set DT2, and can generate the second labor-hour prediction model PM2 by machine learning on the third training data set DT3.

[0135] (2-9-2) Model Variation 2: Three-stage Prediction In this second modification, as shown in FIG. 4, the model includes an incoming quantity forecasting model PM4, a third man-hour forecasting model PM3, and a decomposition model PM5.

[0136] The incoming quantity prediction model PM4 is a model that takes the prediction dataset DP as input and outputs the incoming quantity prediction value PV3. The prediction dataset DP includes, for each of one or more future time periods, a set of, for example, time period identification information TI, day attribute information AI, and predicted number of visitors PV1. However, the prediction dataset DP only needs to include, for example, at least one type of information from time period identification information TI, day attribute information AI, and predicted number of visitors PV1. The time period identification information TI is information that identifies a time period. The day attribute information AI is information regarding the attributes of a day to which the time period belongs. The predicted number of visitors PV1 is a predicted value for the number of visitors on a day to which the time period belongs.

[0137] The third effort forecasting model PM3 is a model that takes the forecasting dataset DP and the arrival volume forecast value PV3 as inputs and outputs the total effort forecast value PV4. The arrival volume forecast value PV3 here is the output of the arrival volume forecasting model PM4.

[0138] The total man-hour prediction value PV4 is the result of predicting the total man-hours for work on a day that includes one or more future time slots. The total man-hours is the sum of the man-hours by two or more time slots and departments that correspond to a combination of one or more time slots and two or more departments.

[0139] The decomposition model PM5 is a model that takes the total man-hour prediction value PV4, time period specification information TI, and day attribute information AI as inputs, and outputs the time period and departmental man-hour prediction value PV2. The total man-hour prediction value PV4 is the output of the third man-hour prediction model PM3. The time period and departmental man-hour prediction value PV2 is the result of predicting the work man-hours by department for each of two or more future time periods.

[0140] The prediction data set DP is associated with department identification information, which is information that identifies a department.

[0141] The prediction unit 122 obtains an incoming quantity prediction value PV3 for each of two or more department-specific information by executing an incoming quantity prediction algorithm PA4 using an incoming quantity prediction model PM4, using as input a prediction dataset DP corresponding to the department-specific information.

[0142] In addition, the prediction unit 122 acquires a total labor-hour prediction value PV4 by executing a third labor-hour prediction algorithm PA3 using a third labor-hour prediction model PM3 as input, using the acquired incoming quantity prediction value PV3 and a prediction dataset DP corresponding to the department-specific information.

[0143] Then, the prediction unit 122 uses the acquired total man-hour prediction value PV4 as input and executes the decomposition algorithm PA5 using the decomposition model PM5 to acquire two or more man-hour prediction values ​​for each department corresponding to one or more time periods in the future.

[0144] In this way, in the second variant of the model, a three-stage forecast is performed in which the incoming volume is predicted from the forecast dataset, the incoming volume forecast value PV3 is used to forecast the total man-hours, and the total man-hours forecast value PV4 is broken down into man-hours by time period and department.

[0145] According to the second modification regarding the model, the accuracy of the man-hour prediction by time period and department can be improved by three-stage prediction using the incoming quantity prediction model PM4, the third man-hour prediction model PM3, and the decomposition model PM5. In addition, the incoming quantity prediction value PV3 and the total man-hour prediction value PV4 obtained in the man-hour prediction process can be used to optimize the work plan.

[0146] Furthermore, in the second modification example related to the model, the training data sets include a second training data set DT2, a third training data set DT3, and a fourth training data set DT4.

[0147] The second learning dataset DT2 includes, for each of one or more past time periods, a set of time period identification information TI, day attribute information AI, and predicted number of customers PV1, as well as an actual amount of goods received AV2. However, the second learning dataset DT2 only needs to include, for example, at least one type of information from time period identification information TI, day attribute information AI, and predicted number of customers PV1, as well as the actual amount of goods received AV2. The actual amount of goods received AV2 is the actual amount of goods received on a day to which the one or more time periods belong.

[0148] The third training data set DT3 includes the second training data set DT2 and a time-slot and department-specific actual man-hour value AV1 for each of one or more past time slots. The time-slot and department-specific actual man-hour value AV1 is the actual man-hours for each department for work in that time slot.

[0149] The fourth training data set DT4 includes, for each of one or more past time periods, for example, a set of time period identification information TI and day attribute information AI, and a time period and department-specific man-hour actual value AV1. However, the fourth training data set DT4 only needs to include, for example, one or more types of information from time period identification information TI and day attribute information AI, and a time period and department-specific man-hour actual value AV1.

[0150] Each of the second training data set DT2, the third training data set DT3, and the fourth training data set DT4 is associated with department identification information.

[0151] The generation unit 121 generates an incoming quantity prediction model PM4 for each of two or more department identification information by executing a machine learning learning algorithm LA1 using the first part of the second learning dataset DT2 corresponding to the department identification information as input data and the second part as training data.

[0152] Furthermore, for each of the two or more pieces of department identification information, the generation unit 121 generates a third man-hour prediction model PM3 by executing a machine learning learning algorithm LA3 using a first portion of the third training dataset DT3 associated with the department identification information and the predicted incoming quantity output from the incoming quantity prediction model PM4 as input data and a second portion as training data.The generation unit 121 then generates a decomposition model PM5 by executing a machine learning learning algorithm LA4 using a first portion of the fourth training dataset DT4 associated with the department identification information as input data and a second portion as training data.

[0153] In this way, in variant example 2 relating to the model, the planning device 1 itself can generate the incoming quantity prediction model PM4 by machine learning on the second training data set DT2, generate the third labor-hour prediction model PM3 by machine learning on the third training data set DT3, and generate the decomposition model PM5 by machine learning on the fourth training data set DT4.

[0154] (2-10) Work plan by time period and department In the present embodiment and the first and second modifications of the model, each of the two or more pieces of personnel information includes personnel identification information and corresponds to department identification information. The personnel identification information included in the personnel information is information that identifies the personnel corresponding to the personnel information. The department identification information is information that identifies the department to which the personnel belongs among the two or more departments.

[0155] The creation unit 124 creates a time period and department-specific work plan for two or more personnel to work in two or more departments during one or more time periods in the future, based on one or more time period and department-specific man-hour prediction values ​​PV2 acquired by the prediction unit 122, two or more pieces of personnel information, and department-specific information corresponding to each of the two or more pieces of personnel information.

[0156] The creation unit 124 may also create work plans by time period and department using two or more pieces of priority information associated with two or more pieces of department identification information.

[0157] However, the creation unit 124 may create the time-zone and department-specific work plan using one or more time-zone and department-specific man-hours and standard man-hours instead of one or more time-zone and department-specific man-hour predicted values ​​PV2.

[0158] In more detail, for example, the receiving unit 11 receives input of the workload, and the setting unit 123 sets the standard man-hours. The input workload is, for example, the total workload obtained by adding up the workloads for all departments and all time periods, but it may also be the workload for each department, or the workload for each time period and department. The set standard man-hours is, for example, a standard man-hour common to all departments and all time periods, but it may also be the standard man-hours for each department, or the standard man-hours for each time period and department. The creation unit 124 converts the input workload into one or more man-hours for each time period and department, for example, based on the set standard man-hours. Then, the creation unit 124 may create a work plan for each time period and department based on the one or more man-hours for each time period and department, two or more pieces of personnel information, and department-specific information associated with each of the two or more pieces of personnel information.

[0159] This also improves the accuracy of creating work plans for different time periods and departments.

[0160] (2-11) Variation of the plan creation device In a modified example of the plan creation device 1, the target is an item. The item is usually an item to be sold (a commodity), but may also be an item to be discarded (scrap).

[0161] The work in this modification is work related to the distribution of goods, such as transportation, home delivery, and storage.

[0162] The quantity information includes handling volume information. Handling volume information is information relating to the handling volume of goods during a period. Handling volume information is preferably a predicted handling volume value, but actual handling volume values ​​may also be used. The predicted handling volume value may be obtained using a handling volume prediction model. The actual handling volume value may be manually entered, for example, or may be automatically calculated.

[0163] According to the modified example of the plan creation device 1, when the work is logistics, the accuracy of man-hour prediction can be improved by focusing on the handling volume as the quantity of work objects.

[0164] (2-12) Update standard man-hours based on variances The reception unit 11 receives difference information from the difference detection device 2. As will be described later, the difference detection device 2 observes, by a camera 3 or the like, the actual work performed based on a work plan created by a plan creation device (creation unit 124), and detects a difference in man-hours between the planned work and the actual work from the work plan and the results of the observation, thereby acquiring difference information.

[0165] The update unit 125 updates the standard man-hours set by the setting unit 123 based on the difference information from the difference detection device 2 so that the difference is reduced.

[0166] This will reduce discrepancies and ultimately optimize work plans.

[0167] (2-13) Variations of prediction (2-13-1) Variation of prediction 1 The prediction unit 122 in this modification 1 executes one-stage prediction using the first man-hour prediction algorithm PA1 and three-stage prediction using the receipt volume prediction algorithm PA4, the third man-hour prediction algorithm PA3, and the decomposition algorithm PA5 in parallel or sequentially. Then, the difference between two prediction results (time period and department-specific man-hour prediction value PV2) corresponding to the two types of prediction is calculated, and if the absolute value of the calculated difference exceeds a predetermined threshold, at least one of the prediction result of the receipt volume prediction algorithm PA4 (received volume prediction value PV3) and the prediction result of the third man-hour prediction algorithm PA3 (total man-hour prediction value PV4) may be delivered to the output unit 13 in addition to the two prediction results.

[0168] The output unit 13 visualizes and outputs the three or four types of forecast results that have been handed over (two forecast results, the one-stage forecast and the three-stage forecast, and at least one of the forecast value PV3 of the incoming quantity and the forecast value PV4 of the total man-hours, which are the intermediate results of the three-stage forecast). This allows the manager to check the two forecast results when there is a large discrepancy between them and examine which one is more appropriate. Furthermore, by also checking at least one of the forecast value PV3 of the incoming quantity and the forecast value PV4 of the total man-hours, which are the intermediate results of the three-stage forecast, a more accurate examination becomes possible.

[0169] If the absolute value of the calculated difference is equal to or less than a predetermined threshold, the prediction unit 122 may deliver only one of the two prediction results to the output unit 13. This makes it possible to reduce the amount of information to be output. However, even if the absolute value of the difference is equal to or less than the threshold, the prediction unit 122 may deliver both of the two prediction results or a combined result (for example, an average) obtained based on a combination of the two prediction results to the output unit 13.

[0170] Although one-stage prediction and three-stage prediction are compared in the first modification, one-stage prediction and two-stage prediction, or two-stage prediction and three-stage prediction may be compared.

[0171] Furthermore, in Modification 1, the absolute value of the difference between the two prediction results is used, but the ratio of the difference to the prediction result may also be used. The ratio is, for example, the ratio of the difference to one of the two prediction results, but it may also be the ratio of the difference to the average value of the two prediction results. Furthermore, at least one of the average value and the variance may be used instead of the absolute value.

[0172] Alternatively, for example, three types of prediction (one-stage, two-stage, and three-stage) may be performed, and the absolute value of the difference between the maximum and minimum values ​​of the prediction results may be calculated for each of the three types of prediction. Then, each of the three differences corresponding to the three types of prediction may be compared with a threshold, and if the difference exceeds the threshold, the prediction result may be output.

[0173] (2-13-2) Variation of prediction 2 The prediction unit 122 in this modification 2 executes two or more types of algorithms in parallel or sequentially from among a first man-hour prediction algorithm PA1 that performs one-stage prediction, a second man-hour prediction algorithm PA2 that performs two-stage prediction, and a third man-hour prediction algorithm PA3 that performs three-stage prediction.The prediction unit 122 then executes a score calculation algorithm using the execution results of each algorithm (time period and department-specific man-hour prediction value PV2) as input, and obtains two or more scores corresponding to the two or more execution results.

[0174] The score calculation algorithm is an algorithm that takes the effort prediction value as input and outputs a score, which is an evaluation value related to the accuracy of the prediction.

[0175] The prediction unit 122 accumulates the scores thus obtained for each of two or more types of prediction algorithms. Then, for each of two or more types of prediction algorithms, the prediction unit 122 may calculate an average value of the scores over a certain period of time in the past, and transfer the execution result corresponding to the prediction algorithm with the largest average value to the creation unit 124.

[0176] According to prediction variant 2, predictions are made using multiple algorithms for the same prediction target, multiple scores corresponding to the multiple prediction results obtained are calculated, and the prediction result with the highest score is adopted, thereby further improving prediction accuracy.

[0177] (2-13-3) Variation of prediction 3 The prediction unit 122 in this third modification corrects the acquired man-hour predicted value and passes the corrected value to the creation unit 124. The correction here includes integer conversion (rounding up / rounding down / rounding to the nearest integer, etc.). The prediction unit 122 also switches between rounding up / rounding down the value so that the difference between the man-hour predicted value before and after integer conversion is minimized. This makes it possible to prevent the number of people from becoming a value that includes a decimal, such as "1.3 people."

[0178] (2-13-4) Prediction Variation 4: Another Three-Step Prediction The prediction unit 122 in this fourth modification acquires a total workload prediction value from the prediction dataset DP using a total workload prediction model. The prediction unit 122 may then decompose the acquired total workload prediction value into workloads for multiple time periods and departments using a decomposition model PM5, and acquire time period and departmental workload prediction values ​​PV2 from each of the multiple decomposed workloads for multiple time periods and departments using a workload prediction model. Note that the above-mentioned tasks are subdivisions of a business. In other words, a business is made up of one or more tasks.

[0179] (2-14) Other variations The processing unit 12 may calculate a department-by-department RE (Reasonable Expectancy) value and a department-by-department fixed workload, for example, using the actual number of visitors or the actual amount of incoming goods. The RE value is the ratio of the appropriate number of personnel (reasonable expectation) to the number of visitors. The fixed workload is the amount of fixed workload that accounts for the total workload. The fixed workload varies, for example, depending on at least one of the workload and the RE value. However, the fixed workload may be a constant amount regardless of the workload or the RE value.

[0180] The calculated departmental RE value and departmental fixed workload are delivered to the setting unit 123, which then uses the delivered departmental RE value and departmental fixed workload to set standard man-hours for each time period and department.

[0181] (3) Difference detection device 1, the difference detection device 2 includes a reception unit 21, a processing unit 22, and an output unit 23. However, the difference detection device 2 may include only the processing unit 22, and the reception unit 21 and the output unit 23 may be included in a terminal device (such as a tablet terminal) separate from the difference detection device 2.

[0182] The processing unit 22 includes an observation unit 221 and a detection unit 222 .

[0183] The reception unit 21 receives various types of information, such as work plans and countermeasure information.

[0184] The receiving unit 21 receives, for example, a work plan from the plan creation device 1 via the network 200. The receiving unit 21 also receives countermeasure information input by the administrator.

[0185] The processing unit 22 executes various types of processing, such as processing by an observation unit 221 and a detection unit 222.

[0186] The observation unit 221 observes the actual work being performed based on the work plan created by the plan creation device 1. Observation includes, for example, taking pictures using the camera 3, detecting positions using the LPS 4, and the like.

[0187] The observation unit 221 captures images of actual work using the camera 3 and acquires image information relating to the personnel performing the work and the work targets (visitors, goods, etc.). The observation unit 221 also performs position detection using the LPS 4 and acquires position information relating to the positions of the personnel and targets in the location (store, warehouse, etc.) where the work is being performed.

[0188] The detection unit 222 detects the difference in labor hours between the planned work and the actual work based on the work plan from the plan creation device 1 and the observation results (image information and position information) of the observation unit 221, and acquires difference information regarding the detected difference.

[0189] The difference in man-hours may be the difference in man-hours itself, or various gaps that occur in relation to the difference in man-hours. Examples of various gaps include the presence of a number of idle staff exceeding a threshold, the presence of a queue at a cash register that is longer than a threshold, the presence of a number of out-of-stock items exceeding a threshold in an item display area, the presence of a number of items waiting to be processed exceeding a threshold in an item processing area, etc.

[0190] The detection unit 222 may detect the above-mentioned various gaps based on, for example, the image information and position information acquired by the observation unit 221, and may acquire difference information indicating the difference in man-hours (hereinafter, man-hour difference information) based on the various detected gaps. strangenessThe information is, for example, the difference between the actual man-hours and the man-hours included in the work plan (usually including a positive or negative sign), but may also be a flag indicating a surplus or shortage (only information corresponding to a positive or negative sign).

[0191] In detail, the detection unit 222 acquires gap information using a gap detection model that receives image information and position information as input and outputs gap information related to various gaps. Next, the detection unit 222 acquires man-hour variance information using a difference detection model that receives gap information as input and outputs man-hour variance information. In this way, the gap information related to various gaps is converted into man-hour variance information that indicates the difference in man-hours.

[0192] The gap detection model is generated, for example, by executing a machine learning learning algorithm (LA) using a first part (a set of image information and position information) of a training dataset including a set of image information and position information and manually input gap information as input data and a second part (gap information) as training data. The difference detection model is generated, for example, by executing a machine learning learning algorithm (LA) using a first part (gap information) of a training dataset including gap information and manually input difference information as input data and a second part (difference information) as training data.

[0193] Moreover, the difference detection device 2 (processing unit 22) may further include a generating unit (not shown) that generates the gap detection model and the difference detection model as described above.

[0194] However, as will be described later, the difference detection device 2 (output unit 23) may visualize and output the observation results of the observation unit 221. A person who sees the visualized observation results may recognize a difference and input difference information related to the recognized difference via an input device, and the difference detection device 2 may accept the input difference information.

[0195] The output unit 23 outputs various types of information, such as the difference information described above, countermeasure information described below, and observation results described below.

[0196] The output unit 23 transfers the difference information acquired by the detection unit 222 to the plan creation device 1 by transmitting it, for example, via the network 200. Note that information may be exchanged between devices not only via a communication medium such as the network 200, but also via a portable recording medium such as a memory card.

[0197] Furthermore, the output unit 23 may output the observation results of the observation unit 221 (for example, display them on a display).

[0198] A person who sees the observation results (for example, a manager who manages the work) recognizes the gaps described above and inputs countermeasure information related to measures to reduce the gaps via an input device such as a keyboard. Countermeasure information may include, for example, information on the work plans and operation methods of departments with fewer gaps, instructions on increasing or decreasing personnel and changing their positions, etc.

[0199] In the difference detection device 2, the receiving unit 21 receives the input countermeasure information, and the output unit 23 outputs the received countermeasure information together with the difference information to the plan creation device 1. Note that the countermeasure information may be included in the difference information and output.

[0200] In the plan creation device 1, the receiving unit 11 receives the difference information and the countermeasure information, the updating unit 125 updates the standard man-hours based on the difference information, and the output unit 13 visualizes and outputs the received countermeasure information. Note that the countermeasure information may be directly transmitted from the manager to the personnel who will perform the work.

[0201] This allows for optimization of the work plan.

[0202] The difference information may be information relating to the total man-hours, or information relating to the man-hours by time period and by department.

[0203] (4) Operation of the planning optimization system The operation of the plan optimization system 100 will be explained below using the flowcharts of Figures 5 and 6. Note that in the explanation of the flowcharts, detailed explanation of the operation of each unit will be omitted.

[0204] The plan creation device 1 constituting the plan optimization system 100 operates, for example, according to the flowchart in Fig. 5. The processing of this flowchart starts when the plan creation device 1 is powered on, and ends when the power is turned off.

[0205] The processing unit 12 constituting the plan creation device 1 determines whether the receiving unit 11 has received a data set via an input device (step S1). If the receiving unit 11 has received a data set via an input device, the process proceeds to step S2. If the data set has not been received, the process proceeds to step S5.

[0206] The processing unit 12 determines whether the dataset accepted in step S1 is for prediction based on a comparison between the date included in the current time information acquired from a built-in clock of the processor or the like and the date included in the dataset (step S2). If the dataset accepted in step S1 is for prediction, the process proceeds to step S4. If the accepted dataset is not for prediction, the accepted dataset is considered to be for learning, and the process proceeds to step S4.

[0207] The data sets determined to be for training are stored in an area for training data sets (not shown) in memory. When a predetermined number of training data sets are stored in memory, the generation unit 121 executes a machine learning learning algorithm (LA) using a first portion of the stored training data sets (DT1 to DT4) as input data and a second portion as training data, thereby generating models (PM1 to PM5) that use the prediction data set (DP) as input and the labor-hour prediction value (PV2) as output (step S3). Thereafter, the process returns to step S1.

[0208] The data sets determined to be for prediction are stored in an area for prediction data sets (not shown) in memory. When a predetermined number of prediction data sets (DP) are stored in memory, the prediction unit 122 uses the stored learning data sets (DA) as input and executes prediction algorithms (PA1 to PA5) using the models (PM1 to PM5) generated in step S3 to obtain one or more predicted effort values ​​(PV2) corresponding to one or more future periods (step S4). Thereafter, the process returns to step S1.

[0209] The processing unit 12 determines whether the receiving unit 11 has received a plan creation instruction via the input device (step S5). If the receiving unit 11 has received a plan creation instruction via the input device, the process proceeds to step S6. If the plan creation instruction has not been received, the process proceeds to step S9.

[0210] The setting unit 123 sets the standard man-hours (step S6). The creation unit 124 creates a work plan for the two or more personnel to perform the work in one or more future periods based on the one or more predicted man-hour values ​​(PV2) acquired in step S4 and two or more pieces of personnel information stored in the memory (step S7).

[0211] The output unit 13 outputs the work plan created in step S7 (step S8). Based on the work plan output here, work is carried out by two or more people. The output work plan is also handed over to the difference detection device 2. Thereafter, the process returns to step S1.

[0212] Processing unit 12 determines whether or not receiving unit 11 has received difference information from difference detection device 2 (step S9). If receiving unit 11 has received difference information from difference detection device 2, the process proceeds to step S10. If difference information has not been received, the process returns to step S1.

[0213] The update unit 125 updates the standard man-hours set in step S6 based on the difference information received in step S9 so as to reduce the difference between the planned work and the actual work (step S10), after which the process returns to step S10.

[0214] The difference detection device 2 operates, for example, according to the flowchart of Fig. 6. The processing of this flowchart starts when the power to the difference detection device 2 is turned on, and ends when the power is turned off.

[0215] The observation unit 221 constituting the difference detection device 2 observes, via the camera 3 and the PLS 4, the work that is actually performed based on the work plan created by the plan creation device 1 (step S21).

[0216] The detection unit 222 detects a difference in man-hours between the work planned by the work plan and the actual work from the plan information and the results of the observation in step S21, and acquires difference information regarding the detected difference (step S22). The output unit 23 outputs the difference information acquired in step S22 (step S23). The output difference information is passed to the plan creation device 1. Thereafter, the process returns to step S1.

[0217] Based on the outputted difference information, the manager or the like who manages the work may input countermeasure information to reduce the difference, the reception unit 21 may receive the input countermeasure information, and the output unit 23 may pass the received countermeasure information to the plan creation device 1.

[0218] (5) Specific examples (5-1) Example 1: One-step prediction Assume that the current time is 23:00 on February 20, 2021. At this time, for example, two or more pieces of personnel information as shown in FIG. 7 are stored in the memory of the plan creation device 1.

[0219] The personnel information in this specific example 1 includes personnel identification information and work performance information. The work performance information is composed of a set of pairs of period identification information and department identification information. The period identification information is time period identification information, and is composed of year, month, date, and time. The personnel information may also include skill information, vacation information, etc.

[0220] The work performance information corresponding to the personnel identification information "aa" is, for example, "···{(2021,2,20,10),AA},{(2021,2,20,11),AA}···{(2021,2,20,17),BB}". The work performance information corresponding to the personnel identification information "bb" is, for example, "···{(2021,2,20,8),BB}···{(2021,2,20,12),BB}". Note that "aa" and "bb" are names, and "AA" and "BB" are department names.

[0221] The memory of the plan creation device 1 further stores two or more (seven in the figure) data sets as shown in Fig. 8. Each data set is associated with an ID ("1", "2", etc.). Hereinafter, a data set associated with ID "i" will be referred to as the "ith data set".

[0222] The data set in this specific example 1 includes department-specific information, man-hours, number of visitors, period-specific information, and period attribute information. Man-hours are man-hours per hour and include predicted man-hours and actual man-hours. Number of visitors is the number of visitors per day and includes predicted number of visitors and actual number of visitors. Period-specific information is time-zone-specific information and is composed of year, month, date, and hour. Period attribute information is day attribute information and includes day of the week, day of the week category, sale day flag, and weather information.

[0223] Figure 8 shows multiple datasets corresponding to multiple time periods (one year and one month) from midnight on February 21, 2020, including the present (February 20, 2021) until 11:00 PM on March 20, 2021. Of these multiple datasets, some (for example, datasets 1 to 5) are for learning, and others (datasets 6 and 7) are for prediction.

[0224] For example, the first dataset, which is the training dataset, is "AA, (120, 130), (5000, 5200), (2020, 2, 21, 5), (Friday, weekday, 0, sunny)". Similarly, the second dataset is "AA, (150, 140), (5000, 5200), (2020, 2, 21, 6), (Friday, weekday, 0, sunny)". The third dataset is "AA, (50, 60), (5000, 5200), (2020, 2, 21, 23), (Friday, weekday, 0, sunny)".

[0225] The fourth dataset is "AA,(0,0),(6000,5900),(2020,2,22,0),(Sat,Sat / Sun / Holiday,0,Cloudy)". The fifth dataset is "AA,(360,380),(8000,7000),(2020,3,20,11),(Fri,Sat / Sun / Holiday,1,Rainy)".

[0226] The sixth dataset, which is a prediction dataset, is "AA,(-,-),(7000,-),(2021,2,21,5),(Sun,Sat,Sun,Holiday,0,-)". Note that "-" indicates that the data does not yet exist. Similarly, the seventh dataset is "AA,(-,-),(7000,-),(2021,2,21,6),(Sun,Sat,Sun,Holiday,0,-)".

[0227] The plurality of data sets are read from memory, and the reception unit 11 constituting the plan creation device 1 receives the read plurality of data sets. The processing unit 12 acquires current time information from a built-in clock of the processor or the like, and determines whether each of the plurality of data sets is for prediction by comparing the date "2021,2,20" included in the acquired current time information with the date (e.g., "2020,2,21") included in each of the read plurality of data sets. Here, the first to fifth data sets, etc. are determined to be for learning, and the sixth, seventh data sets, etc. are determined to be for prediction.

[0228] The generation unit 121 generates a first man-hour prediction model PM1 by executing a machine learning (in this example, LightGBM) learning algorithm LA using a first portion of the first to fifth learning datasets, etc. as input data and a second portion as training data.

[0229] The prediction unit 122 acquires multiple man-hour prediction values ​​PV2 corresponding to multiple future time periods by executing the machine learning (LightGBM) prediction algorithm PA1 using the sixth, seventh, and other prediction data sets as input to the generated first man-hour prediction model PM1. Here, it is assumed that the acquired man-hour prediction values ​​are "140" and "160" corresponding to the time period identification information "2021,2,21,5" and "2021,2,21,6".

[0230] The processing unit 12 adds the acquired man-hour prediction values ​​"140" and "160" to the sixth and seventh data sets for prediction. As a result, the sixth and seventh data sets shown in FIG. 8 are updated as shown in FIG. 9A. In the sixth and seventh data sets shown in FIG. 9A, the man-hour prediction values ​​"-" have been updated to "140" and "160".

[0231] After that, when the current time passes midnight on February 22, 2021, the processing unit 12 acquires from an external device (such as a store server) the actual man-hour values ​​for each time slot and the actual number of visitors for the previous day, February 21, 2021. Here, for example, it is assumed that actual man-hour values ​​of "150" and "160" for the 5:00 and 6:00 a.m. hours and an actual number of visitors value of "7,500" are acquired.

[0232] The processing unit 12 adds the acquired man-hour actual values ​​"150" and "160" and the actual number of visitors value "7500" to the sixth and seventh data sets for prediction. As a result, the sixth and seventh data sets shown in FIG. 9A are updated as shown in FIG. 9B. In the sixth and seventh data sets shown in FIG. 9B, the man-hour actual value "-" has been updated to "150" and "160", and the actual number of visitors value "-" has been updated to "7500".

[0233] (5-2) Example 2: Two-stage prediction Below, only the differences from Specific Example 1 will be described. The dataset in Specific Example 2 further includes information regarding the amount of stock received in addition to the dataset in Specific Example 1 (see FIG. 8). The information regarding the amount of stock received includes a predicted value of stock received PV3 and an actual value of stock received AV2. Hereinafter, the predicted value of stock received PV3 will be referred to as "p" and the actual value of stock received AV2 will be referred to as "q", and the information regarding the amount of stock received will be referred to as "p, q".

[0234] The first dataset in this specific example 2 further includes information on the amount of goods received, "100, 110", in the first dataset of FIG. 8. Similarly, the second dataset further includes information on the amount of goods received, "100, 110", in the second dataset of FIG. 8. The third dataset further includes information on the amount of goods received, "100, 110", in the third dataset of FIG. 8. The fourth dataset further includes information on the amount of goods received, "120, 110", in the fourth dataset of FIG. 8. The fifth dataset further includes information on the amount of goods received, "150, 140", in the fifth dataset of FIG. 8. The sixth dataset further includes information on the amount of goods received, "-, -", in the sixth dataset of FIG. 8. The seventh dataset further includes information on the amount of goods received, "-, -", in the seventh dataset of FIG. 8.

[0235] The generation unit 121 generates an incoming quantity prediction model PM4 by executing a machine learning (LightGBM) learning algorithm LA using a first portion of the first to fifth learning datasets, etc. as input data and a second portion as training data. The generation unit 121 also generates a second man-hour prediction model PM2 by executing the machine learning learning algorithm LA using a first portion of the first to fifth learning datasets, etc. as input data and a second portion as training data.

[0236] The prediction unit 122 acquires multiple predicted arrival quantities PV3 corresponding to multiple future days by executing an arrival quantity prediction algorithm PA4 using an arrival quantity prediction model PM4 with the sixth, seventh, and other prediction data sets as input. The prediction unit 122 also acquires two or more predicted man-hour values ​​corresponding to one or more future time periods by executing a second man-hour prediction algorithm PA2 using a second man-hour prediction model PM2 with the acquired arrival quantity prediction values ​​PV3 and the sixth, seventh, and other prediction data sets as input.

[0237] Here, for example, it is assumed that the predicted arrival quantity value "150" corresponding to February 21, 2021 and the predicted man-hour values ​​"140" and "160" corresponding to the time zone specification information "2021,2,21,5" and "2021,2,21,6" are obtained.

[0238] The processing unit 12 adds the acquired incoming quantity forecast value "150" and man-hour forecast values ​​"140", "160" to the sixth and seventh data sets for the forecast. As a result, in the sixth and seventh data sets, the incoming quantity forecast value "-" is updated to "150", and the man-hour forecast value "-" is updated to "140", "160".

[0239] After that, when the current time passes midnight on February 22, 2021, the processing unit 12 acquires from an external device the actual number of visitors for the day, the actual amount of stock received for the day, and the actual man-hours for each time period for the previous day, February 21, 2021. Here, it is assumed that the actual number of visitors is "7500," the actual amount of stock received is "170," and the actual man-hours for the 5:00 and 6:00 a.m. hours are "150" and "160."

[0240] The processing unit 12 adds the acquired actual number of visitors value "7500" etc., actual amount of received product value "170" etc., and actual man-hours value "150", "160" etc. to the sixth, seventh, etc. data sets for prediction. As a result, in the sixth and seventh data sets, the actual number of visitors value "-" is updated to "7500", the actual amount of received product value "-" is updated to "170", and the actual man-hours value "-" is updated to "150", "160".

[0241] (5-3) Example 3: Three-stage prediction Below, only the differences between Specific Example 3 and Specific Example 2 will be described. The dataset in Specific Example 3 further includes information regarding the total man-hours compared to the dataset in Specific Example 2. The information regarding the total man-hours includes a predicted total man-hour value PV4 and an actual total man-hour value. Below, the predicted total man-hour value PV4 will be referred to as "r" and the actual incoming volume value AV2 will be referred to as "s", and the information regarding the incoming volume will be referred to as "r, s".

[0242] The first dataset in this specific example 3 further includes information on the total man-hours, "4000, 4300", in the first dataset in specific example 2. Similarly, the second dataset further includes information on the total man-hours, "4000, 4300", in the second dataset in specific example 2. The third dataset in the third dataset of FIG. 8 further includes information on the total man-hours, "4000, 4300". The fourth dataset in the fourth dataset of FIG. 8 further includes information on the total man-hours, "4500, 4200". The fifth dataset in the fifth dataset of FIG. 8 further includes information on the total man-hours, "5000, 4500". The sixth dataset in the sixth dataset of FIG. 8 further includes information on the total man-hours, "-,-". The seventh dataset in the seventh dataset of FIG. 8 further includes information on the total man-hours, "-,-".

[0243] The generation unit 121 generates an incoming quantity prediction model PM4 by executing a machine learning (LightGBM) learning algorithm LA using a first portion of the first to fifth training datasets, etc. as input data and a second portion as training data. The generation unit 121 also generates a third man-hour prediction model PM3 by executing the machine learning learning algorithm LA using a first portion of the first to fifth training datasets, etc. as input data and a second portion as training data. The generation unit 121 also generates a decomposition model PM5 by executing the machine learning learning algorithm LA using a first portion of the first to fifth training datasets, etc. as input data and a second portion as training data.

[0244] The prediction unit 122 acquires multiple predicted incoming volume values ​​PV3 corresponding to multiple future days by executing an incoming volume prediction algorithm PA4 using an incoming volume prediction model PM4 with the sixth, seventh, and other prediction data sets as input. The prediction unit 122 also acquires a total man-hour predicted value PV4 for a day to which one or more future time slots belong by executing a third man-hour prediction algorithm PA3 using a third man-hour prediction model PM3 with the acquired incoming volume predicted value PV3 and the sixth, seventh, and other prediction data sets as input. The prediction unit 122 also acquires two or more predicted man-hour values ​​corresponding to one or more future time slots by executing a decomposition algorithm PA5 using a decomposition model PM5 with the acquired total man-hour predicted value PV4 as input.

[0245] Here, for example, the predicted arrival quantity value "150" and total labor hour value "4500" corresponding to February 21, 2021, and the predicted labor hour values ​​"140" and "160" corresponding to the time zone specification information "2021,2,21,5" and "2021,2,21,6" are acquired.

[0246] The processing unit 12 adds the acquired incoming quantity forecast value "150", total man-hour forecast value "4500", and man-hour forecast values ​​"140", "160" to the sixth and seventh data sets for the forecast. As a result, in the sixth and seventh data sets, the incoming quantity forecast value "-" is updated to "150" etc., the total man-hour forecast value "-" is updated to "4500" etc., and the man-hour forecast value "-" is updated to "140", "160".

[0247] After that, when the current time passes midnight on February 22, 2021, the processing unit 12 acquires from an external device the actual number of visitors for the day, the actual amount of stock received for the day, the predicted total man-hours for the day, and the actual man-hours for each time period for the previous day, February 21, 2021. Here, it is assumed that the actual number of visitors is "7500," the actual amount of stock received is "170," the actual total man-hours is "4600," and the actual man-hours for the 5:00 and 6:00 a.m. hours are "150" and "160."

[0248] The processing unit 12 adds the acquired actual number of visitors value "7500", actual amount of stock "170", actual total man-hours value "4600", and actual man-hours values ​​"150", "160" to the sixth and seventh datasets for prediction. As a result, in the sixth and seventh datasets, the actual number of visitors value "-" is updated to "7500", the actual amount of stock "-" is updated to "170", the actual total man-hours value "-" is updated to "4600", and the actual man-hours value "-" is updated to "150", "160".

[0249] (5-4) Example 4: Planning The creation unit 124 creates a work plan (a set of personnel identification information and a set of work schedule information: see Figure 10) for work to be performed by two or more personnel, based on one or more man-hour prediction values ​​acquired by the prediction unit 122 (for example, the man-hour prediction value "140" constituting the sixth data set and the man-hour prediction value "160" constituting the seventh data set, etc.: see Figure 8) and two or more pieces of personnel information (for example, a set of personnel identification information and a set of multiple pieces of work performance information: see Figure 7).

[0250] (6) Planning methods and programs Note that functions similar to those of the plan creation device 1 according to the above embodiment may be embodied as a plan creation method, a (computer) program, or a non-transitory recording medium on which a program is recorded. Note that the plan creation method is a method that includes at least step S4 (prediction step) and step S8 (creation step S8) among the various steps described above. Also, the program is a program for causing a computer to execute the above plan creation method.

[0251] The plan optimization system 100 in the present disclosure includes a computer system. The computer system is primarily composed of a processor and memory as hardware. For example, the computer system includes a processor and memory of a first server constituting the plan creation device 1 and a processor and memory of a second server constituting the difference detection device 2. The computer system may also include a processor and memory of a terminal such as a tablet terminal. The functions of the plan optimization system 100 in the present disclosure are realized by the processor executing a program stored in the memory of the computer system. The program may be pre-stored in the memory of the computer system, provided via a telecommunications line, or provided by being stored on a non-transitory recording medium readable by the computer system, such as a memory card, optical disc, or hard disk drive. The processor of the computer system is composed of one or more electronic circuits, including a semiconductor integrated circuit (IC) or a large-scale integrated circuit (LSI). The integrated circuits, such as ICs and LSIs, are referred to by different names depending on the degree of integration, and include integrated circuits called system LSIs, very large-scale integration (VLSI), or ultra-large-scale integration (ULSI). Furthermore, a field-programmable gate array (FPGA), which is programmed after the LSI is manufactured, or a logic device capable of reconfiguring the connections within the LSI or the circuit partitions within the LSI, can also be employed as a processor. Multiple electronic circuits may be integrated into a single chip or distributed across multiple chips. Multiple chips may be integrated into a single device or distributed across multiple devices. The computer system referred to here includes a microcontroller having one or more processors and one or more memories. Therefore, a microcontroller may also be composed of one or more electronic circuits, including a semiconductor integrated circuit or a large-scale integrated circuit.

[0252] Furthermore, at least some of the functions of the plan optimization system 100, for example, some of the functions of the plan creation device 1, may be realized by cloud (cloud computing) or the like.

[0253] Conversely, in an embodiment, at least some of the functions of the plan optimization system 100 that are distributed across multiple devices may be consolidated into a single housing. For example, some of the functions of the plan optimization system 100 that are distributed across the plan creation device 1 and the difference detection device 2 may be consolidated into a single housing. [Explanation of symbols]

[0254] 1 Planning device 11 Reception 12 Processing section 121 Generation part 122 Prediction Department 123 Settings 124 Creation Department 125 Update section 13 Output section 2. Difference detection device 21 Reception 22 Processing section 221 Observation Section 222 Detection unit 23 Output section 3 Camera 4. LPS 100 Planning Optimization System 200 Network

Claims

1. A planning device for creating a work plan for a future period including multiple days from the next day onwards, a prediction unit that acquires a man-hour prediction value corresponding to the future period by executing a prediction algorithm using a model that receives as input a prediction data set including date information, day attribute information, and arrival amount information or visitor number information for each of the plurality of days, and outputs a man-hour prediction value that predicts the work man-hours for the future period; and a creation unit that creates a work plan for the work target in the future period based on the man-hour prediction value acquired by the prediction unit and personnel information related to personnel who will perform the work, Planning device.

2. The system further comprises a generation unit that generates the model by executing a machine learning learning algorithm using two or more learning datasets including, for each of a plurality of days included in a past period, the date information, the day attribute information, and the arrival amount information or the number of visitors information, and actual labor hours that are the actual labor hours for the work in the past period. The plan creation device according to claim 1 .

3. The work is sales-related work, The day attribute information includes a flag indicating whether the day is a sale day. The plan creation device according to claim 1 or 2.

4. The day attribute information further includes weather information regarding the weather on the day at the location where the work is performed. The plan creation device according to any one of claims 1 to 3.

5. The prediction dataset is associated with department identification information that identifies any one of two or more departments, the prediction unit acquires the man-hour prediction value for each department for each of two or more department identification information corresponding to the two or more departments by inputting the prediction data set associated with the department identification information; The plan creation device according to any one of claims 1 to 4.

6. A setting unit that sets a standard man-hour for the work; a receiving unit that receives difference information from a difference detection device that detects a difference in man-hours between the work planned by the work plan and the actual work, and acquires difference information regarding the detected difference; an updating unit that updates the standard man-hours set by the setting unit based on the difference information so that the difference is reduced, The plan creation device according to any one of claims 1 to 5.

7. A setting unit that sets a standard man-hour for the work; a detection unit that detects a gap that occurs in relation to a difference in man-hours between the work planned by the work plan and the actual work, and acquires difference information regarding the difference in man-hours between the work planned by the work plan and the actual work based on the detected gap; an updating unit that updates the standard man-hours so that the difference information is reduced, The plan creation device according to any one of claims 1 to 6.

8. A planning method executed by one or more processors to generate a work plan for a future period including a plurality of days from the next day onwards, comprising: a prediction step of acquiring a man-hour prediction value corresponding to the future period by executing a prediction algorithm using a model that receives as input a prediction dataset including date information, day attribute information, and arrival amount information or visitor number information for each of the plurality of days, and outputs a man-hour prediction value that predicts the work man-hours for the future period; a creating step of creating a work plan for the work target in the future period based on the man-hour predicted value acquired in the predicting step and personnel information related to personnel who will perform the work, How to create a plan.

9. A program for causing one or more processors to execute a planning method for creating a work plan for a future period including a plurality of days from the next day onward, The plan creation method includes: a prediction step of acquiring a man-hour prediction value corresponding to the future period by executing a prediction algorithm using a model that receives as input a prediction dataset including date information, day attribute information, and arrival amount information or visitor number information for each of the plurality of days, and outputs a man-hour prediction value that predicts the work man-hours for the future period; a creating step of creating a work plan for the work target in the future period based on the man-hour predicted value acquired in the predicting step and personnel information related to personnel who will perform the work, program.

Images