WO2016151640A1

WO2016151640A1 - Learning system, method, and program

Info

Publication number: WO2016151640A1
Application number: PCT/JP2015/001745
Authority: WO
Inventors: 紗和子見上; 洋介本橋
Original assignee: 日本電気株式会社
Priority date: 2015-03-26
Filing date: 2015-03-26
Publication date: 2016-09-29
Also published as: JPWO2016151640A1

Abstract

Provided is a learning system capable of generating learning models having high number-of-clients prediction accuracy. A data storage means 71 stores sets of data that have associated therein a result value for the number of clients at a facility and information relating to reservations for the facility from prior to a point in time corresponding to the result value. A learning model generation means 72 uses the data sets as learning data and generates learning models for predicting the number of clients at the facility at a target point in time for prediction, said learning models using, as explanatory variables, information specified from reservation information.

Description

Learning system, method and program

The present invention relates to a learning system, a learning method, and a learning program that generate a learning model, and more particularly to a learning system, a learning method, and a learning program that generate a learning model used for predicting the number of customers in a facility.

Patent Document 1 describes a method for calculating the expected number of visitors to the venue. In the method described in Patent Document 1, a number obtained by adding the number of ticket sales to the number of visitors is set as the expected number of entrances.

Also, Patent Document 2 describes a system for predicting the number of transportation reservations. The system described in Patent Document 2 determines an optimum reservation pattern from one or more reservation patterns stored in a reservation pattern database based on the accumulated result of reservation status. Then, the system calculates the final reservation number of the transportation facility based on the function of the reservation pattern and the stored reservation situation. The system described in Patent Document 2 calculates the coefficient of the curve function of the reservation pattern by multiple regression analysis or the like based on the past reservation results.

JP 2003-303361 A JP 2004-295532 A

It is preferable that the number of customers can be accurately predicted at various facilities such as movie theaters. For example, if the number of customers can be accurately predicted, convenience such as easy determination of the amount of products (for example, snacks such as popcorn in the case of movie theaters) prepared for sale to customers in the facility can be obtained.

Therefore, an object of the present invention is to provide a learning system, a learning method, and a learning program that can solve the technical problem of enabling generation of a learning model with high customer number prediction accuracy.

The learning system according to the present invention includes a data storage means for storing a set of data in which the actual value of the number of customers of the facility is associated with information related to the reservation for the facility prior to the time corresponding to the actual value; A learning model for predicting the number of customers in a facility, comprising learning model generation means for generating a learning model using information specified from reservation information as an explanatory variable using a set of data as learning data It is characterized by.

Further, the learning method according to the present invention comprises a learning system comprising a data storage means for storing a set of data in which the actual value of the number of customers in the facility is associated with the information related to the reservation for the facility prior to the time corresponding to the actual value. A learning model for predicting the number of customers in a facility at the time of prediction, and a learning model that uses information specified from reservation information as explanatory variables, and a set of data as learning data It is characterized by generating using.

Further, the learning program according to the present invention is a computer provided with data storage means for storing a set of data in which the actual value of the number of customers of the facility is associated with the information related to the reservation for the facility before the time corresponding to the actual value. A learning program installed on a computer, which is a learning model for predicting the number of customers of a facility at the time of prediction, and that uses information specified from information related to reservation as an explanatory variable, A learning model generation process for generating a set using learning data is executed.

The technical means of the present invention can generate a learning model with high customer number prediction accuracy.

It is a block diagram which shows the structural example of the learning system of this invention. It is a schematic diagram which shows the example of the collection of the data memorize | stored in a data storage part. It is a block diagram which shows the structural example of the learning system in case a prediction part is not included. It is a flowchart which shows the example of process progress in which a learning system produces | generates a learning model. It is a schematic diagram which shows the example of the collection of data which a data storage part memorize | stores in the modification of 1st Embodiment. It is a schematic diagram which shows the example of the collection of data memorize | stored in a data storage part in 2nd Embodiment. It is a schematic diagram which shows the example of the data which a data storage part memorize | stores in the modification of 2nd Embodiment. It is a schematic diagram which shows the example of the collection of data memorize | stored in a data storage part in 3rd Embodiment. It is a schematic diagram which shows the example of the data which a data storage part memorize | stores in the modification of 3rd Embodiment. It is a schematic block diagram which shows the structural example of the computer which concerns on each embodiment of this invention. It is a block diagram which shows the outline | summary of the learning system of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In each embodiment described below, a case where the number of customers per day in a movie theater is predicted will be described as an example, but the facility for which the number of customers is predicted is not limited to a movie theater.

Embodiment 1. FIG.
FIG. 1 is a block diagram illustrating a configuration example of a learning system according to the present invention. The learning system 1 of the present invention includes, for example, a data storage unit 2, a learning unit 3, and a prediction unit 4.

The data storage unit 2 stores a set of data in which the actual value of the number of customers of the facility (in this example, a movie theater) is associated with information related to the reservation for the movie theater before the time corresponding to the actual value. Storage device. The time point corresponding to the actual value means the time point when the actual value is measured, and can also be referred to as the actual value measurement time point. In each embodiment, a case where the actual value of the number of customers is measured in units of “1 day” will be described as an example. Therefore, in the example shown below, the time point corresponding to the actual value is the day when the actual value is measured, and is hereinafter referred to as the actual value measurement date.

FIG. 2 is a schematic diagram showing an example of a set of data stored in the data storage unit 2. In each embodiment, the unit of time for measuring the actual value is “1 day”. Therefore, in the example shown in FIG. 2, the actual value of the number of customers in the movie theater is associated with the information related to the reservation for the movie theater before the actual value measurement date every day. However, not only information related to the reservation is associated with the actual value, but also other information may be associated with the actual value in addition to the information regarding the reservation.

FIG. 2 exemplifies data in which the actual value of the number of customers, the number of movie theater reservations 9 days before the actual value measurement date, the day of the week of the actual value measurement date, and the like are associated with each other every day. Each day corresponds to the actual value measurement date. The number of movie theater reservations 9 days before the actual value measurement date is the number of people who have made reservations to use the movie theater 9 days before the actual value measurement date. For example, the number of reservations 9 days ago associated with the actual value on December 24 is 1000 (see FIG. 2). This means that there were 1000 people who made a reservation to use the movie theater on December 24th on December 15th (9 days before December 24th). The same applies to other date data. However, depending on the type of facility, a reservation may be made without specifying a use date. In that case, the number of persons who have reserved the use of the facility without specifying the use date nine days before the actual value measurement date may be the number of reservations nine days before the actual value measurement date. This also applies to each embodiment described later.

In the example shown in FIG. 2, the number of reservations 9 days before the actual value measurement date is shown. “9 days ago” is an example, and there is no particular limitation on when the number of reservation users before the actual value measurement date is stored in the data storage unit 2.

In addition, in the example shown in FIG. 2, in addition to information related to the reservation (the number of reservations 9 days ago), information on the day of the week of the actual value measurement date is also associated. Moreover, forecast values, such as the weather and temperature of a performance value measurement day, may be matched with a performance value. In this way, information other than information related to the reservation may be associated with the actual value. In FIG. 2, the day of the week is illustrated, but what information is associated with the actual value as information other than information related to the reservation is not particularly limited.

For example, an administrator of the learning system 1 (hereinafter simply referred to as an administrator) prepares data in which the actual number of customers, the number of reservations 9 days ago, the day of the week, and the like are associated with each other. A set of data is stored in the data storage unit 2. In the following embodiments, in order to simplify the description, an example will be described in which the administrator stores a set of data in the data storage unit 2, but an aspect in which the set of data is stored in the data storage unit 2 will be described. Is not particularly limited.

The learning unit 3 uses a set of data stored in the data storage unit 2 (see FIG. 2) as learning data to generate a learning model for predicting the number of customers in the movie theater. The learning model is used to predict the number of customers in the movie theater at the prediction target time point (in this example, the prediction target date). In addition, for example, each future day may be set as the prediction target day. In addition, the learning unit 3 generates a learning model having information specified from information related to reservation as an explanatory variable. The information specified from the information related to the reservation may be information related to the reservation itself, or may be information obtained by performing an operation on the information related to the reservation. In the present embodiment, it is assumed that the number of reservations 9 days before the prediction target date is given when the number of customers is predicted. And the learning part 3 shall generate | occur | produce the learning model which uses the number of reservation persons of the movie theater 9 days before a prediction object day as an explanatory variable according to learning data (refer FIG. 2). However, the explanatory variables used in the learning model are not limited to the number of reservations 9 days before the prediction target date, and other items are also used as explanatory variables. For example, when learning data including day information is used as shown in FIG. 2, the day of the week is also used as an explanatory variable in the learning model.

Note that a variable representing data used as a parameter in prediction is called an “explanatory variable”, and a variable representing a prediction target is called an “object variable”.

The method by which the learning unit 3 generates the learning model is not particularly limited. For example, the learning unit 3 may generate a learning model by regression analysis using learning data. Alternatively, the learning unit 3 may generate a learning model using another machine learning algorithm. This also applies to each embodiment described later.

The learning model may be, for example, a prediction formula for calculating the value of the objective variable. Hereinafter, in order to simplify the explanation, a case where the learning model is a prediction formula represented by the following formula (1) will be described as an example. However, the format of the learning model is not limited to the format of the prediction formula.

_{_{_{_{y = a 1 x 1 + a}}}} 2 x 2 + ··· + a n x n + b formula (1)

y is an objective variable representing a predicted value. In this example, y represents the predicted value of the number of customers in the movie theater on the prediction target date. x ₁ to x _n are explanatory variables. _a ₁ ~ a _n are coefficients of the explanatory variables. b is a constant term. The value of _a ₁ ~ a _n and b are based on training data, as determined by the learning unit 3.

In each embodiment, a case where each day is a prediction target day will be described as an example. The value of each explanatory variable used for prediction of the number of customers in the movie theater on the prediction target day is input to the prediction unit 4 from the administrator for each day. In the present embodiment, in accordance with the learning data (see FIG. 2), the number of movie theater reservations, the day of the week, etc. 9 days before the prediction target date are input for each prediction target date. The number of movie theater reservations 9 days before the prediction target date is the number of people who have made reservations to use the movie theater on the prediction target date 9 days before the prediction target date. However, as described above, depending on the type of facility, it may be possible to make a reservation without specifying a use date. In that case, the number of persons who have reserved the use of the facility without specifying the use date 9 days before the prediction target day may be the number of reservations 9 days before the prediction target date. This also applies to each embodiment described later.

The prediction unit 4 calculates the predicted value y of the number of customers in the movie theater on the prediction target day by applying the value of each input explanatory variable to the learning model. As in this example, when the learning model is represented by the prediction expression shown in Expression (1), the prediction unit 4 sets values for x ₁ to x _n in the prediction expression according to the value of the input explanatory variable. Is used to calculate the predicted value y. Hereinafter, an operation in which the prediction unit 4 substitutes values for x ₁ to x _n in the prediction formula according to the value of the explanatory variable will be described.

・・・ There are continuous type and categorical type as explanatory variable types.

* Continuous variables take numerical values. For example, the number of reservations 9 days before the prediction target date is a continuous variable.

Categorical variables take items as values. For example, the day of the week is a categorical variable.

One continuous variable corresponds to one of the explanatory variables x ₁ to x _n in the prediction formula. The prediction unit 4 assigns the value (numerical value) of the explanatory variable corresponding to the continuous variable to the corresponding explanatory variable in the prediction formula. For example, it is assumed that among explanatory variables x ₁ to x _n in the prediction formula, x ₂ corresponds to the number of reservations 9 days ago. In that case, the prediction unit 4 substitutes the number of subscriber 9 days prior to the prediction target date x _2.

Each value of one categorical variable corresponds to one of the explanatory variables x ₁ to x _n in the prediction formula. For example, each possible value of the categorical variable “day of the week” (each item such as “Sunday”, “Monday”) corresponds to one of the explanatory variables x ₁ to x _n in the prediction formula, respectively. To do. The prediction unit 4 substitutes one of two values (in this example, 0 and 1) for each explanatory variable in the prediction formula corresponding to each value of the categorical variable. For example, it is assumed that the value of “day of the week” that has been input is “Monday”. In this case, the prediction unit 4 substitutes 1 for the explanatory variable in the prediction formula corresponding to Monday, and substitutes 0 for each explanatory variable in the prediction formula corresponding to each day of the week other than Monday.

As described above, the prediction unit 4 substitutes values for x ₁ to x _n in the prediction formula in accordance with the value of the given explanatory variable, thereby obtaining the predicted value y of the number of customers of the movie theater on the prediction target day. calculate.

The learning unit 3 and the prediction unit 4 are realized by a CPU of a computer that operates according to a learning program, for example. In this case, for example, the CPU reads a learning program from a program recording medium such as a program storage device (not shown in FIG. 1) of the computer, and operates as the learning unit 3 and the prediction unit 4 according to the learning program. Further, the learning unit 3 and the prediction unit 4 may be realized by different hardware.

Further, the learning system 1 may have a configuration in which two or more physically separated devices are connected by wire or wirelessly.

Further, the prediction unit 4 may be provided separately as a predictor outside the learning system 1. In this case, as illustrated in FIG. 3, the learning system 1 may include the data storage unit 2 and the learning unit 3 without including the prediction unit 4. The data storage unit 2 and the learning unit 3 shown in FIG. 3 are the same as the data storage unit 2 and the learning unit 3 shown in FIG.

These points are the same in each embodiment described later.

FIG. 4 is a flowchart showing an example of processing progress in which the learning system 1 generates a learning model. The learning unit 3 reads a set of data (see FIG. 2) stored in advance in the data storage unit 2 (step S1). Subsequently, the learning unit 3 uses the set of data as learning data to generate a learning model in which the number of reservations 9 days before the prediction target date is an explanatory variable (step S2). Note that the learning unit 3 may generate a learning model that uses not only the number of reservations 9 days before the prediction target date but also other items (for example, day of the week) as explanatory variables. As already described, the method by which the learning unit 3 generates the learning model is not particularly limited.

In addition, when the value of each item that is an explanatory variable in the learning model generated in step S2 is input, the prediction unit 4 applies the value to the learning model, so Calculate the predicted number of customers. Since the operation of the prediction unit 4 has already been described by taking as an example the case where the learning model is a prediction equation, description thereof is omitted here.

In each of the embodiments described below and modifications of each embodiment, the data stored in the data storage unit 2 and the explanatory variables used in the learning model are different from those in the first embodiment, but a learning model is generated. The progress of the process to be performed can be represented by the flowchart shown in FIG. Accordingly, in each of the embodiments described later and modifications of each embodiment, the flowchart is not shown.

According to the present embodiment, the learning unit 3 generates a learning model that uses information related to reservation (in the present embodiment, the number of reservation persons a certain period before the prediction target date) as an explanatory variable. Therefore, the learning system 1 can generate a learning model that can accurately predict the number of customers.

Next, a modification of the first embodiment will be described. Explanation of matters similar to those already described is omitted. The configuration of the learning system 1 in each modification of the first embodiment is the same as the configuration already described (see FIG. 1 or FIG. 3), and the description is omitted.

First, a first modification of the first embodiment will be described.
The data storage unit 2 may store a set of data in which the actual number of customers in the movie theater is associated with the number of reservations for each attribute of the reservations. FIG. 5 is a schematic diagram illustrating an example of a set of data stored in the data storage unit 2 according to the modification of the first embodiment. In the example shown in FIG. 5, the attributes of the reservation person are classified into “general”, “student”, and “infant”. Each day, the actual value of the number of customers is associated with the number of reservations of each attribute before a certain period of time on the actual value measurement date (9 days before in the example shown in FIG. 5). For example, the actual value on December 24 corresponds to the number of reservations corresponding to “general” 9 days before the actual value measurement date (December 15) and “student” 9 days before the actual value measurement date. The number of reservations and the number of reservations corresponding to “infant” 9 days before the actual value measurement date are associated with each other. In the example shown in FIG. 5, for ease of explanation, only three types of reservation person attributes are shown. However, more types of reservation person attributes are used in accordance with the types of tickets reserved by the reservation person. May be classified. For example, instead of “student”, more detailed attributes such as “high school student / university student” and “elementary school student / junior high school student” may be used.

Also, information other than information related to the reservation (the number of reservations of each attribute 9 days ago) may be associated with the actual value. FIG. 5 illustrates the case where the day of week information is also associated with the actual value, as in the case illustrated in FIG. 2.

In the present modification, the learning unit 3 generates a learning model using a set of data (see FIG. 5) stored in the data storage unit 2 as learning data. The learning unit 3 matches the learning data (see FIG. 2) with the number of reservations corresponding to “general” 9 days before the prediction date, the number of reservations corresponding to “student” 9 days before the prediction date, A learning model is generated in which the number of reservations corresponding to “infant” 9 days before the prediction target date is an explanatory variable. Further, when learning data including day information is used as shown in FIG. 5, the day of the week is also used as an explanatory variable in the learning model.

The prediction unit 4 includes the number of reservations corresponding to “general” 9 days before the prediction date, the number of reservations corresponding to “student” 9 days before the prediction date, and the “infant” ”And information such as the day of the week to be predicted are input. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

According to this modification, the number of predictors before a certain period before the prediction target date is not set as one explanatory variable, but is divided into different explanatory variables according to attributes. Therefore, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

Next, a second modification of the first embodiment will be described.
In the second modification of the first embodiment, the data storage unit 2 stores data similar to that in the first modification. The data storage unit 2 stores, for example, a set of data illustrated in FIG.

The learning unit 3 generates a learning model using a set of data stored in the data storage unit 2 (see FIG. 5) as learning data. However, the learning unit 3 uses the number of reservations of each attribute 9 days before the actual value measurement date to measure the actual value for the sum of the number of reservations of each attribute 9 days before the actual value measurement date. The ratio of the number of reservations of a specific attribute 9 days before the day is calculated. Here, it is assumed that the specific attribute is “general”. For example, regarding the number of reservations of each attribute associated with the actual value on December 24, the learning unit 3 sets “700” for “general” and “250” for “student”. The ratio of the “general” number of reservations “700” to the sum of the number of reservations “50” (see FIG. 5) of “infant” is calculated. That is, the learning unit 3 calculates 700 / (700 + 250 + 50) = 700/1000. The learning unit 3 performs the same calculation with respect to data of other actual value measurement dates. Then, the learning unit 3 uses the above-mentioned ratio obtained for each actual value measurement date, the day of the week of each actual value measurement date, and the actual value, each attribute before a prediction target day (9 days in this example). A learning model is generated in which the ratio of the number of reservations of a specific attribute (“general” in this example) to the sum of the number of reservations is an explanatory variable. Further, when learning data including day information is used as shown in FIG. 5, the day of the week is also used as an explanatory variable in the learning model.

The prediction unit 4 includes information such as the ratio of the number of “general” reservations 9 days before the prediction target date to the sum of the number of reservations of each attribute 9 days before the prediction target date, and the day of the week for the prediction target date. Entered. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

The ratio of the number of reservations with a specific attribute to the total number of reservations may be particularly effective in predicting the number of customers. In such a case, in the second modification, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

In the second modification, the actual value used as learning data may be the number of specific customers (for example, the number of customers who purchased the same day ticket without making a reservation). For example, it is assumed that there is a high correlation between the number of customers who purchased the same day ticket without making a reservation and the ratio of the number of reservations with a specific attribute to the total number of reservations. In such a case, the second modified example is particularly effective. In addition, the learning unit 3 may generate a learning model for each admission mode of a customer, such as a learning model for predicting the number of customers who purchased the same day ticket without making a reservation. And the prediction part 4 may calculate the predicted value of the number of customers of each entrance aspect for every learning model, and may calculate the total number of customers of a prediction object day as the sum total.

Embodiment 2. FIG.
Similar to the learning system of the first embodiment, the learning system of the second embodiment of the present invention can be represented by the block diagram shown in FIG. Hereinafter, a second embodiment will be described with reference to FIG. Explanation of matters similar to those in the first embodiment will be omitted as appropriate.

FIG. 6 is a schematic diagram illustrating an example of a set of data stored in the data storage unit 2 in the second embodiment. In the example shown in FIG. 6, the actual value of the number of customers, the number of reservations before a certain period of time on the actual value measurement date (in this embodiment, 9 days before as an example), and the actual value measurement for each day. Example of data that correlates the actual value of the number of customers on the day one year before the day, the number of reservations a certain period before (9 days before) the day one year ago, the day of the week, etc. is doing. Each day corresponds to the actual value measurement date.

Hereafter, the day one year before the actual value measurement date is referred to as the same day of the previous year as the actual value measurement date. Similarly, a day one year before the prediction target date is referred to as the same day of the previous year as the prediction target date.

The number of reservations 9 days before the actual value measurement date is the number of persons who have made a reservation to use the movie theater 9 days before the actual value measurement date.

In addition, the number of reservations 9 days before the same day of the actual value measurement date is the number of reservations made on the same day of the actual value measurement day 9 days before the same day of the actual value measurement day. Is the number of

The number of reservations 9 days before the actual value measurement date and the number of reservations 9 days prior to the same day of the actual value measurement date are the number of persons who have simply reserved to use the facility without specifying the use date. May be.

The actual number of customers, the number of reservations 9 days before the actual value measurement date, the actual value of the number of customers the same day of the previous year of the actual value measurement date, the number of reservations 9 days before the same day of the actual value measurement date, and the day of the week For example, the manager prepares the associated data every day (every actual value measurement date), and stores a set of data for each day in the data storage unit 2.

The learning unit 3 uses a set of data stored in the data storage unit 2 (see FIG. 6) as learning data, and generates a learning model for predicting the number of customers in the movie theater. In the second embodiment, the learning unit 3 calculates, for each day, the ratio of the number of reservations 9 days before the actual value measurement date to the number of reservations 9 days before the same day of the previous year on the actual value measurement date. For example, with respect to the data of the actual value measurement date “December 24” shown in FIG. 6, the number of reservations “700” 9 days before the same day of the actual value measurement date is 9 days before the actual value measurement date. The ratio of the number “1000” is calculated. That is, the learning unit 3 calculates 1000/700. The learning unit 3 performs the same calculation with respect to data of other actual value measurement dates. And the learning part 3 produces | generates a learning model using said ratio calculated | required for every actual value measurement day, the actual value of the number of customers of the previous year same day of each actual value measurement day, the day of the week of each actual value measurement day, and an actual value To do. At this time, the learning unit 3 calculates the ratio of the number of reservations before a certain period (9 days ago) to the number of reservations before a certain period (9 days ago) on the same day of the previous year as the prediction target, A learning model is generated in which the actual value of the number of customers on the same day, the day of the week to be predicted, and the like are explanatory variables.

As in the first embodiment, the learning model generation method is not particularly limited. For example, the learning unit 3 may generate a learning model by regression analysis, or may generate a learning model by another machine learning algorithm.

The prediction unit 4 includes the ratio of the number of reservations 9 days before the prediction target day to the number of reservations 9 days before the same day of the prediction target day, the actual value of the number of customers on the same day of the prediction target day, and the prediction date Information such as day of the week is input. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

According to the second embodiment, the learning unit 3 determines the ratio of the number of reservations before a certain period of the prediction target day to the number of reservations before the fixed period on the same day of the previous prediction date, A learning model using the actual number of customers as an explanatory variable is generated. Accordingly, when the situation of the actual value fluctuation is similar to the situation of the actual value fluctuation in the same period of the previous year, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved. .

Next, a modification of the second embodiment will be described.
In the present modification, the data storage unit 2 adds the number of reservations for a certain period before the actual value measurement date (in this example, 9 days in advance) for each attribute of the reservation to the actual value of the number of customers in the theater. A set of data is stored in which the actual value of the number of customers on the same day of the previous year of the actual value measurement date and the number of reservations 9 days before the same day of the previous year of the actual value measurement date are associated.

FIG. 7 is a schematic diagram illustrating an example of data stored in the data storage unit 2 according to a modification of the second embodiment. In the example shown in FIG. 7, the attributes of the reservation person are classified into “general”, “student”, and “infant”. However, as already described, the type of attribute of the reservation person is not limited to this example. The type of the attribute of the reservation may be classified into more types according to the type of ticket reserved by the reservation.

As illustrated in FIG. 7, for example, for the actual value of the number of customers on the actual value measurement date “December 24”, the reservation person corresponding to “general” 9 days before the actual value measurement date (December 15) The number “700”, the actual value “2000” of the number of customers corresponding to “general” on the same day of the previous year of the actual value measurement date, and the “general” on 9 days before the same day of the actual value measurement date (December 15 of the previous year) ”Is associated with the number of reservation users“ 500 ”.

Similarly, the actual value of the number of customers on the actual value measurement date “December 24” includes the number of reservations “250” corresponding to “students” 9 days before the actual value measurement date, The actual value “700” of the number of customers corresponding to “student” is associated with the number of reservations “160” corresponding to “student” 9 days before the same day of the actual value measurement date (see FIG. 7). .

Similarly, the actual value of the number of customers on the actual value measurement date “December 24” includes the number of reservations “50” corresponding to “infant” 9 days before the actual value measurement date, The actual value “200” of the number of customers corresponding to “Infant” is associated with the number of reservations “40” corresponding to “Infant” 9 days before the same day of the previous year on the actual value measurement date (see FIG. 7). .

FIG. 7 illustrates the case where the day of week information is also associated with the actual value, as in the case shown in FIG.

As described above, for example, an administrator prepares data in which various types of information are associated with the actual number of customers as described above, and a set of these data is stored in the data storage unit 2.

The learning unit 3 generates a learning model using a set of data stored in the data storage unit 2 (see FIG. 7) as learning data.

Here, for each day, the learning unit 3 compares the number of reservations corresponding to “general” 9 days before the actual value measurement date with respect to the number of reservations corresponding to “general” 9 days before the same day of the previous year of the actual value measurement date. Calculate the percentage of. For example, the learning unit 3 calculates 700/500 for the data of the actual value measurement date “December 24” shown in FIG.

Similarly, the learning unit 3 compares the number of reservations corresponding to “students” 9 days before the actual value measurement date with respect to the number of reservations corresponding to “students” 9 days prior to the same day of the previous year on the actual value measurement day. Calculate the percentage of. For example, the learning unit 3 calculates 250/160 for the data of the actual value measurement date “December 24” shown in FIG.

Similarly, the learning unit 3 compares the number of reservations corresponding to “infant” 9 days before the actual value measurement date with respect to the number of reservations corresponding to “infant” 9 days before the same day of the previous year on the actual value measurement day. Calculate the percentage of. For example, the learning unit 3 calculates 50/40 for the data of the actual value measurement date “December 24” shown in FIG.

The learning unit 3 performs the same calculation for each attribute with respect to data of other actual value measurement dates.

The learning unit 3 calculates, for each attribute, the ratio of the number of reservations 9 days before the actual value measurement date to the number of reservations 9 days prior to the same day the previous year on the actual value measurement date, The learning model is generated using the actual value, the day of each measurement value measurement day, the actual value, and the like. At this time, the learning unit 3 generates a learning model having the following items as explanatory variables.
(1) The ratio of the number of reservations corresponding to “general” nine days before the prediction target day to the number of reservations corresponding to “general” nine days before the same day of the prediction target date.
(2) Actual value of the number of customers corresponding to “general” on the same day of the previous year as the prediction target date.
(3) The ratio of the number of reservations corresponding to “students” nine days before the prediction target day to the number of reservations corresponding to “students” nine days before the same day of the prediction target date.
(4) The actual value of the number of customers corresponding to “students” on the same day of the previous year as the prediction target date.
(5) The ratio of the number of reservations corresponding to “infant” 9 days before the prediction target day to the number of reservations corresponding to “infant” 9 days before the same day of the prediction target day.
(6) The actual value of the number of customers corresponding to “infants” on the same day of the previous year as the prediction target date.
(7) Day of the week to be predicted.

Note that the learning unit 3 may generate a learning model in which the items (1) to (7) are used as explanatory variables, and other items are also used as explanatory variables.

The information such as (1) to (7) above is input to the prediction unit 4. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

According to this modification, the ratio of the number of reservations in a certain period before the prediction target date to the number of reservations in the fixed period on the same day of the previous year on the prediction target date, and the actual value on the same day of the previous year on the prediction target day Individual explanatory variables. Therefore, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

Embodiment 3. FIG.
The learning system of the third embodiment of the present invention can be represented by the block diagram shown in FIG. 1 or FIG. 3, similarly to the learning system of the first embodiment. Hereinafter, a third embodiment will be described with reference to FIG. Explanation of matters similar to those in the first embodiment will be omitted as appropriate.

In the third embodiment, a plurality of predetermined periods are determined in advance as a predetermined period that goes back in the past from the actual value measurement date and the prediction target date. Hereinafter, the plurality of predetermined periods are defined as first to nth predetermined periods. n is an integer of 2 or more. In addition, it is assumed that the predetermined periods from the first to the n-th are determined so that the lengths are in descending order. That is, the first predetermined period is the longest and the nth predetermined period is the shortest.

In this embodiment, in order to simplify the description, a case where three predetermined periods from the first to the third are defined will be described as an example. Further, the case where the first predetermined period is 14 days, the second predetermined period is 12 days, and the third predetermined period is 9 days will be described as an example. However, the types of the predetermined period are not limited to three types and may be two or more types. Moreover, the lengths of 14th, 12th, 9th, etc. are examples, and the length of each predetermined period may not be the illustrated length.

FIG. 8 is a schematic diagram illustrating an example of a set of data stored in the data storage unit 2 in the third embodiment. In the example shown in FIG. 8, the actual value of the number of customers, the number of reservations before the first predetermined period (14 days before) from the actual value measurement date, and the second predetermined period from the actual value measurement date for each day Exemplified data associating the previous (12 days ago) number of reservations, the number of reservations before the third predetermined period (9 days before) the actual value measurement date, the day of the week, etc. Yes. Each day corresponds to the actual value measurement date.

The number of reservations before the first predetermined period (14 days before) from the actual value measurement date is the number of persons who have made a reservation to use the movie theater on the actual value measurement date 14 days before the actual value measurement date. It is. Similarly, the number of reservations before the second predetermined period (12 days before) the actual value measurement date is reserved to use the movie theater on the actual value measurement date 12 days before the actual value measurement date. Number of people. Similarly, the number of reservations before the third predetermined period (9 days before) the actual value measurement date is reserved to use the movie theater on the actual value measurement date 9 days before the actual value measurement date. Number of people. Alternatively, the number of each reservation person may be the number of persons who have reserved the use of the facility without specifying the use date.

Data in which the actual number of customers, the number of reservations 14 days before the actual value measurement date, the number of reservations 12 days before the actual value measurement date, the number of reservations 9 days before the actual value measurement date, the day of the week, etc. For example, the administrator prepares every day (actual value measurement date), and stores a set of data for each day in the data storage unit 2.

The learning unit 3 uses the set of data stored in the data storage unit 2 (see FIG. 8) as learning data to generate a learning model for predicting the number of customers in the movie theater. The learning data 3 includes, according to the learning data, the number of reservations before the first predetermined period (14 days before) from the prediction target date, the number of reservations before the second predetermined period (12 days before) from the prediction target date, and A learning model is generated in which the number of reservations before the third predetermined period (9 days ago) from the prediction target date, the day of the week of the prediction target date, and the like are explanatory variables.

The prediction unit 4 has information such as the number of reservations 14 days before the prediction target date, the number of reservations 12 days before the prediction target date, the number of reservations 9 days before the prediction target date, and the day of the prediction target day. Entered. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

In the third embodiment, the learning unit 3 generates a learning model in which the number of reservations at a plurality of time points before the prediction target date is an explanatory variable. Therefore, more information is used as information used for predicting the number of customers on the prediction target day than in the first embodiment of the present invention. Therefore, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

In the third embodiment, the actual value of the number of customers, the number of reservations before the first predetermined period (14 days before) the actual value measurement date, and the second predetermined period from the actual value measurement date A set of data in which the number of previous reservations (12 days ago), the number of reservations before the third predetermined period (9 days before) the actual value measurement date, the day of the week, etc. of the actual measurement date It is stored in the storage unit 2. Instead of the number of reservations before each predetermined period based on the actual value measurement date, a moving average value of the number of reservations at the time before each predetermined period based on the actual value measurement date may be used. For example, for each day, the actual value of the number of customers, the moving average value of the number of reservations at the time before the first predetermined period (14 days before) from the actual value measurement date, and the second predetermined value from the actual value measurement date The moving average value of the number of reservations before the period (12 days ago), the moving average value of the number of reservations before the third predetermined period (9 days ago) from the actual value measurement date, and the actual results The manager may prepare data in which the day of the week or the like is associated with the value measurement date, and the data storage unit 2 may store the set of data.

In this case, the learning data 3 includes the moving average value of the number of reservations at the time before the first predetermined period (14 days before) the prediction target date and the second predetermined period before the prediction target date according to the learning data. The moving average value of the number of reservations as of (12 days ago), the moving average value of the number of reservations as of the third predetermined period (9 days ago) from the prediction target date, the day of the prediction target day, and the like A learning model is generated as an explanatory variable.

Even when the moving average value is used as described above, the same effect as in the third embodiment can be obtained.

Next, a modification of the third embodiment will be described.
In this modified example, the data storage unit 2 measures the actual value of the number of customers in the movie theater, the number of reservations and the actual value measurement before the first predetermined period (14 days before) the actual value measurement date for each attribute of the reservation. A set of data in which the number of reservations before the second predetermined period of the day (12 days before) and the number of reservations before the third predetermined period (9 days before) of the actual value measurement date are stored is stored.

FIG. 9 is a schematic diagram illustrating an example of data stored in the data storage unit 2 in a modification of the third embodiment. In the example shown in FIG. 9, the attributes of the reservation person are classified into “general”, “student”, and “infant”. However, as already described, the type of attribute of the reservation person is not limited to this example. The type of the attribute of the reservation may be classified into more types according to the type of ticket reserved by the reservation.

As illustrated in FIG. 9, for example, the actual value of the number of customers on the actual value measurement date “December 24” includes the number of reservations “50” corresponding to “general” 14 days before the actual value measurement date, The number of reservations “180” corresponding to “general” 12 days before the value measurement date and the number of reservations “700” corresponding to “general” 9 days before the actual value measurement date are associated with each other.

Similarly, the actual value of the number of customers on the actual value measurement date “December 24” includes the number of reservations “40” corresponding to “students” 14 days before the actual value measurement date, The number of reservations “100” corresponding to “student” is associated with the number of reservations “250” corresponding to “student” 9 days before the actual value measurement date.

Similarly, the actual value of the number of customers on the actual value measurement date “December 24” includes the number of reservations “10” corresponding to “infant” 14 days before the actual value measurement date, The number of reservation persons “20” corresponding to “Infant” is associated with the number of reservation persons “50” corresponding to “Infant” 9 days before the actual value measurement date.

FIG. 9 illustrates the case where the day of week information is also associated with the actual value as in the case shown in FIG.

As described above, for example, an administrator prepares data in which various types of information are associated with the actual value of the number of customers, and stores a set of these data in the data storage unit 2.

The learning unit 3 generates a learning model using a set of data stored in the data storage unit 2 (see FIG. 9) as learning data. The learning data 3 includes the number of reservations corresponding to “general” before the first predetermined period (14 days ago) from the prediction target date and the second predetermined period (12 days before the prediction target date) in accordance with the learning data. The number of reservations corresponding to “general” and the number of reservations corresponding to “general” before the third predetermined period (9 days) before the forecast date and “students” before the first predetermined period from the prediction date The number of reservations corresponding to, the number of reservations corresponding to “students” in the second predetermined period before the forecast date, the number of reservations corresponding to “students” in the third predetermined period before the prediction date The number of reservations corresponding to “infant” before the first predetermined period from the target date, the number of reservations corresponding to “infant” before the second predetermined period from the prediction target date, and the third from the prediction target date Scheduled to fall under “Infant” before the specified period Number of interrupted, as well as to generate a learning model for the explanatory variables the day of the week, etc. of the prediction target date.

The forecasting unit 4 sets the number of reservations corresponding to “general” before the first predetermined period (14 days before) from the prediction target date, and “general” before the second predetermined period (12 days before) from the prediction target date. Number of applicable reservations, number of reservations corresponding to “general” before the third predetermined period (9 days) before the target date of prediction, and reservations corresponding to “students” before the first predetermined period from the target date of prediction Number, number of reservations corresponding to “student” in the second predetermined period before the prediction target date, number of reservations corresponding to “student” in the third predetermined period before the prediction target date, first from the prediction target date The number of reservations corresponding to “infant” before the predetermined period of time, the number of reservations corresponding to “infant” before the second predetermined period from the prediction target date, and the number of reservation persons before the third predetermined period from the prediction target date Number of reservations that fall under the category “Infants” and forecast targets Information such as the day of the week is input. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

According to this modification, the number of reservations before each predetermined period from the prediction target date is an individual explanatory variable for each attribute. Therefore, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

Also in this modification, instead of the number of reservations before each predetermined period based on the actual value measurement date, the moving average value of the number of reservations at the time before each predetermined period based on the actual value measurement date May be used. For example, for each day, the actual value of the number of customers, the moving average value of the number of reservations corresponding to “general” at the time before the first predetermined period (14 days before) the actual value measurement date, the actual value measurement date The moving average value of the number of reservations corresponding to “general” at the time before the second predetermined period (12 days before), the “3” before the third predetermined period (9 days before) from the actual measurement date. The moving average value of the number of reservations corresponding to “general”, the moving average value of the number of reservations corresponding to “students” as of the first predetermined period before the date of measurement of the actual value, and the date of measurement of the actual value Moving average value of the number of reservations corresponding to “students” as of 2 before the predetermined period, and movement of the number of reservations corresponding to “students” as of the 3rd period before the measurement date Average value, moving average value of the number of reservations corresponding to “infant” at the time before the first predetermined period from the actual value measurement date The moving average value of the number of reservations corresponding to “Infants” at the time before the second predetermined period from the actual value measurement date, and “Infants” at the time before the third predetermined period from the actual value measurement date The administrator may prepare data in which the moving average value of the corresponding number of reservation users and the day of the week of the actual value measurement date are associated with each other, and the data storage unit 2 may store the set of data.

In this case, the learning data 3 is based on the learning data, the moving average value of the number of reservations corresponding to “general” at the time before the first predetermined period (14 days before) from the prediction target date, and the prediction target date. Moving average value of the number of reservations corresponding to “general” at the time before the second predetermined period (12 days before), “general” at the time before the third predetermined period (9 days before) from the prediction target date The moving average value of the corresponding number of reservations, the moving average value of the number of reservations corresponding to “students” at the time before the first predetermined period from the prediction target date, the time before the second predetermined period from the prediction target date The moving average value of the number of reservations corresponding to “students” in Japan, the moving average value of the number of reservations corresponding to “students” as of the third predetermined period before the prediction target date, and the first from the prediction target date The moving average value of the number of reservations corresponding to “infant” at the time before the predetermined period of, the second place from the forecast date The moving average value of the number of reservations corresponding to “Infant” at the time before the period, and the moving average value of the number of reservations corresponding to “Infant” at the time before the third predetermined period from the prediction target date, In addition, a learning model is generated in which the day of the week to be predicted is an explanatory variable.

Even when the moving average value is used as described above, the same effect as that of the modified example of the third embodiment can be obtained.

Embodiment 4 FIG.
The learning system of the third embodiment of the present invention can be represented by the block diagram shown in FIG. 1 or FIG. 3, similarly to the learning system of the first embodiment. Hereinafter, a fourth embodiment will be described with reference to FIG. Explanation of matters similar to those in the first embodiment will be omitted as appropriate.

In the fourth embodiment, as in the third embodiment, a plurality of predetermined periods are determined in advance as predetermined periods that go back in the past from the actual value measurement date and the prediction target date. As in the third embodiment, the plurality of predetermined periods are defined as the first to nth predetermined periods. n is an integer of 2 or more. In addition, it is assumed that the predetermined periods from the first to the n-th are determined so that the lengths are in descending order. That is, the first predetermined period is the longest and the nth predetermined period is the shortest.

In order to simplify the description, a case where three predetermined periods from the first to the third are defined as in the third embodiment will be described as an example. Further, the case where the first predetermined period is 14 days, the second predetermined period is 12 days, and the third predetermined period is 9 days will be described as an example. However, the types of the predetermined period are not limited to three types and may be two or more types. Moreover, the lengths of 14th, 12th, 9th, etc. are examples, and the length of each predetermined period may not be the illustrated length.

The data storage unit 2 of the fourth embodiment stores the same data set as the data storage unit 2 of the third embodiment. For example, the data storage unit 2 of the fourth embodiment stores a set of data illustrated in FIG. Since the data stored in the data storage unit 2 of the fourth embodiment is the same as the data stored in the data storage unit 2 of the third embodiment, detailed description thereof is omitted.

The learning unit 3 uses the set of data stored in the data storage unit 2 (see FIG. 8) as learning data to generate a learning model for predicting the number of customers in the movie theater.

When generating the learning model, the learning unit 3 calculates the actual value from the actual value measurement date for the number of reservations before the m−1 predetermined period from the actual value measurement date when m is an integer from 2 to n. The ratio of the number of reservations before the m-th predetermined period is calculated for each actual value measurement date. And the learning part 3 produces | generates a learning model using those calculation results, the actual value of each day, a day of the week, etc. which are contained in learning data. Specifically, the learning unit 3 sets the mth predetermined number from the prediction target date for the number of reservations before the m−1th predetermined period from the prediction target date when m is an integer from 2 to n. A learning model is generated in which the ratio of the number of reservations before the period is an explanatory variable, and the day of the week is also an explanatory variable in accordance with the learning data.

In this example, since three types of predetermined periods are defined, n = 3. Therefore, the learning unit 3 calculates the above ratios when m is set to 2 and 3, respectively. That is, the learning unit 3 sets m = 2, and the reservation person before the second predetermined period (12 days before) the actual value measurement date for the number of reservations before the first predetermined period (14 days before) the actual value measurement date. Calculate the percentage of numbers. Similarly, the learning unit 3 sets m = 3, and reserves before the third predetermined period (9 days ago) from the actual value measurement date for the number of reservations before the second predetermined period (12 days ago) from the actual value measurement date. Calculate the percentage of people. The learning unit 3 calculates these ratios for each actual value measurement date.

An explanation will be given taking the actual value measurement date “December 24” shown in FIG. 8 as an example. The learning unit 3 calculates the ratio of the number of reservations “300” 12 days before the actual value measurement date to the number of reservations “100” 14 days before the actual value measurement date (that is, 300/100). Further, the learning unit 3 calculates the ratio of the number of reservations “1000” 9 days before the actual value measurement date to the number of reservations “300” 12 days before the actual value measurement date (that is, 1000/300). The learning unit 3 performs the same calculation for other days.

The learning unit 3 generates a learning model using the ratios calculated for each day in this way and the actual values and days of the week included in the learning data. In this example, the learning unit 3 sets the number of reservation persons before the mth predetermined period from the prediction target date to the number of reservation persons before the m−1 predetermined period from the prediction target date when m = 2, 3. Is used as an explanatory variable, and a learning model is generated in which the day of the week is also an explanatory variable. More specifically, the learning unit 3 determines the ratio of the number of reservations 12 days before the prediction target date to the number of reservations 14 days before the prediction target date, and the prediction target for the number of reservations 12 days before the prediction target date. A learning model is generated in which the ratio of the number of reservations 9 days before the date and the day of the week to be predicted are explanatory variables.

The prediction unit 4 has a ratio of the number of reservations 12 days before the prediction target date to the number of reservations 14 days before the prediction target date, and 9 days before the prediction target date for the number of reservations 12 days before the prediction target date. Information such as the ratio of the number of reservations and the day of the week to be predicted is input. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

In the fourth embodiment, a learning model is generated in which the rate of change in the number of reservations in a period before the prediction target date is used as an explanatory variable. Therefore, the predicted value of the number of customers can be obtained with high accuracy using the learning model.

Next, a modification of the fourth embodiment will be described. In the modification of the fourth embodiment, the data storage unit 2 stores the same set of data as the data storage unit 2 in the modification of the third embodiment. For example, the data storage unit 2 in the modification of the fourth embodiment stores a set of data illustrated in FIG. Since the data stored in the data storage unit 2 in the modification of the fourth embodiment is the same as the data stored in the data storage unit 2 in the modification of the third embodiment, detailed description thereof is omitted.

The learning unit 3 uses a set of data stored in the data storage unit 2 (see FIG. 9) as learning data to generate a learning model for predicting the number of customers in the movie theater.

When generating the learning model, the learning unit 3 calculates the actual value from the actual value measurement date for the number of reservations before the m−1 predetermined period from the actual value measurement date when m is an integer from 2 to n. The ratio of the number of reservations before the m-th predetermined period is calculated for each attribute of the reservations. Moreover, the learning part 3 calculates those ratios for every performance value measurement day.

And the learning part 3 produces | generates a learning model using those calculation results, the actual value of each day, a day of the week, etc. which are contained in learning data. Specifically, the learning unit 3 predicts the number of reservation users for a predetermined period of the (m-1) th period before the prediction target date for each attribute of the reservation person, where m is an integer from 2 to n. A learning model is generated in which the ratio of the number of reservations before the mth predetermined period from the target date is an explanatory variable. Further, in accordance with the learning model, days of the week and the like are also used as explanatory variables in the learning model.

In this example, since three types of predetermined periods are defined, n = 3. Accordingly, the learning unit 3 calculates the above ratios when m is set to 2 and 3, respectively.

First, the learning unit 3 sets m = 2, and before the second predetermined period from the actual value measurement date for the number of reservations corresponding to “general” before the first predetermined period (14 days before) from the actual value measurement date ( The ratio of the number of reservations corresponding to “general” 12 days ago) is calculated. Similarly, the learning unit 3 reserves persons corresponding to “students” in the second predetermined period before the actual value measurement date for the number of reservation persons corresponding to “students” in the first predetermined period before the actual value measurement date. Calculate the percentage of numbers. Similarly, the learning unit 3 sets the reservation person who corresponds to the “infant” in the second predetermined period before the actual value measurement date for the number of reservation persons corresponding to the “infant” in the first predetermined period before the actual value measurement date. Calculate the percentage of numbers.

Further, the learning unit 3 sets m = 3, and before the third predetermined period from the actual value measurement date for the number of reservations corresponding to “general” before the second predetermined period (12 days before) from the actual value measurement date ( The ratio of the number of reservations corresponding to “general” in (9 days ago) is calculated. Similarly, the learning unit 3 reserves persons corresponding to “students” in the third predetermined period before the actual value measurement date for the number of reservation persons corresponding to “students” in the second predetermined period before the actual value measurement date. Calculate the percentage of numbers. Similarly, the learning unit 3 reserves a person who corresponds to the “infant” in the third predetermined period before the actual value measurement date for the number of reservation persons corresponding to the “infant” in the second predetermined period before the actual value measurement date. Calculate the percentage of numbers.

The learning unit 3 calculates these ratios for each actual value measurement date.

Referring to the data of the actual value measurement date “December 24” shown in FIG. The learning unit 3 has a ratio of the number of reservations “180” corresponding to “general” 12 days before the actual value measurement date to the number of reservations “50” corresponding to “general” 14 days before the actual value measurement date (that is, , 180/50). Similarly, the learning unit 3 sets the number of reservation persons “100” corresponding to “students” 12 days before the actual value measurement date to the number of reservations “40” corresponding to “students” 14 days before the actual value measurement date. The ratio (ie 100/40) is calculated. Similarly, the learning unit 3 sets the number of reservation persons “20” corresponding to “toddlers” 12 days before the actual value measurement date to “10” reservation persons corresponding to “infant” 14 days before the actual value measurement date. The ratio (ie 20/10) is calculated.

In addition, the learning unit 3 determines the ratio of the number of reservations “700” corresponding to “general” 9 days before the actual value measurement date to the number of reservations “180” corresponding to “general” 12 days before the actual value measurement date. (Ie 700/180) is calculated. Similarly, the learning unit 3 sets the number of reservations “250” corresponding to “students” 9 days before the actual value measurement date for the number of reservations “100” corresponding to “students” 12 days before the actual value measurement date. The ratio (ie 250/100) is calculated. Similarly, the learning unit 3 sets the number of reservation persons “50” corresponding to “Infant” 9 days before the actual value measurement date to the number “20 infants” corresponding to “Infant” 12 days before the actual value measurement date. The ratio (ie 50/20) is calculated.

The learning unit 3 performs the same calculation for other days.

The learning unit 3 generates a learning model by using the ratios calculated for each day in this way and the actual values and days of the week included in the learning data. In this example, the learning unit 3 generates a learning model having the following items as explanatory variables.
(A) Reservation corresponding to “general” in the second predetermined period (12 days ago) from the prediction target date for the number of reservation persons corresponding to “general” in the first predetermined period (14 days before) from the prediction target date Percentage of people.
(B) The ratio of the number of reservations corresponding to “students” before the second predetermined period from the prediction target date to the number of reservations corresponding to “students” before the first predetermined period from the prediction target date.
(C) The ratio of the number of reservation users corresponding to “infant” in the second predetermined period before the prediction target date to the number of reservation persons corresponding to “infant” in the first predetermined period before the prediction target date.
(D) Reservation corresponding to “general” in the third predetermined period (9 days ago) from the prediction target date for the number of reservations corresponding to “general” in the second predetermined period (12 days) before the prediction target date Percentage of people.
(E) The ratio of the number of reservations corresponding to “students” before the third predetermined period from the prediction target date to the number of reservations corresponding to “students” before the second predetermined period from the prediction target date.
(F) The ratio of the number of reservation persons corresponding to “infant” in the third predetermined period before the prediction target day to the number of reservation persons corresponding to “infant” in the second predetermined period before the prediction target day.
(G) The day of the week to be predicted.

The learning unit 3 may generate a learning model in which the items (a) to (g) are used as explanatory variables and other items are also used as explanatory variables.

The information such as (a) to (g) described above is input to the prediction unit 4. The prediction unit 4 calculates the predicted value of the number of customers on the prediction target day by applying such information to the learning model.

According to this modification, the rate of change in the number of reservations in the period before the prediction target date is an individual explanatory variable for each attribute. Therefore, the accuracy of the predicted value of the number of customers calculated using the learning model can be further improved.

In each of the above-described embodiments and modifications thereof, the case where the number of customers per day in the movie theater is the target of prediction has been described as an example, but the target of prediction is that of various facilities that can be reserved such as amusement parks and theme parks. It may be the number of customers.

In each of the above-described embodiments and modifications thereof, the case where the actual value is measured and the number of customers is predicted with “one day” as a unit has been described as an example. The length of time used as a unit when measuring the actual value or predicting the number of customers is not particularly limited, and may not be “one day”.

FIG. 10 is a schematic block diagram showing a configuration example of a computer according to each embodiment of the present invention. The computer 1000 includes a CPU 1001, a main storage device 1002, an auxiliary storage device 1003, an interface 1004, and an input device 1006.

The learning system 1 of the present invention is implemented in a computer 1000. The operation of the learning system 1 is stored in the auxiliary storage device 1003 in the form of a program. The CPU 1001 reads out the program from the auxiliary storage device 1003, develops it in the main storage device 1002, and executes the above processing according to the program.

The auxiliary storage device 1003 is an example of a tangible medium that is not temporary. Other examples of the non-temporary tangible medium include a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, and a semiconductor memory connected via the interface 1004. When this program is distributed to the computer 1000 via a communication line, the computer 1000 that has received the distribution may develop the program in the main storage device 1002 and execute the above processing.

Further, the program may be for realizing a part of the above-described processing. Furthermore, the program may be a differential program that realizes the above-described processing in combination with another program already stored in the auxiliary storage device 1003.

Next, the outline of the present invention will be described. FIG. 11 is a block diagram showing an outline of the learning system of the present invention. The learning system of the present invention includes data storage means 71 and learning model generation means 72.

The data storage means 71 (for example, the data storage unit 2) stores a set of data in which the actual value of the number of customers of the facility is associated with information related to the reservation for the facility before the time corresponding to the actual value.

The learning model generation means 72 (for example, the learning unit 3) is a learning model for predicting the number of customers of a facility at a prediction target time point, and uses a learning model having information specified from information related to reservation as an explanatory variable, Generate a set of data as learning data.

Such a configuration makes it possible to generate a learning model with high customer number prediction accuracy.

The above embodiments can be described as in the following supplementary notes, but are not limited to the following.

(Additional remark 1) The data storage means which memorize | stores the collection of the data which matched the track record value of the number of customers of a facility, and the information regarding the reservation with respect to the said facilities before the time corresponding to the said track record value, and the said time in prediction object Learning model generation means for predicting the number of customers of a facility, wherein a learning model is generated using information specified from the information related to the reservation as an explanatory variable, using the set of data as learning data; A learning system characterized by comprising.

(Supplementary note 2) The data storage means stores a set of data in which the actual value of the number of customers in the facility is associated with the number of reservations at a certain time before the time corresponding to the actual value, and a learning model is generated The learning system according to appendix 1, wherein the means uses the set of data as learning data to generate a learning model in which the number of reservations at the time before the certain period before the prediction target time is an explanatory variable.

(Supplementary Note 3) The data storage means is a set of data in which the actual value of the number of customers in the facility is associated with the number of reservations for each attribute of the reservation at a time before a certain period before the time corresponding to the actual value The learning model generation means generates a learning model using the set of data as learning data and using the number of reservations for each attribute at a time before the predetermined period from the prediction target time as an explanatory variable. The learning system according to Supplementary Note 1 or Supplementary Note 2.

(Supplementary note 4) The data storage means is a set of data in which the actual value of the number of customers of the facility is associated with the number of reservation persons for each attribute of the reservation person at a certain time before the time corresponding to the actual value The learning model generation means uses the set of data as learning data, and the number of reservations of a specific attribute with respect to the sum of the number of reservations for each attribute at the time before the predetermined period from the prediction target time The learning system according to supplementary note 1, wherein a learning model having the ratio of the above as an explanatory variable is generated.

(Supplementary note 5) The data storage means includes the actual value of the number of customers of the facility, the number of reservations at a certain period before the time corresponding to the actual value, and the time one year before the time corresponding to the actual value A set of data in which the actual value of the number of customers corresponding to the number and the number of reservations at a time before the predetermined period from the time point one year ago are stored, and the learning model generation means learns the data set Use as data, the actual value of the number of customers corresponding to the time point one year before the prediction target time point, and the prediction target time point relative to the number of reservations at a time point a certain period before the time point one year before the prediction target time point The learning system according to supplementary note 1, wherein a learning model is generated in which a ratio of the number of reservations at a point before a certain period of time is an explanatory variable.

(Supplementary note 6) The data storage means includes the actual value of the number of customers in the facility, the number of reservations at a certain period before the time corresponding to the actual value, and the time corresponding to the actual value for each attribute of the reservation A set of data in which the actual value of the number of customers corresponding to a point in time one year ago and the number of reservations at a point in time before the predetermined period from the point in time one year ago are stored, and learning model generation means Using the set of data as learning data, the actual value of the number of customers for each attribute corresponding to the time point one year before the prediction target time point, and the time point one year before the prediction target time point for each attribute The learning system according to appendix 1 or appendix 5, wherein a learning model is generated in which a ratio of the number of reservations at a time before the prediction target time to the number of reservations at a time before the fixed period is an explanatory variable.

(Supplementary note 7) The data storage means includes the actual value of the number of customers in the facility, and the reservation person at the time before each predetermined period from the first to the n-th (n is an integer of 2 or more) from the time corresponding to the actual value. A set of data associated with the number is stored, and the learning model generation unit uses the set of data as learning data, and uses the first to nth predetermined time periods before the prediction target time point as the learning data. The learning system according to appendix 1, which generates a learning model having the number of reservations as an explanatory variable.

(Additional remark 8) The data storage means is a predetermined period from the 1st to the n-th (n is an integer of 2 or more) from the time corresponding to the actual value of the actual number of customers of the facility and the reservation person. A set of data associated with the number of reservations at the previous time point is stored, and the learning model generation unit uses the set of data as learning data, and uses the first time from the prediction target time point for each attribute. The learning system according to supplementary note 1 or supplementary note 7, wherein a learning model is generated in which the number of reservations at a time before each predetermined period up to the nth is an explanatory variable.

(Supplementary note 9) The data storage means includes the actual value of the number of customers of the facility and the reservation person at the time before each predetermined period from the first to the n-th (n is an integer of 2 or more) from the time corresponding to the actual value The learning model generating means stores a set of data in which the number is associated, and the learning model generation unit corresponds to the number of reservation persons before a prediction target time point m−1 when m is an integer from 2 to n. The learning system according to appendix 1, wherein a learning model is generated in which the ratio of the number of reservations before the m-th predetermined period is an explanatory variable.

(Additional remark 10) A data storage means is each predetermined period from the 1st to nth (n is an integer greater than or equal to 2) from the time corresponding to the said actual value for every actual number of customers of a facility, and a reservation person's attribute. A set of data associated with the number of reservations at the previous time point is stored, and the learning model generation unit is configured to determine the number of predictions for each attribute when m is an integer from 2 to n. The learning system according to appendix 1 or appendix 9, wherein a learning model is generated in which a ratio of the number of reservations before the mth predetermined period to the number of reservations before the predetermined period of m−1 is an explanatory variable.

(Supplementary Note 11) The data storage means includes the actual value of the number of customers in the facility and the reservation person at the time before each predetermined period from the first time to the nth (n is an integer of 2 or more) from the time corresponding to the actual value. A set of data associated with the moving average value of the number is stored, and the learning model generating means uses the set of data as learning data, and uses each of the first to nth periods before the prediction target time point The learning system according to appendix 1, which generates a learning model having the moving average value of the number of reservations at the time of as an explanatory variable.

(Additional remark 12) The data storage means is a predetermined period from the 1st to the nth (n is an integer of 2 or more) from the time corresponding to the actual value of the number of customers in the facility and the attribute of the reservation person. A set of data associated with the moving average value of the number of reservations at the previous time point is stored, and the learning model generation unit uses the set of data as learning data, from each prediction target time point for each attribute. The learning system according to supplementary note 1 or supplementary note 11, wherein a learning model is generated in which the moving average value of the number of reservations at a time before each of the first to nth predetermined periods is an explanatory variable.

(Additional remark 13) It is applied to the learning system provided with the data storage means which memorize | stores the collection of the data which matched the actual value of the number of customers of a facility, and the information regarding the reservation with respect to the said facility before the time corresponding to the said actual value. In the learning method, a learning model for predicting the number of customers of the facility at the prediction target time point, the learning model having information specified from the information related to the reservation as an explanatory variable, the set of data as learning data A learning method characterized by generating using.

(Additional remark 14) It mounts in the computer provided with the data storage means which memorize | stores the collection of the data which matched the actual value of the customer number of facilities, and the information regarding the reservation with respect to the said facility before the time corresponding to the said actual value. A learning program, which is a learning model for predicting the number of customers of the facility at a prediction target time point, the learning model having information specified from the information related to the reservation as an explanatory variable. A learning program for executing learning model generation processing for generating a set of learning data as learning data.

Industrial applicability

The present invention is preferably applied to a learning system that generates a learning model used for predicting the number of customers of a facility that can be reserved.

DESCRIPTION OF SYMBOLS 1 Learning system 2 Data storage part 3 Learning part 4 Prediction part

Claims

Data storage means for storing a set of data in which the actual value of the number of customers of the facility is associated with information related to the reservation for the facility prior to the time corresponding to the actual value;
A learning model for predicting the number of customers of the facility at the time of prediction is generated using information specified from the information related to the reservation as an explanatory variable, using the set of data as learning data A learning system comprising learning model generation means.
The data storage means stores a set of data in which the actual value of the number of customers of the facility is associated with the number of reservations at a certain time before the time corresponding to the actual value,
2. The learning system according to claim 1, wherein the learning model generation unit generates a learning model using the set of data as learning data and having the number of reservations at a time before the predetermined period from the prediction target time as an explanatory variable. .
The data storage means stores a set of data in which the actual value of the number of customers of the facility is associated with the number of reservations for each attribute of the reservation at a certain time before the time corresponding to the actual value,
The learning model generation means uses the set of data as learning data to generate a learning model that uses the number of reservations for each attribute at a time before the predetermined period from the prediction target time as an explanatory variable. The learning system according to claim 2.
The data storage means stores a set of data in which the actual value of the number of customers of the facility is associated with the number of reservations for each attribute of the reservation at a certain time before the time corresponding to the actual value,
The learning model generating means uses the set of data as learning data to explain the ratio of the number of reservations of a specific attribute to the sum of the number of reservations for each attribute at a time before the predetermined period from the prediction target time The learning system according to claim 1, wherein a learning model is generated as a variable.
The data storage means includes the actual value of the number of customers of the facility, the number of reservations at a certain time before the time corresponding to the actual value, and the number of customers corresponding to the time one year before the time corresponding to the actual value. A set of data that associates the actual value of No. 1 with the number of reservations at a time before the certain period from the time one year ago,
The learning model generation means uses the set of data as learning data, the actual value of the number of customers corresponding to the time point one year before the prediction target time point, and the predetermined period before the time point one year before the prediction target time point. The learning system according to claim 1, wherein a learning model is generated in which a ratio of the number of reservations at a point in time before the prediction target time to the number of reservations at the point in time is an explanatory variable.
The data storage means includes the actual value of the number of customers of the facility, the number of reservations at a certain period before the time corresponding to the actual value, and one year before the time corresponding to the actual value, for each attribute of the reservation Storing a set of data that correlates the actual value of the number of customers corresponding to the time point and the number of reservations at the time point before the certain period from the time point one year ago,
The learning model generation means uses the set of data as learning data, the actual value of the number of customers for each attribute corresponding to the time point one year before the prediction target time point, and the year before the prediction target time point for each attribute. 6. A learning model is generated in which the ratio of the number of reservations at a time before the prediction target time to the number of reservations at a time before the fixed period from the time of the point is an explanatory variable. The learning system described in.
The data storage means associates the actual value of the number of customers of the facility with the number of reservations at each of the first to n-th (n is an integer of 2 or more) before each predetermined period from the time corresponding to the actual value. Remember the set of attached data,
The learning model generation unit generates a learning model using the set of data as learning data and using the number of reservations at each of the first to nth predetermined periods before the prediction target time as an explanatory variable. The learning system according to claim 1.
The data storage means is a time point before each predetermined period from the first to nth (n is an integer of 2 or more) from the time point corresponding to the actual value of the number of customers of the facility and the attribute of each reservation person. Store a set of data that correlates the number of reservations
The learning model generation means uses the set of data as learning data, and uses the number of reservations for each attribute as the explanatory variable at a point in time before each of the first to nth periods from the prediction target point in time. The learning system according to claim 1 or 7, wherein a learning model is generated.
The data storage means associates the actual value of the number of customers of the facility with the number of reservations at each of the first to n-th (n is an integer of 2 or more) before each predetermined period from the time corresponding to the actual value. Remember the set of attached data,
The learning model generation means calculates the ratio of the number of reservations before the mth predetermined period to the number of reservations before the m−1 predetermined period from the prediction target time point when m is an integer from 2 to n. The learning system according to claim 1, wherein a learning model is generated as each explanatory variable.
The data storage means is a time point before each predetermined period from the first to nth (n is an integer of 2 or more) from the time point corresponding to the actual value of the number of customers of the facility and the attribute of each reservation person. Store a set of data that correlates the number of reservations
The learning model generation means reserves before the mth predetermined period with respect to the number of reservations before the m−1 predetermined period from the prediction target time for each attribute when m is each integer from 2 to n. The learning system according to claim 1 or 9, wherein a learning model is generated in which the ratio of the number of persons is an explanatory variable.
The data storage means includes the actual value of the number of customers in the facility, and the moving average of the number of reservations at each of the first to n-th (n is an integer of 2 or more) before each predetermined period from the time corresponding to the actual value. Stores a set of data associated with values,
The learning model generation means uses the set of data as learning data, and learns using the moving average value of the number of reservations at a time before each of the first to nth predetermined periods from the prediction target time as an explanatory variable. The learning system according to claim 1, wherein a model is generated.
The data storage means is a time point before each predetermined period from the first to nth (n is an integer of 2 or more) from the time point corresponding to the actual value of the number of customers of the facility and the attribute of each reservation person. A set of data associated with a moving average value of the number of reservations of
The learning model generation means uses the set of data as learning data, and calculates a moving average value of the number of reservations for each attribute at a point in time before each predetermined period from the prediction target time point to the first to nth time points. The learning system according to claim 1 or 11, wherein a learning model is generated as an explanatory variable.
In a learning method applied to a learning system including a data storage unit that stores a set of data in which an actual value of the number of customers of a facility is associated with information related to a reservation for the facility prior to the time corresponding to the actual value ,
A learning model for predicting the number of customers of the facility at the time of prediction is generated using information specified from the information related to the reservation as an explanatory variable, using the set of data as learning data A learning method characterized by that.
A learning program installed in a computer having data storage means for storing a set of data in which the actual value of the number of customers at a facility is associated with information related to a reservation for the facility prior to the time corresponding to the actual value. And
In the computer,
A learning model for predicting the number of customers of the facility at the time of prediction is generated using information specified from the information related to the reservation as an explanatory variable, using the set of data as learning data A learning program for executing the learning model generation process.