Prediction device, prediction method, prediction program, and model generation method

Abstract

A technique for predicting the manifestation of volunteer behavior is provided. [Solution] A prediction device according to one aspect of the present invention acquires target student data related to a target student, including attribute data indicating the attributes of the target student, and provides the target student data to a trained prediction model generated by machine learning to predict the manifestation of the target student's application behavior to a target school, and outputs information related to the predicted manifestation. A model generation device according to one aspect of the present invention generates a trained prediction model by machine learning using multiple datasets.

Description

[Technical Field]

[0001] The present invention relates to a prediction device, a prediction method, a prediction program, a model generation method, and a model generation device. [Background technology]

[0002] In recent years, machine learning techniques have been used in various fields. For example, Patent Document 1 proposes a system that uses an admission probability estimation model to predict the admission probability of an examinee based on entrance examination data and examinee attribute data. This system makes it possible to predict the admission probability of each examinee according to the entrance examination. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-072919 Summary of the Invention [Problem to be solved by the invention]

[0004] The system of Patent Document 1 makes it possible to estimate the number of students to be admitted in advance by predicting the probability of admission for each examinee based on the entrance examination. However, the inventors of the present invention have discovered the following problem with this system. In other words, when recruiting students, it is important whether or not students will take action to apply, such as whether or not to participate in an open school (open campus) or whether or not to submit an application for admission. If it is possible to approach students who are likely to take action to apply when recruiting students, it is possible to efficiently increase the number of students to be admitted. However, the system of Patent Document 1 only predicts the probability of admission, and it is unclear whether or not students will take such action to apply.

[0005] In one aspect, the present invention has been made in view of the above circumstances, and one of its objectives is to provide a technique for predicting the manifestation of volunteering behavior. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, the present invention employs the following configurations, which can be combined as appropriate.

[0007] A prediction device according to one aspect of the present invention includes a control unit configured to acquire target student data related to a target student, the target student data including attribute data indicating attributes of the target student, predict a manifestation pattern of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning, and output information related to the predicted manifestation pattern.

[0008] For example, students who live near a school are more likely to participate in / apply to an open school; the attributes of the school a student belongs to (such as deviation value, availability of recommendation slots, etc.) can affect the school they apply to; students from families with high annual household incomes are more likely to apply to schools with high deviation values ​​("The annual income of parents of Tokyo University students is 10 million yen"). Over 40% of students are from households with high incomes, which explains why many students come from high-income families," [online], [Reiwa 7 Searched February 3, Internet<URL:https: / / diamond.jp / articles / - / 318710> ) may be related to the manifestation of volunteer behavior. In other words, the manifestation of volunteer behavior can be predicted from student attributes. In this configuration, by utilizing machine learning technology based on this perspective, it is possible to provide a technology for predicting the manifestation of volunteer behavior.

[0009] In the prediction device according to the above aspect, the volunteering behavior may include participation in an open school event. The manifestation of the volunteering behavior may include a degree of participation in the open school event. With this configuration, a prediction result of the degree of participation in the open school event can be obtained.

[0010] In the prediction device according to the above aspect, the attribute data may include the number of designated school recommendation slots held by the school to which the target student belongs. With this configuration, it is possible to expect improved prediction accuracy, as shown in experimental examples described below.

[0011] In the prediction device according to the above aspect, the application behavior may include submitting an application for admission. The manifestation of the application behavior may include the manner in which the application for admission is submitted. With this configuration, a prediction result of the manner in which the application for admission is submitted can be obtained.

[0012] In the prediction device according to the above aspect, the target student data may further include behavioral data related to application before applying. For example, behavior related to application before applying may be related to the manifestation of application behavior, such as students who request information materials being more likely to attend open school events, or students who attend open school events being more likely to apply / may also attend open school events of different types. Therefore, with this configuration, by including behavioral data in the target student data, which is an explanatory variable, it is possible to expect improved prediction accuracy for the manifestation of application behavior.

[0013] In the prediction device according to the above aspect, the behavioral data may include at least one of (1) the date of first contact, (2) the number of requests for materials for at least one of the route, means, and method, and (3) the number of participations in open school events for at least one of the event types and holding methods. With this configuration, improvement in prediction accuracy can be expected, as shown in the experimental examples described below.

[0014] In the prediction device according to the above aspect, the target student data may further include aspiration pattern data. Aspiration patterns may be related to the manifestation of aspiration behavior, such as whether a student is likely to attend / apply to an open school offering a desired department or faculty, whether they are likely to attend an open school offering a desired school, whether they are likely to apply directly to their desired school, or whether they are likely to apply exclusively to a desired school. Therefore, according to this configuration, by including aspiration pattern data in the target student data, which is an explanatory variable, it is possible to expect improved prediction accuracy for the manifestation of aspiration behavior.

[0015] In the prediction device according to the above aspect, predicting the manifestation pattern of application behavior may be configured by predicting the manifestation pattern of application behavior and the probability of admission of the target student to the target school. Outputting information related to the predicted manifestation pattern may be configured by outputting information related to the predicted manifestation pattern and the predicted admission probability. With this configuration, the predicted admission probability can further be obtained.

[0016] Furthermore, aspects of the present invention may not be limited to the inference (prediction) stage. Aspects of the present invention may also be directed to a machine learning stage for generating a trained predictive model used in the prediction. For example, one aspect of the present invention may be a model generation method executed by a computer for generating a trained predictive model.

[0017] For example, a model generation method according to one aspect of the present invention may be executed by a computer. The model generation method may include acquiring multiple datasets each composed of a combination of training student data on a sample student and a ground truth label indicating the true value of the expression form of the sample student's application behavior for the sample school, controlling machine learning of a predictive model using the acquired multiple datasets, and outputting the results of the machine learning. The training student data may include attribute data indicating the attributes of the sample student. The learning may be configured by training a predictive model for each of a plurality of datasets so that the results of predicting the manifestation of volunteering behavior from the training student data match the correct label. With this configuration, a trained predictive model for predicting the manifestation of volunteering behavior can be generated. Note that this model generation method is a method for producing an object (a trained predictive model).

[0018] In the model generation method according to the above aspect, the volunteering behavior may include participation in an open school event. The manifestation of the volunteering behavior may include the degree of participation in the open school event. With this configuration, a trained prediction model capable of predicting the degree of participation in the open school event can be generated.

[0019] In the model generation method according to the above aspect, the attribute data may include the number of designated school recommendation slots for the sample school held by the school to which the sample student belongs. This configuration is expected to generate a trained prediction model with high prediction accuracy.

[0020] In the model generation method according to the above aspect, the application behavior may include submitting an application for admission. The manifestation of the application behavior may include the application behavior of the application. With this configuration, a trained prediction model capable of predicting the application behavior of the application can be generated.

[0021] In the model generation method according to the above aspect, the trainee data may further include behavioral data related to volunteering before applying. As described above, behavior related to volunteering before applying may be involved in the manifestation of volunteering behavior. Therefore, with this configuration, by further using the behavioral data, it is possible to generate a trained prediction model with improved prediction accuracy for the manifestation of volunteering behavior.

[0022] In the model generation method according to the above aspect, the behavioral data may include at least one of (1) the date of first contact, (2) the number of requests for materials by at least one of route, means, and method, and (3) the number of participations in open school events by at least one of event type and event type. This configuration is expected to generate a trained prediction model with high prediction accuracy.

[0023] In the model generation method according to the above aspect, the trainee data may further include aspirational behavior data. As described above, aspirational behavior may be involved in the expression of volunteering behavior. Therefore, according to this configuration, by further using aspirational behavior data, it is possible to generate a trained prediction model with improved prediction accuracy of the expression pattern of volunteering behavior.

[0024] In the model generation method according to the above aspect, the ground truth label may be configured to further indicate the true value of the admission probability of the sample student. The machine learning may include training the predictive model for each of the multiple data sets so that the result of predicting the admission probability from the training student data matches the ground truth label. This configuration allows for the generation of a trained predictive model that has further acquired the ability to predict the admission probability.

[0025] The present invention is not limited to the above-described prediction device (information processing device). As another mode of the prediction device according to each of the above aspects, one aspect of the present invention may be a prediction method (information processing method) that realizes all or part of each of the above configurations, or may be a program (prediction program), or may be a storage medium that stores such a program and is readable by a machine such as a computer. The machine-readable storage medium may be a non-transitory medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action. The non-transitory storage medium may be a storage medium (CD, DVD, semiconductor memory, etc.), It may include an auxiliary storage device of a computer, an external storage device connected to a computer, etc. Furthermore, the form of the present invention is not limited to the above-mentioned model generation method (information processing method). As another mode of the model generation method according to each of the above aspects, one aspect of the present invention may be a model generation device (information processing device) that realizes all or part of each of the above configurations, or may be a program (model generation program), or may be a storage medium that stores such a program and is readable by a machine such as a computer.

[0026] For example, a prediction method according to an aspect of the present invention may be executed by a computer. The prediction method may include acquiring target student data about a target student, the target student data including attribute data indicating attributes of the target student, predicting a manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning, and outputting information related to the predicted manifestation.

[0027] For example, a prediction program according to an aspect of the present invention may be a program for causing a computer to execute a prediction method. The prediction method may include acquiring target student data about a target student, the target student data including attribute data indicating attributes of the target student, predicting a manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning, and outputting information related to the predicted manifestation.

[0028] For example, a model generation device according to one aspect of the present invention may include a control unit. The control unit may be configured to acquire multiple datasets, each of which is composed of a combination of training student data on a sample student and a correct answer label indicating a true value of the expression form of the volunteering behavior of the sample student for the sample school, control machine learning of a predictive model using the acquired multiple datasets, and output the results of the machine learning. The training student data may include attribute data indicating attributes of the sample student. The machine learning may be configured by training a predictive model for each of the multiple datasets so that the result of predicting the expression form of the volunteering behavior from the training student data matches the correct answer label.

[0029] For example, a model generation program according to one aspect of the present invention may be a program for causing a computer to execute a model generation method. The model generation method may include acquiring multiple datasets, each of which is composed of a combination of training student data on a sample student and a correct answer label indicating a true value of the expression form of the sample student's volunteering behavior for the sample school, controlling machine learning of a predictive model using the acquired multiple datasets, and outputting the results of the machine learning. The training student data may include attribute data indicating the attributes of the sample student. The machine learning may be performed by training a predictive model for each of the multiple datasets so that the results of predicting the expression form of the volunteering behavior from the training student data match the correct answer label. [Effects of the Invention]

[0030] According to the present invention, a technique for predicting the manifestation of volunteering behavior can be provided. [Brief explanation of the drawings]

[0031] [Figure 1] FIG. 1 shows a schematic diagram of an example of a situation in which the present invention is applied. [Figure 2] FIG. 2 is a diagram illustrating an example of input and output of a prediction model. [Figure 3]FIG. 3 is a diagram illustrating an example of a hardware configuration of the model generating device. [Figure 4] FIG. 4 is a diagram illustrating an example of a hardware configuration of the prediction device. [Figure 5] FIG. 5 is a diagram illustrating an example of the software configuration of the model generating device. [Figure 6] FIG. 6 is a diagram illustrating an example of the software configuration of the prediction device. [Figure 7] FIG. 7 is a flowchart illustrating an example of a processing procedure of the model generating device. [Figure 8] FIG. 8 is a flowchart illustrating an example of a processing procedure of the prediction device. DETAILED DESCRIPTION OF THE INVENTION

[0032] An embodiment according to one aspect of the present invention (hereinafter also referred to as "the present embodiment") will be described below with reference to the drawings. However, the present embodiment described below is merely an example of the present invention in all respects. Needless to say, various improvements and modifications can be made without departing from the scope of the present invention. In other words, when implementing the present invention, specific configurations according to the embodiment may be appropriately adopted. Note that, although data appearing in the present embodiment are described in natural language, more specifically, they are specified using computer-recognizable pseudo-language, commands, parameters, machine language, etc.

[0033] §1 Application Examples FIG. 1 schematically illustrates an example of a scenario in which the present invention is applied. The system according to this embodiment includes a model generation device 1 and a prediction device 2. The model generation device 1 is one or more computers configured to generate a trained prediction model 4 by controlling the implementation of machine learning. The prediction device 2 is one or more computers configured to predict the manifestation of application behavior of a target student TS to a target school using the trained prediction model 4.

[0034] The model generation device 1 according to this embodiment acquires multiple datasets 3. Each of the multiple datasets 3 is composed of a combination of training student data 30 related to a sample student RS and a correct answer label 35 indicating the true value of the manifestation form of the volunteering behavior of the sample student RS at the sample school. The training student data 30 includes attribute data 301 indicating the attributes of the sample student RS. The model generation device 1 controls machine learning of a prediction model 4 using the acquired multiple datasets 3. The machine learning is performed by training the prediction model 4 so that the results of predicting the manifestation form of volunteering behavior from the training student data 30 for each of the multiple datasets 3 match the correct answer label 35. By performing this machine learning, a trained prediction model 4 capable of predicting the manifestation form of the volunteering behavior of the corresponding student from student data including attribute data can be generated. The model generation device 1 outputs the results of the machine learning. The generated trained prediction model 4 may be provided to the prediction device 2 in any manner and at any timing.

[0035] Meanwhile, the prediction device 2 according to this embodiment acquires target student data 50 relating to a target student TS. The target student data 50 includes attribute data 501 indicating the attributes of the target student TS. The prediction device 2 predicts the manifestation form of the target student TS's application behavior toward the target school by providing the target student data 50 to a trained prediction model 4 generated by machine learning. The prediction device 2 outputs information relating to the results of predicting the manifestation form.

[0036] As described above, student attributes may be related to the manifestation of volunteer behavior. In other words, the manifestation of volunteer behavior can be predicted from student attributes. According to this embodiment, by utilizing machine learning technology based on this perspective, the model generation device 1 can generate a trained prediction model 4 for predicting the manifestation of volunteer behavior. The prediction device 2 can predict the manifestation of volunteer behavior of the target student TS. By utilizing this prediction result, it is expected that student recruitment will become more efficient.

[0037] [Student / School] The target students TS are students whose application behavior is predicted. The sample students RS are students from whom the training sample (dataset 3) used for machine learning is obtained. The type of student may be selected arbitrarily. In one example, the students may include anyone who wishes to enroll in any type of school. The students may include alumni.

[0038] The target school is a school for which the expression pattern of the application behavior of the target student TS is predicted. The sample school is a school for which the true value (correct answer label 35) of the expression pattern of the application behavior of the corresponding sample student RS is obtained. The type of school is not particularly limited and may be selected appropriately depending on the embodiment. In one example, the school may include any type of school for which application behavior is related to admission, such as elementary school, middle school, high school, university, vocational school, graduate school, etc.

[0039] [Student Data] The target student data 50 is student data of the target student TS, and the training student data 30 is student data of the sample student RS. As long as it contains information on students who may be involved in the manifestation of volunteer behavior, the configuration of the student data (training student data 30, target student data 50) is not particularly limited and may be determined appropriately depending on the embodiment.

[0040] FIG. 2 schematically illustrates an example of input and output of the prediction model 4 according to this embodiment. In one example, the student data may include attribute data indicating the attributes of the corresponding student. The target student data 50 may include attribute data 501 of the target student TS, and the training student data 30 may include attribute data 301 of the sample student RS. The configuration of the attributes is not particularly limited and may be selected appropriately depending on the embodiment. The attributes may include any characteristics of the student that may be related to school application.

[0041] In one example, student attributes may include address, annual household income, commuting attributes to the target school / sample school, attributes of the school (currently attending / graduated from) (school attributes), whether or not contact information is registered, age, date of birth, gender, activity history, orientation, grades, etc. Items constituting attributes may be selected as appropriate. Items of each attribute may be defined as appropriate depending on the embodiment.

[0042] For example, an address may be expressed as a single location or an area (address area). An address may include both location and area information. An address area may include attributes of households belonging to the address area (e.g., average annual household income). Commuting attributes may include commuting characteristics that may be related to application, such as commute time and commuting costs. Commuting attributes may be obtained individually or calculated from the address. School attributes may include school characteristics such as location (address / address area), deviation score, whether the school has designated school recommendation slots for the target school / sample school, the number of designated school recommendation slots, whether the school is coeducational / single-sex, and the type of governing body (private, public, etc.). In one example, the attribute data (301, 501) may include the number of designated school recommendation slots for the target school / sample school that the student (target student TS / sample student RS) belongs to has. This can be expected to improve prediction accuracy, as demonstrated in the experimental example described below. Furthermore, students who are more proactive in applying tend to register their contact information for information gathering, etc. Therefore, by including whether or not contact information is registered in the attribute data (301, 501), the proactiveness of application behavior can be taken into consideration. The activity history may include club activities (positions, tournament results, etc.), student council activities, volunteer activities, extracurricular activities, qualifications obtained, test results (English Proficiency Test, Kanji Aptitude Test, etc.), internship experience, study abroad experience, and other activities. The orientation may include at least one tendency of ability and personality, such as goal orientation, relationship orientation, growth orientation, creativity orientation, and stability orientation. The orientation may be obtained through tests such as personality assessments and aptitude tests. The aptitude may include academic aptitude, skill aptitude, personality aptitude, value aptitude, etc. Publicly known tests may be used to test the orientation. The grades may be obtained arbitrarily. For example, the grades may be obtained as the results of publicly known tests such as national mock exams.

[0043] In one example, the student data may consist of only attribute data. However, factors other than the student's attributes may also be involved in the manifestation of volunteer behavior. Therefore, in another example, in order to improve the accuracy of predicting the manifestation of volunteer behavior, the student data may further include any data other than attribute data. Any data other than attribute data may be included as long as it is involved in predicting the manifestation of volunteer behavior. The data is not particularly limited and may be selected appropriately depending on the embodiment. In one example, the student data may further include at least one of behavioral data and aspirational morphology data. That is, the target student data 50 may further include at least one of behavioral data 503 and aspirational morphology data 505 of the target student TS. The training student data 30 may further include at least one of behavioral data 303 and aspirational morphology data 305 of the sample student RS.

[0044] (Behavioral data) The behavioral data (303, 503) relates to applications before application. The configuration of the behavioral data (303, 503) is not particularly limited and may be appropriately selected depending on the embodiment, as long as it includes information indicating any behavior related to applications before application. For example, the behavioral data (303, 503) may include information such as contact time, number of requests (frequency) for materials, participation history in open school events, and other behavioral history. Contact may be involvement with the school, such as requesting materials, participating in open school events, or other behaviors (browsing a webpage, visiting the school outside of open school events, etc.). The contact time may be recorded by stage, such as the date of first contact or the date of last contact (the date of most recent contact). The contact time may be expressed in terms of elapsed time (number of days, number of months, etc.). Materials may include pamphlets, etc. Requests for materials may be made through any vendor. The number of requests for materials may be tallied in one go or by vendor. The number of requests for materials may be tallied by at least one of route, means, and method. The route may be defined, for example, as a request source for materials, such as directly to the school, via a vendor, or other route. The means may be defined, for example, as a method for obtaining materials, such as using media (information magazines, the web, telephone, etc.), attending a guidance session, or other means. The means may be defined for each route. The method may be defined, for example, as a specific method for obtaining materials, such as the web, telephone, information magazines, the type of guidance session (on-site information session, on-campus information session, etc.), or other methods. The method may be defined for each means. An open school may be an event held to open school buildings to students considering enrolling and to help them deepen their understanding of the school. An open school may include an open campus. The participation history may include any information related to participation in an open school. In one example, the participation history for an open school may include the number of times the student attended the open school, the results of activities within the open school, the results of responses to a questionnaire at the open school, whether or not a chaperone was present when attending, the relationship with the chaperone (parent, friend, etc.), etc. The open school may include a variety of event types, such as mock lectures, practical training, school tours, individual consultations, etc. The open school may be held in a variety of ways, such as on-site or online.The number of open school attendances may be aggregated all at once or by event type and / or event format. Activity results may include whether or not an interview was conducted, whether or not an event was attended at a specific venue, etc. In one example, the behavioral data (303, 503) may include (1) the date of first contact, (2) the number of requests for materials by route, means, and / or format, and (3) the number of open school attendances by event type and / or event format. The date of contact may be expressed as elapsed time (days, months, etc., any unit is acceptable). This is expected to improve prediction accuracy, as shown in the experimental example described below.

[0045] As described above, behavior related to volunteering before applying can contribute to the manifestation of volunteering behavior. Therefore, according to one example of this embodiment, in the prediction device 2, the target student data 50 further includes behavioral data 503, which is expected to improve the prediction accuracy of the manifestation of volunteering behavior. Also, according to one example of this embodiment, in the model generation device 1, the trainee student data 30 further includes behavioral data 303, which is expected to generate a trained prediction model 4 with improved prediction accuracy of the manifestation of volunteering behavior.

[0046] (Application form data) The application form data (305, 505) relates to the application form of the students (sample student RS, target student TS). If any information about the application form is included, the application form data (30 The configuration of the application form data (305, 505) is not particularly limited and may be selected appropriately depending on the embodiment. In one example, the application form data (305, 505) may include information such as application form, application content, and application status. The application form may include whether the application is for a single school or multiple schools. The application content may include the desired department, faculty, major, course, etc. The application status may include the number of times the applicant has applied, the time period when the applicant has applied, etc. The number of times the applicant has applied may be the number of times the applicant has expressed their desire to enroll in the target school through at least one of a request for information and an open school event. The time period when the applicant has expressed their desire to enroll in the target school through at least one of a request for information and an open school event.

[0047] In one example, the aspiration form data (305, 505) may include information about schools other than the target school in addition to the above information about the target school (target school, sample school). For example, when multiple schools are available for application, such as first choice, second choice, etc., the schools other than the target school may include schools to which the student is also applying. As a specific example, when the target school is the student's first choice, the aspiration form data (305, 505) may include information about the second choice and subsequent schools (application form, aspiration content, aspiration status, etc.). In another example, the aspiration form data (305, 505) may be configured to indicate the student's current aspiration form. The aspiration form data (305, 505) may also be configured to include a history of past aspiration forms.

[0048] As described above, aspirational behavior may be involved in the manifestation of volunteering behavior. Therefore, according to one example of this embodiment, in the prediction device 2, by further including aspirational behavior data 505 in the target student data 50, it is expected that the prediction accuracy of the manifestation of volunteering behavior will be improved. Also, according to one example of this embodiment, by further including aspirational behavior data 305 in the trainee student data 30 in the model generation device 1, it is expected that a trained prediction model 4 with improved prediction accuracy of the manifestation of volunteering behavior will be generated.

[0049] (others) Note that data other than the attribute data included in the student data need not be limited to the behavioral data and aspirational data, and may be selected appropriately depending on the embodiment. In one example, the student data may further include approach data from the school to the student, along with at least one of the behavioral data and aspirational data, or instead of the behavioral data and aspirational data. The target student data 50 may further include approach data from the target school to the target student TS. The training student data 30 may further include approach data from the sample school to the sample student RS.

[0050] The approach data relates to direct or indirect approaches by a target school (target school, sample school) to a target student (target student TS, sample student RS). The structure of the approach data is not particularly limited and may be appropriately selected depending on the embodiment, as long as it includes any information related to the approach. For example, the approach may be direct, such as providing materials (e.g., pamphlets), sending mail (e.g., postcards), sending emails, or making phone calls, or indirect, such as approaching the student's current school. For example, the approach data may include the number and timing of providing materials, the number and timing of sending mail, the number and timing of sending emails, the number and timing of phone calls, and the number and timing of approaches to the student's current school. Approaching the student's current school may include, for example, holding an optional event (information session, etc.) by the target school at a school one level below the target school, such as a university holding an entrance information session at a high school.

[0051] Approaches from schools, such as applying to enroll after seeing a pamphlet or applying to enroll in a target school at an entrance information session held at the school where the student is currently enrolled, can contribute to the occurrence of application behavior. Therefore, according to one example of this embodiment, by further including approach data in the target student data 50 in the prediction device 2, it is expected that the prediction accuracy of the manifestation of application behavior will be improved. Furthermore, according to one example of the present embodiment, in the model generation device 1, the trainee data 30 further includes approach data, so that it is possible to expect the generation of a trained prediction model 4 with improved prediction accuracy for the manifestation of volunteering behavior.

[0052] [Manifestation of volunteer behavior] The application behavior may be any behavior related to applying to a target school (target school, sample school). In one example, the application behavior may include participating in an open school event, applying, etc. The manifestation form may be any way in which the application behavior is manifested. The manifestation form may include, for example, whether or not to take action (whether or not to attend / visit school, whether or not to apply), a probabilistic element of the behavior (probability of attending / visiting school, probability of applying), etc.

[0053] As shown in FIG. 2 , in one example, application behavior may include at least one of participating in an open school event and submitting an application for admission. If application behavior includes participating in an open school event, the manifestation of application behavior may include the degree of participation in the open school event. The degree of participation in the open school event may include whether or not to participate in the open school event, the probability of whether or not to participate, the number of times (frequency) to participate in the open school event, etc. If the open school event includes multiple event types, the degree of participation in the open school event may be predicted collectively or for each event type. Furthermore, if application behavior includes submitting an application for admission, the manifestation of application behavior may include the application mode of the application for admission. The application mode of the application for admission may include whether or not to apply, the probability of whether or not to apply, which application format to select, the probability of selecting the application format (single application, multiple application, etc.), which department / faculty / major / course to select, the probability of selecting the department / faculty / major / course, etc. Whether or not to apply and the probability of doing so may be predicted all at once, for each application type, or for each department / faculty / major / course. Note that the probabilistic element of behavior (such as the probability of whether or not to attend the open school) may be used as is, or may be converted into a binary judgment result (negative, positive). If the latter is adopted, the manifestation of application behavior may include the result of a binary judgment on the probabilistic element of behavior. The threshold used for the binary judgment may be specified appropriately depending on the embodiment. For example, the threshold may be specified based on a known criterion such as a recall criterion.

[0054] According to one example of this embodiment, the manifestation of volunteer behavior includes the degree of participation in open school events, allowing the prediction device 2 to obtain a predicted result of the degree of participation in open school events for the target student TS. The model generation device 1 can generate a trained prediction model 4 that has the ability to predict the degree of participation in open school events. In particular, when predicting the degree of participation in open school events, in one example, the attribute data (301, 501) may include the number of designated school recommendation slots for the target school / sample school held by the student's (target student TS / sample student RS) affiliated school. This, as shown in the experimental example described below, can be expected to improve the accuracy of predicting the degree of participation in open school events in the prediction device 2. The model generation device 1 can be expected to generate a trained prediction model 4 that accurately predicts the degree of participation in open school events. In another example, the behavioral data (303, 503) may include (1) the date of first contact, (2) the number of requests for materials for at least one of the route, means, and method, and (3) the number of participations in open school events for at least one of the event type and event type. As a result, as shown in the experimental example described below, it is expected that the prediction device 2 will be able to predict with improved accuracy the degree of participation in open school events. It is also expected that the model generation device 1 will be able to generate a trained prediction model 4 that has high accuracy in predicting the degree of participation in open school events.

[0055] According to an example of this embodiment, the manifestation form of the application behavior includes the application form of the admission application, so that the prediction device 2 predicts the application form of the admission application for the target student TS. The model generation device 1 can generate a trained prediction model 4 that has the ability to predict the application behavior of an application. Note that predicting the manifestation of application behavior may include inferring potential interest in application at the time of prediction (current time) (i.e., predicting whether it will lead to application behavior in the future).

[0056] [Data acquisition method] The model generating device 1 may acquire each dataset 3 by any method. Each dataset 3 may be collected as needed. Each dataset 3 may be generated by one or more arbitrary computers. The one or more arbitrary computers involved in generating the dataset 3 may or may not include the model generating device 1. The computer that stores the dataset 3 may be the same as or different from the computer that generates the dataset 3. At least a portion of each dataset 3 may be generated on the model generating device 1. The model generating device 1 may acquire at least a portion of each dataset 3 from an external computer. Acquiring the multiple datasets 3 may include at least one of generating at least a portion of the multiple datasets 3 and acquiring at least a portion of the multiple datasets 3 from an external computer. The external computer may include an external storage device such as a NAS (Network Attached Storage).

[0057] The information sources of the trainee data 30 and the correct answer labels 35 may be selected appropriately depending on the embodiment. In one example, the values ​​of each item of the trainee data 30 may be obtained appropriately from an inquiry (such as an application to participate in an open school) made by a student (sample student RS) to a school (sample school), a student's response to a request for information, or a student's response to a questionnaire. The corresponding correct answer labels 35 may be obtained from the student's (sample student RS) performance at the school (sample school). In one example, at least a portion of the values ​​of the trainee data 30 and the correct answer labels 35 may be manually entered by a student, a vendor employee, a school official, or the like. At least a portion of the values ​​of the trainee data 30 and the correct answer labels 35 may be obtained by computer processing, such as image analysis of answers written on paper or calculation of cumulative values ​​(such as number of times). In one example, the trainee data 30 and the correct answer labels 35 may be obtained collectively or by aggregating (integrating) partial data generated separately from multiple information sources.

[0058] Similarly, the prediction device 2 may acquire the target student data 50 in any manner. The target student data 50 may be generated by one or more arbitrary computers. The one or more arbitrary computers involved in generating the target student data 50 may or may not include the prediction device 2. The computer that stores the target student data 50 may be the same as or different from the computer that generates the target student data 50. At least a portion of the target student data 50 may be generated on the prediction device 2. The prediction device 2 may acquire at least a portion of the target student data 50 from an external computer. Acquiring the target student data 50 may include at least one of generating at least a portion of the target student data 50 and acquiring at least a portion of the target student data 50 from the external computer. The external computer may include an external storage device.

[0059] The information source of the target student data 50 may be selected appropriately depending on the embodiment. In one example, the values ​​of each item of the target student data 50 may be obtained appropriately from an inquiry by a student (target student TS) to a school (target school), a student's response to a request for information, a student's response to a questionnaire, etc. In one example, at least a portion of each value of the target student data 50 may be manually entered by a student, a vendor employee, a school official, etc. At least a portion of each value of the target student data 50 may be obtained by computer processing such as image analysis of responses written on paper or calculation of a cumulative value (number of times, etc.). In one example, the target student data 50 may be obtained in one batch, or by aggregating (integrating) partial data generated separately from multiple information sources.

[0060] [Prediction model] The prediction model 4 is configured to predict the manifestation of volunteer behavior from a given input (such as student data). The trained prediction model 4 has acquired the ability to predict the manifestation of volunteer behavior from a given input through machine learning. As long as such a prediction (inference) process can be performed, the configuration of the prediction model 4 is not particularly limited and may be selected appropriately depending on the embodiment. The prediction model 4 may be configured using any type of machine learning model.

[0061] The machine learning model is configured to have one or more calculation parameters that can be adjusted by machine learning. The one or more calculation parameters are used to calculate the desired inference (prediction of the manifestation form of volunteer behavior). The machine learning model may be configured, for example, with a neural network, a regression model, a decision tree model, a support vector machine, or other functional formulas (calculation models). The machine learning method may be selected appropriately depending on the machine learning model to be adopted. For example, known optimization methods such as backpropagation, regression analysis (multiple regression analysis, etc.), and random forest may be adopted as the machine learning method.

[0062] In one example, the prediction model 4 may include a neural network. The structure of the neural network is not particularly limited and may be determined appropriately depending on the embodiment. The structure of the neural network may be specified, for example, by the number of layers from the input layer to the output layer, the type of each layer, the number of nodes (neurons) included in each layer, the connection relationships between the nodes in each layer, etc. In one example, the neural network may include any mechanism such as a recursive structure or an attention mechanism. Furthermore, the neural network may include any layer such as a fully connected layer, a convolutional layer, a pooling layer, a deconvolutional layer, an unpooling layer, a normalization layer, a dropout layer, or an LSTM (Long Short-Term Memory). The neural network may include a diffusion model. The model may include any type of model, such as a neural network model, a transformer model, or a generative model. A large-scale language model (LLM) is a large-scale network. It may include large-scale models such as the Large Vision-Language Model (LVLM). When a large-scale model is included, each dataset 3 may be used for at least one of fine tuning and in-context learning (provided as a sample set of the correct answer). The weights of the connections between the nodes included in the neural network and the thresholds of the nodes are examples of calculation parameters. The data format of the input and output of the prediction model 4 is not particularly limited and may be selected appropriately depending on the embodiment.

[0063] Machine learning involves adjusting (optimizing) values ​​of computational parameters using training samples (datasets 3). In one example, the model generation device 1 may perform supervised learning using multiple datasets 3 as a machine learning process. Training student data 30 are input samples (training samples), and correct labels 35 are output samples (teacher signals). In supervised learning, the values ​​of computational parameters of the prediction model 4 may be adjusted (optimized) so that the output obtained from the prediction model 4 by providing input samples (training student data 30, etc.) matches the corresponding output samples (correct labels 35). However, the machine learning method is not limited to this example and may be modified as appropriate depending on the embodiment.

[0064] In one example, a trained prediction model 4 may be generated for each school. That is, machine learning of the prediction model 4 may be performed using only the dataset 3 of the school for which the trained prediction model 4 is to be generated. This makes it possible to obtain a trained prediction model 4 specialized for predictions for the target school. However, the method for generating the trained prediction model 4 is not limited to this example. In another example, the multiple datasets 3 may include datasets derived from multiple schools. This allows the model generation device 1 to generate a generalized trained prediction model 4.

[0065] As an option when the multiple datasets 3 include datasets from multiple schools, in one example, the prediction model 4 may be a conditional model that further accepts input of school attribute data. As long as the school conditions can be identified, the configuration of the school attribute data is not particularly limited and may be selected appropriately depending on the embodiment. In one example, the attribute data may include the school's location, deviation score, number of years since opening (foundation year), departments / faculties / majors / courses offered, whether the school has a designated school recommendation quota, whether it is coeducational or single-sex, the type of governing body, a list of club activities, characteristics of the student council, and other school characteristics. In the model generation device 1, each dataset 3 may further include attribute data of a sample school. In machine learning, the attribute data of the sample school may be used as an input sample (training sample) along with the trainee data 30. The attribute data of the sample school may be acquired by any method, similar to the trainee data 30. The prediction device 2 may further acquire attribute data of a target school and provide the target student data 50 and the attribute data of the target school to the trained prediction model 4. The attribute data of the target school may be obtained in any manner, similar to the target student data 50.

[0066] Furthermore, the input / output configuration of the prediction model 4 may be changed as appropriate depending on the embodiment. Regarding the input / output configuration of the prediction model 4, components can be added as appropriate depending on the embodiment. The input of the prediction model 4 may consist of only student data, or may further include any data other than student data (such as the above-mentioned school attribute data). In one example, the prediction model 4 may be configured to further accept input of data other than student data. Furthermore, the output of the prediction model 4 may consist of only a prediction result of the manifestation form of volunteering behavior, or may further include any inference result other than the manifestation form of volunteering behavior. In one example, the prediction model 4 may be configured to further output an inference result other than the manifestation form of volunteering behavior.

[0067] As shown in FIG. 2 , in one example, the prediction model 4 may further be configured to output a predicted result of the student's admission probability. That is, in the prediction device 2, predicting the expression pattern of application behavior may include predicting the expression pattern of application behavior and the admission probability of the target student TS to the target school. Outputting information regarding the predicted result of the expression pattern may include outputting information regarding the predicted result of the expression pattern and the predicted admission probability. When predicting the admission probability for each of multiple target students TS, predicting the admission probability may include predicting the number of students admitted. Furthermore, in the model generation device 1, the correct answer label 35 may be configured to further indicate the true value of the admission probability of the sample student RS. The true value of the admission probability may be obtained by any method. In one example, the true value of the admission probability may be obtained from the admission record of the sample student RS. The machine learning may include training the prediction model 4 for each of multiple datasets 3 so that the predicted result of the admission probability from the training student data 30 matches the correct answer label 35.

[0068] For example, a student's attributes, behavior, and aspirations may affect the school they attend. For example, the attributes of the school they attend (e.g., deviation score, availability of recommendation slots, etc.) may affect the school they attend. Students are more likely to select a school from among schools that participate in open school events. Students are more likely to select a school that offers a department / faculty / major / course that matches their desired department / faculty / major / course. Therefore, a student's admission probability can be predicted from student data. According to one example of this embodiment, the prediction device 2 can further obtain a prediction result of the admission probability for the target student TS. Furthermore, according to one example of this embodiment, the model generation device 1 can generate a trained prediction model 4 that further acquires the ability to predict admission probability.

[0069] The input / output format of the prediction model 4 is not particularly limited and may be determined appropriately depending on the embodiment. In one example, the input data (student data, etc.) may be provided to the prediction model 4 as is, or may be provided after any preprocessing has been applied. In another example, the output of the prediction model 4 is configured to directly or indirectly indicate the manifestation form of volunteer behavior. When the output of the prediction model 4 is configured to indirectly indicate the manifestation form of volunteer behavior, the manifestation form of volunteer behavior may be obtained by performing any information processing (interpretation processing, etc.) on the output of the prediction model 4.

[0070] [Information about predicted results] In a typical example, outputting information regarding the result of the predicted expression morphology may be configured by outputting the result of the predicted expression morphology as is. However, as long as the information is related to the predicted expression morphology, the content of the information to be output is not limited to this example and may be determined appropriately depending on the embodiment. In another example, the prediction device 2 may perform any information processing when predicting the expression morphology or on the result of the predicted expression morphology. Outputting information regarding the result of the predicted expression morphology may be configured by outputting the result of this information processing together with or instead of the predicted result.

[0071] Information processing of the prediction results may be selected appropriately depending on the embodiment. In one example, the prediction device 2 may process the prediction results in accordance with a user interface definition and output the processed prediction results. The user interface definition may be provided appropriately depending on the embodiment. For example, the prediction device 2 may predict the manifestation of application behavior for each of multiple target students TS. The prediction device 2 may output the prediction results for each target student TS in a list. The output may include displaying the results on a display. In this case, the prediction device 2 may display each target student TS in a format according to the rank of the probabilistic element. For example, the prediction device 2 may extract students who are likely to participate in or apply to an open school from multiple students. Whether the probability is high or not may be evaluated using any method, such as threshold judgment. The threshold may be specified arbitrarily. The prediction device 2 may display the extracted students in a color different from other students.

[0072] In addition, in one example, the prediction device 2 may calculate the influence of each item of input data (such as the target student data 50) on the prediction result. The influence may be calculated by any method. For example, the influence may be calculated by using the inference of a machine learning model such as SHAP (SHapley Additive exPlanations). The prediction device 2 may output the calculated influence of each item of the input data together with the predicted expression pattern.

[0073] [Utilizing predicted results] The output information on the prediction results may be used for any purpose related to student recruitment. In one example, the information on the prediction results may be used by a student recruitment officer at a target school to select students to approach. That is, by using the information on the prediction results, the officer may identify, from among multiple students, students who are likely to take application action, such as attending an open school or applying. By actively approaching the identified students, it is expected that student recruitment will become more efficient.

[0074] Furthermore, when predicting the expression phenotypes of each of multiple target students TS, the target student data 50 for each target student TS may be obtained through a vendor. The target student data 50 may be obtained from multiple vendors. In this case, the prediction device 2 may calculate, for each vendor, the number of students likely to apply for a school, as indicated by the prediction results. This may allow the extent of recruitment of students likely to apply to be evaluated for each vendor. The probability may be evaluated using any method, such as threshold determination. The threshold may be defined arbitrarily. The evaluation results for each vendor may be used by the student recruitment officer to select vendors to focus on for recruitment in the next term and beyond. For example, by strengthening student recruitment requests to vendors that have a large number of students likely to apply, it is possible to expect more efficient use of student recruitment costs.

[0075] In addition, when predicting the expression form of each of multiple target student TSs, the predicted expression form of each target student TS is The prediction results may be output for each school. Information about the prediction results may be used by a student recruitment officer to select schools to actively approach. In other words, by utilizing the information about the prediction results, the officer may identify, from among multiple schools, schools with many students who are likely to apply or have a high probability of enrollment. By actively approaching the identified schools, it is expected that student recruitment will become more efficient.

[0076] Furthermore, when calculating the influence of each item of input data on the prediction result, the calculation result of this influence may also be utilized for various purposes related to student recruitment. In one example, the calculation result of the influence may be utilized as information for determining how to approach students. For example, if a calculation result indicates that participation in an open school event has a significant impact on the application probability, a student recruitment officer may proactively approach the target student TS for whom this calculation result was obtained to encourage them to participate in the open school event. This is expected to improve the efficiency of student recruitment. In one example, the calculation result of the influence may be utilized as information for interviews with the target student TS. For example, if club activities have a significant impact on the application probability, by conducting the interview while focusing on items with a high influence, such as by discussing club activities, it is expected to improve the likelihood of the student applying. In another example, the calculation result of the influence may be utilized as information for determining how to approach the school to which the student belongs. For example, by holding an open school information session at a school where many students belong, for which participation in an open school event has a significant impact on the application probability, it is expected to improve the efficiency of student recruitment.

[0077] [Predicting] The trained prediction model 4 may be deployed on any computer. The trained prediction model 4 may be deployed on the prediction device 2, or on an external computer other than the prediction device 2. When the trained prediction model 4 is deployed on the prediction device 2, the calculations of the trained prediction model 4 may be executed on the prediction device 2. Predicting the manifestation of volunteering behavior by the prediction device 2 may be configured by executing the calculations of the trained prediction model 4 to obtain a predicted result of the manifestation of volunteering behavior. On the other hand, when the trained prediction model 4 is deployed on an external computer, the calculations of the trained prediction model 4 may be executed on the external computer. Predicting the manifestation of volunteering behavior by the prediction device 2 may be configured by issuing a request for a prediction calculation including input data (such as target student data 50) to the external computer and obtaining a prediction result by the trained prediction model 4 from the external computer. The external computer may be configured by one or more arbitrary computers. For example, the external computer may be a server device. The external computer may be connected to the prediction device 2 via a network or directly connected to the prediction device 2. The type of network may be selected arbitrarily.

[0078] [Controlling Machine Learning] The machine learning arithmetic processing of the predictive model 4 may be executed by any computer. The machine learning arithmetic processing may be executed on the model generation device 1, or may be executed by an external computer other than the model generation device 1. Controlling machine learning may include at least one of executing machine learning within the model generation device 1 and issuing instructions to an external computer to execute machine learning. That is, in one example, the model generation device 1 may execute machine learning arithmetic processing as a machine learning control process. In another example, the model generation device 1 may issue an instruction to execute machine learning of the predictive model 4 to an external computer. The external computer may acquire multiple datasets 3 to be used for machine learning by any method. The multiple datasets 3 may be provided to the external computer from the model generation device 1, or may be provided by other methods such as using an external storage device. The external computer may execute machine learning arithmetic processing of the predictive model 4 in response to instructions from the model generation device 1. The model generation device 1 may acquire machine learning results from the external computer. In yet another example, the model generation device 1 executes a part of the machine learning arithmetic processing of the predictive model 4. The external computer may execute the remaining calculation processing in response to instructions from the model generation device 1. The external computer may be configured with one or more arbitrary computers. For example, the external computer may be a server device. The external computer may be connected to the model generation device 1 via a network, or may be directly connected to the model generation device 1. The type of network may be selected arbitrarily.

[0079] [Machine learning result output] Outputting the machine learning result may include outputting any information related to the machine learning result. In one example, outputting the machine learning result may include generating learning result data indicating the machine learning result and storing the generated learning result data in any storage area. In another example, outputting the machine learning result may include outputting a convergence result of training in machine learning. The convergence result may include, for example, a loss, a number of time steps, a learning curve, etc.

[0080] [System Configuration] In one example, as shown in Fig. 1, the model generation device 1 and the prediction device 2 may be connected to each other via a network. The type of network may be appropriately selected from, for example, the Internet, a wireless communication network, a mobile communication network, a telephone network, a dedicated network, etc. However, the method of exchanging data between the model generation device 1 and the prediction device 2 is not limited to this example and may be appropriately selected depending on the embodiment. In another example, data may be exchanged using a storage medium.

[0081] In addition, in the example of FIG. 1, the model generation device 1 and the prediction device 2 are each separate computers. However, the configuration of the system according to this embodiment is not limited to this example and may be determined appropriately depending on the embodiment. In another example, the model generation device 1 and the prediction device 2 may be configured as a single computer. In yet another example, at least one of the model generation device 1 and the prediction device 2 may be configured as multiple computers. In one example, the prediction device 2 may predict the manifestation form of the volunteer behavior of the target student TS using a trained prediction model 4 generated by a device other than the model generation device 1.

[0082] §2 Configuration example [Hardware configuration] (Model generation device) 3 is a schematic diagram illustrating an example of the hardware configuration of the model generation device 1 according to this embodiment. In one example, the model generation device 1 may be configured as a computer to which a control unit 11, a storage unit 12, a communication module 13, an input device 14, and an output device 15 are electrically connected.

[0083] The control unit 11 is configured to execute information processing based on programs and various data. For example, the control unit 11 includes a hardware processor such as a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory). The control unit 11 (CPU) is an example of a processor resource.

[0084] The storage unit 12 is configured to hold any data. For example, the storage unit 12 may include a hard disk drive, a solid state drive, a semiconductor memory, etc. The storage unit 12, RAM, and ROM are examples of memory resources. In one example of this embodiment, the storage unit 12 may store various types of information such as a model generation program 81, multiple datasets 3, and learning result data 125.

[0085] The model generation program 81 is a program for causing the model generation device 1 to execute information processing (see FIG. 7, which will be described later) related to machine learning of the prediction model 4. includes a series of instructions for the information processing. The learning result data 125 is configured to indicate information about the generated trained prediction model 4. In this embodiment, the learning result data 125 may be generated as a result of executing the model generation program 81. Note that the configuration of the learning result data 125 is not particularly limited as long as it can hold information for executing the calculation processing of the trained prediction model 4, and may be determined appropriately depending on the embodiment. In one example, the learning result data 125 may be configured to include information indicating the values ​​of calculation parameters adjusted by machine learning. The learning result data 125 may be configured to include information indicating the configuration of the prediction model 4 (e.g., the structure of a neural network, etc.).

[0086] In one example, at least one of the model generation program 81, the plurality of datasets 3, and the learning result data 125 may be stored in a storage medium 91 instead of or together with the storage unit 12. The storage medium 91 is configured to store various information (such as stored programs) by electrical, magnetic, optical, mechanical, or chemical action so that a machine such as a computer can read the information. The storage unit 12 and the storage medium 91 are examples of non-transitory storage media. The model generation device 1 may acquire at least one of the model generation program 81 and the plurality of datasets 3 from the storage medium 91. The storage medium 91 may be a disk-type storage medium (such as a CD or a DVD) or a non-disk-type storage medium such as a semiconductor memory (such as a flash memory). In one example, any drive device may be used to read information stored in the storage medium 91. The type of drive device may be selected depending on the storage medium 91. The drive device may be connected to the model generation device 1 in any manner. The storage medium 91 may include an external storage device.

[0087] The communication module 13 is configured to perform wired or wireless communication via a network. The communication module 13 may be configured, for example, by a wired LAN (Local Area Network) module, a wireless LAN module, or the like. The network standard is not particularly limited and may be selected appropriately depending on the embodiment. The model generation device 1 may perform data communication with other computers (such as the prediction device 2) via the communication module 13.

[0088] The input device 14 is configured to accept input of information. The input device 14 may be configured, for example, by a mouse, a keyboard, a touch panel, an operator, etc. The output device 15 is configured to output information. The output device 15 may be configured, for example, by a display, a speaker, etc. The model generation device 1 may be operated using the input device 14 and the output device 15. The input device 14 and the output device 15 may be directly connected to the model generation device 1, or may be indirectly connected via at least one of the communication module 13 and an external interface. The external interface may be appropriately configured to connect to an external device via a wired or wireless connection, for example, a USB (Universal Serial Bus) port, a dedicated port, etc. The input device 14 and the output device 15 may be at least partially integrated into a touch panel display, etc.

[0089] Regarding the specific hardware configuration of the model generating device 1, components can be omitted, replaced, or added as appropriate depending on the embodiment. For example, the control unit 11 may include multiple hardware processors. The hardware processors may be a microprocessor, a field-programmable gate array (FPGA), a digital signal processor (DSP), a GPU (Gateway Processor), a 3D processor, a 3D image ... The communication module 13, the input device 14, and the output device 15 may be configured by a PU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), etc. At least one of the above may be omitted. The model generation device 1 may be configured with a plurality of computers. In this case, the hardware configurations of the computers may or may not be the same. Furthermore, the model generation device 1 may be configured with an information processing device designed specifically for the service to be provided, a general-purpose server device, a general-purpose PC (Personal Computer), etc. The terminal device may be a smartphone, a tablet, or the like. The user terminals may include terminals such as:

[0090] (Prediction device) 4 schematically illustrates an example of a hardware configuration of the prediction device 2 according to this embodiment. In one example, the prediction device 2 may be configured as a computer to which a control unit 21, a storage unit 22, a communication module 23, an input device 24, and an output device 25 are electrically connected.

[0091] The control unit 21 to the output device 25 and the storage medium 92 of the prediction device 2 may be configured similarly to the control unit 11 to the output device 15 and the storage medium 91 of the model generation device 1. The control unit 21 (CPU) is an example of a processor resource of the prediction device 2, and the storage unit 22, RAM, and ROM are examples of memory resources of the prediction device 2. In one example of this embodiment, the storage unit 22 may store various information such as a prediction program 82 and learning result data 125.

[0092] The prediction program 82 is a program for causing the prediction device 2 to execute information processing (see FIG. 8 described below) related to predicting the manifestation form of volunteering behavior for the target student TS. The prediction program 82 includes a series of instructions for this information processing. The learning result data 125 may be managed separately from the prediction program 82, or may be incorporated into the prediction program 82. At least one of the prediction program 82 and the learning result data 125 may be stored in a storage medium 92 instead of or together with the storage unit 22. The prediction device 2 may acquire at least one of the prediction program 82 and the learning result data 125 from the storage medium 92. When the calculation processing of the trained prediction model 4 is executed on an external computer, the learning result data 125 may be omitted from the storage unit 22.

[0093] The prediction device 2 may perform data communication with other computers (such as the model generation device 1) via a communication module 23. The prediction device 2 may be operated using an input device 24 and an output device 25.

[0094] Note that, with regard to the specific hardware configuration of the prediction device 2, components can be omitted, replaced, or added as appropriate depending on the embodiment. For example, the control unit 21 may include multiple hardware processors. The hardware processor may be configured with a microprocessor, FPGA, DSP, GPU, ASIC, etc. At least one of the communication module 23, the input device 24, and the output device 25 may be omitted. The prediction device 2 may be configured with multiple computers. In this case, the hardware configurations of the computers may or may not be the same. The prediction device 2 may be an information processing device designed specifically for the service to be provided, as well as a general-purpose server device, a general-purpose PC, a tablet PC, a terminal device, etc.

[0095] [Software configuration] (Model generation device) 5 schematically shows an example of the software configuration of the model generation device 1 according to this embodiment. The control unit 11 of the model generation device 1 executes instructions included in the model generation program 81 stored in the storage unit 12 using the CPU. As a result, the model generation device 1 operates as a computer including an acquisition unit 111, a training unit 112, and an output processing unit 113 as software modules. That is, in one example, each software module of the model generation device 1 may be realized by the control unit 11 (CPU).

[0096] The acquisition unit 111 is configured to acquire a plurality of data sets 3 each composed of a combination of training student data 30 relating to a sample student RS and a correct answer label 35 indicating the true value of the expression form of the application behavior of the sample student RS for the sample school. The training student data 30 of set 3 includes attribute data 301 indicating attributes of the sample student RS. The training unit 112 is configured to control machine learning of the prediction model 4 using the acquired multiple data sets 3. The machine learning is configured by training the prediction model 4 so that the results of predicting the manifestation form of volunteer behavior from the training student data 30 for each of the multiple data sets 3 match the correct label 35. The output processing unit 113 is configured to output the results of the machine learning.

[0097] (Prediction device) 6 schematically shows an example of the software configuration of the prediction device 2 according to this embodiment. The control unit 21 of the prediction device 2 executes instructions included in the prediction program 82 stored in the storage unit 22 using a CPU. As a result, the prediction device 2 operates as a computer including an acquisition unit 211, a prediction unit 212, and an output processing unit 213 as software modules. That is, in one example, each software module of the prediction device 2 may also be realized by the control unit 21 (CPU).

[0098] The acquisition unit 211 is configured to acquire target student data 50 related to a target student TS, the target student data 50 including attribute data 501 indicating attributes of the target student TS. The prediction unit 212 is configured to predict the manifestation form of the target student TS's application behavior toward a target school by providing the target student data 50 to a trained prediction model 4. In one example, when the calculation processing of the trained prediction model 4 is executed on the prediction device 2, the prediction unit 212 may be provided with a trained prediction model 4 generated by machine learning by holding learning result data 125. The output processing unit 213 is configured to output information related to the result of predicting the manifestation form.

[0099] (others) In the example of the present embodiment, each software module of the model generation device 1 and the prediction device 2 is implemented by a general-purpose CPU. However, the method of implementing each of the above modules is not limited to this example and may be changed as appropriate depending on the embodiment. Some or all of the above software modules may be implemented by one or more dedicated processors or chipsets. Each of the above modules may be implemented as a hardware module. With regard to the software configuration of the model generation device 1 and the prediction device 2, modules may be omitted, replaced, or added as appropriate depending on the embodiment.

[0100] §3 Example of operation [Model generation device] FIG. 7 is a flowchart showing an example of the processing procedure of the model generation device 1 according to this embodiment. The following processing procedure is an example of a model generation method (information processing method) executed by a computer. This model generation method is a method for producing a trained prediction model 4. However, the processing procedure of the model generation device 1 is merely an example, and each step may be changed as much as possible. Furthermore, steps in the following processing procedure may be omitted, replaced, or added as appropriate depending on the embodiment.

[0101] In step S101, the control unit 11 operates as an acquisition unit 111 and acquires a plurality of data sets 3. Each of the plurality of data sets 3 is composed of a combination of training student data 30 related to a sample student RS and a correct answer label 35 indicating the true value of the expression form of the sample student RS's volunteer behavior for the sample school. The training student data 30 includes attribute data 301 indicating the attributes of the sample student RS.

[0102] In one example, the volunteering behavior of the sample student RS may include participating in an open school. The manifestation of the volunteering behavior may include the degree of participation in the open school. The application behavior of the sample student RS may include submitting an application for admission. The manifestation of the application behavior may include the application form for admission.

[0103] In one example, the training student data 30 may further include sample student RS behavior data 303 related to pre-application application. In one example, the training student data 30 may further include sample student RS application form data 305.

[0104] In addition, in one example, the correct answer label 35 may be configured to further indicate the true value of the admission probability of the sample student RS. After acquiring a plurality of data sets 3, the control unit 11 proceeds to the next step S102.

[0105] (Step S102) In step S102, the control unit 11 operates as a training unit 112 and controls machine learning of the prediction model 4 using the acquired multiple data sets 3. The machine learning is performed by training the prediction model 4 so that the results of predicting the manifestation form of volunteering behavior from the training student data 30 for each of the multiple data sets 3 match the true value indicated by the corresponding correct answer label 35. This makes it possible to generate a trained prediction model 4 that has acquired the ability to predict the manifestation form of volunteering behavior from the student data.

[0106] In one example, the machine learning may include training the prediction model 4 so that the result of predicting the admission probability from the training student data 30 for each of the multiple data sets 3 matches the true value indicated by the corresponding correct answer label 35. This makes it possible to generate a trained prediction model 4 that has acquired the ability to further predict the admission probability from the student data. After generating the trained prediction model 4, the control unit 11 proceeds to the next step S103.

[0107] (Step S103) In step S103, the control unit 11 operates as the output processing unit 113 and outputs the results of the machine learning.

[0108] The output destination and the content of the output information may be selected appropriately depending on the embodiment. In one example, the control unit 11 may generate learning result data 125 indicating the results of machine learning as an output process and store the generated learning result data 125 in a predetermined storage area. The predetermined storage area may be, for example, RAM within the control unit 11, the storage unit 12, an external storage device, a storage medium, or a combination thereof. The storage medium may be, for example, a CD, a DVD, a semiconductor memory, or the like. The external storage device may be, for example, a data server such as a NAS. The external storage device may be, for example, an external storage device. When machine learning of the prediction model 4 is executed on an external computer, the learning result data 125 may be generated by the external computer. In another example, the control unit 11 may output information (e.g., loss, etc.) obtained during the machine learning calculation process as an output process. The output destination may be, for example, RAM within the control unit 11, the storage unit 12, the output device 15, an external computer, an external storage device, a storage medium, or a combination thereof.

[0109] When the output of the machine learning results is completed, the control unit 11 ends the processing procedure of the model generation device 1 according to this operation example.

[0110] The generated learning result data 125 may be provided to the prediction device 2 at any timing and by any method. For example, the control unit 11 may transmit the learning result data 125 to the prediction device 2 as part of the output process of step S103 or separately from the output process of step S103. The prediction device 2 may acquire the learning result data 125 (trained prediction model 4) by receiving it. For example, the prediction device 2 may connect to the model generation device 1 or a data server. The prediction device 2 may acquire the learning result data 125 by accessing the learning result data 125 via a network. For example, the prediction device 2 may acquire the learning result data 125 via a storage medium 92. Furthermore, for example, the learning result data 125 may be pre-installed in the prediction device 2.

[0111] Furthermore, the control unit 11 may update or generate new learning result data 125 by periodically or irregularly repeatedly executing the series of processes from step S101 to step S103. During this repetition, at least a portion of the samples used for machine learning may be changed, modified, added, deleted, or the like as appropriate. Then, the control unit 11 may update the learning result data 125 held by the prediction device 2 by providing the updated or newly generated learning result data 125 to the prediction device 2 by any method.

[0112] [Prediction device] FIG. 8 is a flowchart showing an example of a processing procedure of the prediction device 2 according to this embodiment. The following processing procedure is an example of a prediction method (information processing method) executed by a computer. However, the processing procedure of the prediction device 2 is merely an example, and each step may be changed as much as possible. Furthermore, steps in the following processing procedure may be omitted, replaced, or added as appropriate depending on the embodiment.

[0113] (Step S201) In step S201, the control unit 21 operates as the acquisition unit 211 and acquires target student data 50 related to the target student TS. The acquired target student data 50 includes attribute data 501 indicating the attributes of the target student TS.

[0114] In one example, the target student data 50 may further include behavioral data 503 of the target student TS regarding pre-application application. In one example, the target student data 50 may further include application form data 505 of the target student TS. Upon acquiring the target student data 50, the control unit 21 proceeds to the next step S202.

[0115] (Step S202) In step S202, the control unit 21 operates as the prediction unit 212 and predicts the manifestation of the target student TS's application behavior for the target school by providing the acquired target student data 50 to the trained prediction model 4. At least a part of the calculation processing of the trained prediction model 4 may be executed by the prediction device 2 (control unit 21), or may be executed by an external computer other than the prediction device 2.

[0116] In one example, the application behavior of the target student TS may include participating in an open school event. The predicted manifestation of the application behavior may include the degree of participation in the open school event. In one example, the application behavior of the target student TS may include submitting an application for admission. The predicted manifestation of the application behavior may include the manner in which the application is submitted. If the trained prediction model 4 has further acquired the ability to predict the probability of admission, the control unit 21 may predict the probability of admission of the target student TS, along with the manifestation of the application behavior of the target student TS, by providing the acquired target student data 50 to the trained prediction model 4. Once the manifestation of the application behavior has been predicted, the control unit 21 proceeds to the next step S203.

[0117] (Step S203) In step S203, the control unit 21 operates as the output processing unit 213 and outputs information relating to the results of predicting the manifestation form of the application behavior. In one example, if the admission probability of the target student TS is further predicted, the control unit 21 may output information relating to the predicted manifestation form of the application behavior of the target student TS and the admission probability.

[0118] The content of the information to be output and the output destination may be selected appropriately depending on the embodiment. In one example, the control unit 21 may output the predicted result of the manifestation form (and the admission probability) as is. In another example, the control unit 21 may perform any information processing when predicting the manifestation form or on the predicted result of the manifestation form (and the admission probability). The control unit 21 may output the result of this information processing. For example, the control unit 21 may process the prediction result in accordance with a definition in the user interface and output the processed prediction result. Furthermore, for example, the control unit 21 may calculate the influence of each item of input data (such as the target student data 50) on the prediction result. The control unit 21 may output the calculated influence of each item of input data together with the predicted result of the manifestation form. The output destination may be, for example, RAM, the memory unit 22, the output device 25, the storage medium 92, an external computer, an external storage device, etc. When the output of the information related to the prediction result is completed, the control unit 21 terminates the processing procedure related to this operation example.

[0119] [Features] As described above, student attributes may be related to the manifestation of volunteer behavior. In other words, the manifestation of volunteer behavior can be predicted from student attributes. According to the model generation device 1 of this embodiment, a trained prediction model 4 for predicting the manifestation of volunteer behavior can be generated by the processing of step S102. Furthermore, according to the prediction device 2 of this embodiment, the manifestation of volunteer behavior of the target student TS can be predicted by the processing of steps S201 and S202. By utilizing the information on this prediction result output by the processing of step S203, it is expected that student recruitment can be made more efficient.

[0120] §4 Variations Although the embodiments of the present disclosure have been described in detail above, the above description is merely an example of the present disclosure in every respect. It goes without saying that various improvements or modifications can be made without departing from the scope of the present disclosure. The processes and means described in the present disclosure can be freely combined and implemented as long as no technical contradiction occurs.

[0121] §5 Experimental Examples The following experiment was conducted to verify whether it is possible to predict the manifestation of volunteer behavior from student data, although the present disclosure is not limited to the following experimental example.

[0122] First, we collected a dataset (training student data and correct labels) for 41,293 students (high school students) from a certain university (hereafter referred to as "University A") who graduated between 2020 and 2023. The training student data included attribute data such as gender, address area, commute time (travel time), commute costs, the number of designated school recommendation slots the student's school has for University A (by department / total), the location (address area) of the school (high school), whether coeducational or single-sex, the type of governing body (private / public), and deviation score (high school deviation score). The training student data also included behavioral data such as contact time (number of months since first contact, number of months since last contact), number of requests for information, and participation history in open school events. The number of requests for information included aggregated values ​​for route, means, and method, as well as the total values ​​for each. The number of times materials were requested was calculated by route, means, and method, and included the number of requests from vendor magazines, the number of requests via the school's website, the number of requests via the vendor's website, and the number of requests at vendor-sponsored guidance sessions. The history of participation in open schools included the number of times each event type and event type participated, as well as the total number of times each event type participated. The history of participation in open schools included the number of times each event type and event type participated in practical training experiences. The trainee data included the number of times each department applied as application type data. The correct answer labels included the true values ​​of the degree of participation in open schools and whether or not they applied. Furthermore, target student data was collected for 9,330 students who will graduate in 2024 at University A. The target student data included values ​​for each of the above items in the trainee data. For each of the 9,330 students who will graduate in 2024, the true values ​​of the degree of participation in open schools and whether or not they applied were calculated. The values ​​were also collected along with the target student data.

[0123] Next, using a dataset of 41,293 students, machine learning was performed on the prediction models for Experimental Example 1 and Experimental Example 2. For the prediction model for Experimental Example 1, the following data items were used as inputs (explanatory variables): gender, address area, commute time (travel time), commute costs, location of the school (address area), coeducational / single-sex school status, type of operating organization (private / public), deviation score (high school deviation score), time of contact (number of months since last contact), number of requests for information (total), total number of participations in open school programs, and number of applications (by department) from the above data items. On the other hand, the predictive model for the second experimental example used the following inputs (explanatory variables): gender, address area, commute time (travel time), commute costs, school location (address area), coeducational / single-sex school, type of governing body (private / public), deviation score (high school deviation score), time of contact (number of months since first contact, number of months since last contact), number of requests for information (by route, means, and method), number of open school attendances (by event type and method, and total), and number of applications (by department). A decision tree model was used for the predictive models for the first and second experimental examples. Random forest was used as the machine learning method. Trained predictive models for the first and second experimental examples were then generated. The trained predictive models were then used to predict whether or not 9,330 students in 2024 would apply. Because the output of the prediction model was expressed as a probability, a threshold was determined using a recall standard of 95%. Based on the threshold, each student was then asked whether to participate in the open school program and whether to apply. In other words, a positive or negative result was determined using the threshold determined according to the 95% recall standard. The accuracy of the trained prediction model for each experimental example was then verified based on whether the predicted results matched the true value.

[0124] As a result of the verification, in the first experimental example, the positive rate (probability of being judged as "will participate") for whether or not to participate in open school was 42.6%, with a prediction accuracy of 93.8%. For whether or not to apply, the positive rate (probability of being judged as "will apply") was 18.5%, with a prediction accuracy of 96.7%. Meanwhile, in the second experimental example, the positive rate for whether or not to participate in open school was 34.3%, with a prediction accuracy of 96.9%. For whether or not to apply, the positive rate was 16.0%, with a prediction accuracy of 96.2%. These results suggest that it is possible to appropriately predict the manifestation of application behavior based on student data.

[0125] Furthermore, compared to the first experimental example, the second experimental example showed improved prediction accuracy regarding whether or not students would attend open school, while the prediction accuracy regarding whether or not they would apply was roughly the same. These results suggest that including the number of designated school recommendation slots in the attribute data can improve prediction accuracy. Furthermore, we found that including at least one of the following behavioral data can improve prediction accuracy: (1) the date of first contact; (2) the number of requests for information by route, means, and method; and (3) the number of open school attendances by event type and event type. These explanatory variables were particularly useful when predicting the degree of open school participation. It is likely that the improvement in prediction accuracy was due to the fact that the number of requests for information and the number of open school attendances were broken down by type rather than the total value. Therefore, these results suggest that including at least one of the following behavioral data can improve prediction accuracy: (1) the date of first contact; (2) the number of requests for information by route, means, and method; and (3) the number of open school attendances by event type and event type.

[0126] Furthermore, using SHAP, the influence (contribution) of input variables on the prediction results of the prediction model for each experimental example was calculated. As a result, in the first experimental example, each item of attribute data, such as address area (0.80%), commuting time (2.40%), and high school deviation score (3.90%), had a moderate influence. From these results, it is possible to appropriately predict the manifestation of volunteer behavior based on attribute data. This suggests that predicting volunteer behavior is possible. In the second experimental example, behavioral data items such as the number of months since last contact (12.40%), the number of months since first contact (4.00%), the number of requests from vendor information magazines (2.80%), the number of requests via the school's website (1.90%), the number of requests via vendor websites (0.90%), the number of requests at vendor-sponsored guidance sessions (0.40%), the total number of open school attendances (12.00%), and the number of internships (1.10%) had a moderate influence. These results suggest that including behavioral data in student data can improve the accuracy of predicting the manifestation of volunteer behavior. In particular, the following were suggested to be useful explanatory variables for predicting the manifestation of volunteer behavior: (1) the date of first contact; (2) the number of requests for materials by route, means, and / or method; and (3) the number of open school attendances by event type and / or method. Furthermore, in the first and second experimental examples, each item of the aspiration pattern data, such as the number of times a student applied (0.70%), had a moderate degree of influence. These results suggest that by including aspiration pattern data in student data, it is possible to improve the accuracy of predicting the manifestation of student application behavior. [Explanation of symbols]

[0127] 1...Model generation device, 11...control unit, 12...storage unit, 3…dataset, 30...Training student data, 301...Attribute data, 303...Behavioral data, 305...Aspirational data, 35...correct label, 2...Prediction device, 21...control unit, 22...storage unit, 50...Target student data, 501...Attribute data, 503...Behavioral data, 505...Aspirational data, 4. Predictive model

Claims

1. acquiring target student data relating to the target student, the target student data including attribute data indicating attributes of the target student; Predicting the manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning; and outputting information about the results of the predicted expression pattern; a control unit configured to perform The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The volunteering activity includes participating in an open school; The manifestation of the volunteer behavior includes the degree of participation in the open school. Prediction device.

2. The attribute data of the target student includes the number of designated school recommendation slots held by the school to which the target student belongs, for the target school; The attribute data of the sample student includes the number of designated school recommendation slots for the sample school held by the school to which the sample student belongs. The prediction device according to claim 1 .

3. The application includes submitting an application for admission; The form of expression of the application behavior includes the form of application of the admission application. The prediction device according to claim 1 .

4. Obtaining target student data relating to a target student, the target student data including attribute data indicating attributes of the target student; By providing the target student data to a trained prediction model generated by machine learning, , predicting the manifestation of the target student's application behavior to the target school; and outputting information about the results of the predicted expression pattern; a control unit configured to perform The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The target student data and the trainee student data further include pre-application behavioral data regarding volunteering; The behavioral data includes at least one of: (1) the date of first contact; (2) the number of requests for materials by at least one of route, means, and method; and (3) the number of participations in open school events by at least one of event type and holding method. Prediction device.

5. The target student data and the training student data further include aspiration data; The prediction device according to claim 1 .

6. The correct answer label is configured to further indicate a true value of the admission probability of the sample student; The machine learning includes training the prediction model for each of the plurality of datasets so that a result of predicting the admission probability from the training student data matches the correct answer label; predicting the expression pattern of the application behavior by predicting the expression pattern of the application behavior and the probability of the target student being admitted to the target school; outputting information about the predicted expression pattern; and outputting information about the predicted expression pattern and the predicted admission probability. The prediction device according to claim 1 .

7. 1. A computer-implemented prediction method comprising: acquiring target student data relating to the target student, the target student data including attribute data indicating attributes of the target student; Predicting the manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning; and outputting information about the results of the predicted expression pattern; Including, The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The volunteering activity includes participating in an open school; The manifestation of the volunteer behavior includes the degree of participation in the open school. Forecasting methods.

8. A computer-implemented prediction method comprising: Target student data relating to target students, including attribute data indicating the attributes of the target students Obtaining targeted student data, Predicting the manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning; and outputting information about the results of the predicted expression pattern; Including, The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The target student data and the trainee student data further include pre-application behavioral data regarding volunteering; The behavioral data includes at least one of: (1) the date of first contact; (2) the number of requests for materials by at least one of route, means, and method; and (3) the number of participations in open school events by at least one of event type and holding method. Forecasting methods.

9. A prediction program for causing a computer to execute a prediction method, The prediction method includes: acquiring target student data relating to the target student, the target student data including attribute data indicating attributes of the target student; Predicting the manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning; and outputting information about the results of the predicted expression pattern; Including, The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The volunteering activity includes participating in an open school; The manifestation of the volunteer behavior includes the degree of participation in the open school. Prediction program.

10. A prediction program for causing a computer to execute a prediction method, comprising: The prediction method includes: acquiring target student data relating to the target student, the target student data including attribute data indicating attributes of the target student; Predicting the manifestation of the target student's application behavior to a target school by providing the target student data to a trained prediction model generated by machine learning; and outputting information about the results of the predicted expression pattern; Including, The machine learning uses a plurality of data sets each composed of a combination of training student data on sample students and correct answer labels indicating true values ​​of the expression forms of the application behavior of the sample students to the sample schools; the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The target student data and the trainee student data further include pre-application behavioral data regarding volunteering; The behavioral data includes at least one of: (1) the date of first contact; (2) the number of requests for materials by at least one of route, means, and method; and (3) the number of participations in open school events by at least one of event type and holding method. Prediction program.

11. 1. A computer-implemented method for generating a model, comprising: Obtaining a plurality of datasets each composed of a combination of training student data on a sample student and a ground truth label indicating a true value of the expression form of the sample student's application behavior for the sample school; Controlling machine learning of a predictive model using the acquired plurality of datasets; and outputting the results of the machine learning; Including, the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The volunteering activity includes participating in an open school; The manifestation of the volunteer behavior includes the degree of participation in the open school. Model generation method.

12. The attribute data includes the number of designated school recommendation slots for the sample school held by the school to which the sample student belongs, The model generation method of claim 11.

13. The application includes submitting an application for admission; The form of expression of the application behavior includes the form of application of the admission application. The model generation method of claim 11.

14. A computer-implemented method for generating a model, comprising: Obtaining a plurality of datasets each composed of a combination of training student data on a sample student and a ground truth label indicating a true value of the expression form of the sample student's application behavior for the sample school; Controlling machine learning of a predictive model using the acquired plurality of datasets; and outputting the results of the machine learning; Including, the training student data includes attribute data indicating attributes of the sample students; the machine learning is configured by training the prediction model for each of the plurality of datasets so that a result of predicting an expression form of the volunteer behavior from the training student data matches the correct answer label; The trainee data further includes pre-application behavioral data regarding the applicant; The behavioral data includes at least one of: (1) the date of first contact; (2) the number of requests for materials by at least one of route, means, and method; and (3) the number of participations in open school events by at least one of event type and holding method. Model generation method.

15. The training student data further includes aspiration data.

15. A model generation method according to any one of claims 11 to 14.

16. The ground truth label is configured to further indicate the true value of the admission probability of the sample student; The machine learning includes training the prediction model for each of the plurality of datasets so that a result of predicting the admission probability from the training student data matches the correct answer label.

15. A model generation method according to any one of claims 11 to 14.

Images