JP7754643B2 - Prediction device, learning device, prediction method, learning information production method, and program - Google Patents

Description

The present invention relates to a prediction device that predicts one or more test values corresponding to a weight value.

In the past, there have been information processing devices that can provide subjects with useful information related to health management. These information processing devices include a factor extraction unit that evaluates, for each subject, the correlation between each lifestyle index item related to physical activity or lifestyle habits and each bioindicator item related to the physiological state of the body, and extracts sets of lifestyle indexes and bioindicators for items where the evaluated correlation is higher than a predetermined correlation as factors that may affect health; and an input support unit that, when one of the acquired lifestyle indexes and bioindicators is a lifestyle index or bioindicator for an item related to a set extracted by the factor extraction unit, outputs input guidance for the other related to the set to which the one belongs, in accordance with changes from the previous value of the other (see, for example, Patent Document 1).

JP 2016-189085 A

However, conventional technology was unable to present predicted test values for target body weight (hereinafter referred to as "target body weight" where appropriate). In particular, conventional technology was unable to present predicted test values for two or more target body weights, which meant that users were unable to grasp the gradual weight status required to improve to or approach the target test value, and this did not sufficiently motivate them to lose weight. Test values include, for example, blood pressure, triglycerides, HDL, LDL, and blood glucose levels.

The prediction device of this first invention is a prediction device comprising: a learning information storage unit that stores learning information acquired using two or more pieces of training data that associate a person's weight measurement value with one or more other types of test values; a target acquisition unit that acquires two or more future target weight values for a user; a prediction unit that uses the learning information to acquire predicted values for each type of one or more future test values that correspond to each of the two or more target weight values; and a prediction output unit that outputs one or more predicted values.

This configuration makes it possible to present predicted test values for two or more target weight values. As a result, it is possible to provide the subject with useful information regarding health management.

Furthermore, the prediction device of the second invention differs from the prediction device of the first invention in that the training data includes test values of the same type at two or more different time periods and weight measurements at two or more different time periods.

With this configuration, historical test value information can be used to present predicted test values for two or more target weight values.

Furthermore, the prediction device of the third invention is a prediction device according to the first or second invention, in which the learning information is an arithmetic formula obtained by performing linear regression analysis on two or more pieces of training data.

This configuration makes it easy to present predicted test values for two or more target weight values.

Furthermore, the prediction device of the fourth invention is a prediction device in which, compared to the third invention, the teacher data includes two or more other types of test values, and the learning information is information obtained using two or more degenerate teacher data that are pairs of one or more test values and weight measurements corresponding to each of the other one or more types selected based on the importance of each of the other two or more types of test values.

This configuration makes it possible to easily and accurately present predicted test values for two or more target weight values.

Furthermore, the prediction device of the fifth invention is a prediction device according to any one of the first to fourth inventions, in which the target acquisition unit acquires two or more target weight values using a single user's most recent weight measurement value.

This configuration makes it possible to present predicted test values for two or more appropriate target weight values.

Furthermore, the prediction device of the sixth invention is a prediction device in which, compared to the fifth invention, the target acquisition unit acquires a value obtained by subtracting two or more predetermined values from a single user's most recent weight measurement value.

This configuration makes it possible to present predicted test values for two or more appropriate target weight values.

Furthermore, the prediction device of the seventh invention is a prediction device according to any one of the first to sixth inventions, wherein the learning information storage unit stores learning information corresponding to the type of each of the two or more predicted values to be output, the prediction unit acquires learning information corresponding to each of the two or more predicted value types and uses the learning information to acquire predicted values corresponding to each of the two or more target weight values, and the prediction output unit outputs the predicted value for each of the two or more predicted value types acquired by the prediction unit.

With this configuration, two or more types of highly accurate prediction values can be presented by using learning information corresponding to the acquired type of prediction value.

Furthermore, the prediction device of the eighth invention is a prediction device according to any one of the first to seventh inventions, in which the training data corresponds to one or more attribute values of a single person, the learning information storage unit stores learning information for each of two or more conditions using one or more attribute values, and the prediction unit obtains one or more predicted values corresponding to each of two or more target weight values using the learning information corresponding to the condition that matches one or more attribute values of a single user.

This configuration allows for highly accurate predictions to be presented by using learning information based on the user's attribute values.

The prediction device of the ninth invention is a prediction device according to any one of the first to eighth inventions, further comprising a component that generates output information that associates two or more target weight values with one or more types of predicted values acquired by the prediction unit, and the prediction output unit outputs the output information.

This configuration makes it possible to clearly present the relationship between two or more target weight values and predicted values.

The prediction device of the tenth invention is a prediction device according to any one of the first to eighth inventions, further comprising a reference value storage unit in which reference values for one or more other types of test values are stored, a configuration unit that generates output information according to the comparison result between the predicted value obtained by the prediction unit and the reference value corresponding to the predicted value, and a prediction output unit that outputs the output information.

This configuration allows for appropriate predictions to be presented.

The learning device of this eleventh invention is a learning device that includes a training data storage unit that stores two or more pieces of training data that correspond to weight measurements obtained from a single person at the same time and one or more other types of test values, a training information acquisition unit that acquires training information using the two or more pieces of training data, and a training information storage unit that accumulates the training information.

This configuration makes it possible to obtain information for obtaining predicted test values for two or more target weight values.

The prediction device of the present invention can present predicted test values for two or more target weight values.

Block diagram of learning device 1 according to embodiment 1 A diagram illustrating an example of the acquisition process of the training data. A flowchart illustrating an example of the operation of the learning device 1. A flowchart illustrating an example of the explanatory variable selection process Figure showing the teacher data management table Block diagram of a prediction device 2 according to a second embodiment A flowchart illustrating an example of the operation of the prediction device 2 A flowchart illustrating an example of the target weight acquisition process. Figure showing an example of the output Block diagram of the prediction device 3 Block diagram of the prediction system A Overview of the computer system in the above embodiment Block diagram of the computer system

Embodiments of a learning device, a prediction device, etc. will be described below with reference to the drawings. Note that components with the same reference numerals in the embodiments perform similar operations, and therefore may not be described again.

(Embodiment 1)
In this embodiment, a learning device that acquires learning information using two or more pieces of teacher data will be described.

Figure 1 is a block diagram of a learning device 1 in this embodiment. The learning device 1 includes a storage unit 11, a learning information acquisition unit 12, and a learning information accumulation unit 13. The storage unit 11 includes a teacher data storage unit 111.

The storage unit 11 stores various types of information. Examples of such information include training data (described below), learning devices (described below), selection conditions (described below), and attribute value conditions (described below).

The teacher data storage unit 111 stores two or more sets of teacher data. The teacher data corresponds to, for example, a user identifier and one or more attribute values of a person. The user identifier is an attribute value of a person, such as an ID, name, email address, or telephone number. The person is the person from whom the test values contained in the teacher data were obtained.

Attribute values are attribute values of the person from whom the test value was obtained. Attribute values include, for example, gender, age, generation, occupation, and information about lifestyle habits (e.g., whether or not the person smokes, how much they smoke, whether or not they exercise, the amount of exercise they do in a specified period of time, how much they exercise, whether or not they drink alcohol, and how much they drink).

The training data is information that associates a weight measurement value (hereinafter referred to as "weight value" or "weight") with one or more other types of test values. Note that a weight measurement value is typically a measured weight value, but it does not need to be accurate and may be a self-reported value by the user, although this is not preferred.

Teacher data is, for example, information that includes a weight measurement value and one or more other types of test values. Teacher data is, for example, link information that links a weight measurement value with one or more other types of test values. Note that the data structure, etc., is not important as long as the weight measurement value and one or more other types of test values can be obtained using the teacher data. Furthermore, the one or more other types of test values are, for example, the values resulting from a health checkup for a person.

The weight measurement value and one or more other types of test values corresponding to one training data are information obtained from one person at the same time. "Same time" means at the same time, for example, on the same day, within one week, within one month, etc., and the range of time can vary.

The one or more other types of test values are test values other than body weight. The one or more other types of test values are, for example, values related to obesity, values related to blood pressure, values related to lipids, and values related to blood glucose. Values related to obesity are, for example, waist circumference and BMI. Values related to blood pressure are, for example, systolic blood pressure and diastolic blood pressure. Values related to lipids are, for example, triglycerides, HDL cholesterol, and LDL cholesterol. Values related to blood glucose are, for example, fasting blood glucose level, HbA1C, and the presence or absence of urinary sugar. Note that test values are typically any of a number of multiple values, but may also be any of two-value information (for example, (-) or (+)).

Training data is a set of values for one or more explanatory variables and values for one or more objective variables.

For example, the value of one or more explanatory variables is weight, and the value of one or more response variables is a single laboratory value. An example of the single laboratory value is any one or more other types of laboratory values.

For example, the values of one or more explanatory variables are one or more past weights and one or more past test values, and the value of one or more dependent variables is a single test value. For example, the values of one or more explanatory variables are (weight this year, weight last year, weight two years ago, weight three years ago, blood pressure last year, blood pressure two years ago, blood pressure three years ago), and the value of one or more dependent variables is (blood pressure this year). For example, the values of one or more explanatory variables are (weight this year, weight last year, weight two years ago, weight three years ago, blood pressure this year, blood pressure last year, blood pressure two years ago, blood pressure three years ago), and the value of one or more dependent variables is (blood pressure this year).

For example, the values of one or more explanatory variables consist of a set of a person's most recent weight and two or more different past time points for that person, and the value of the dependent variable consists of one or more most recent test values. The set consists of a weight at one time point and one or more other types of test values. It is preferable that the value of the dependent variable be one other test value, but it may also be two or more other test values.

The other types of test values contained in the training data may be just one type of test value (e.g., blood pressure) or two or more types of test values (e.g., blood pressure and blood glucose level).

It is preferable that each value constituting the training data be associated with time information that specifies when it was acquired. Time information may be, for example, information indicating the date, time, month, year, or period of time elapsed since the present, but the level of detail and data structure are not important.

The learning information acquisition unit 12 acquires two or more pieces of teacher data from the teacher data storage unit 111 and acquires learning information using the two or more pieces of teacher data. The learning information is information for acquiring the values of one or more objective variables when the values of one or more explanatory variables are given. The learning information acquisition unit 12 may also acquire learning information using two or more pieces of reduced teacher data, as described below. Note that reduced teacher data can also be considered teacher data.

When the training data for one person in the training data storage unit 111 is a set of laboratory values including weights for α years (α is a natural number greater than or equal to 2), it is preferable that the learning information acquisition unit 12 acquires two or more sets of laboratory values including weights for β years (β < α) from the training data for one or more people, staggering the years, acquires the two or more partial training data, and uses the two or more partial training data to acquire learning information. Note that the partial training data is also training data for acquiring learning information.

For example, if a set of test values for one individual exists for six years (teaching data "A"), the learning information acquisition unit 12 shifts the base year and acquires three partial teaching data sets ("A1," "A2," and "A3") for four years from the set of test values for that individual. As a result, if there are R records of teaching data for six years, for example, the learning information acquisition unit 12 acquires 3R pieces of partial teaching data (see Figure 2). The learning information acquisition unit 12 then acquires learning information using the 3R pieces of partial teaching data.

For example, for each of two or more attribute value conditions, the learning information acquisition unit 12 acquires two or more pieces of training data that satisfy the attribute value conditions from the training data storage unit 111, and acquires learning information using the two or more pieces of training data. Note that an attribute value condition is a condition related to one or more attribute values. Examples of attribute value conditions are "male and over 60 years old," "male and in his 40s," and "yes, smoking habit." The content of the attribute value condition is not important. The attribute value conditions are stored in the storage unit 11.

The training data may contain two or more candidate explanatory variable values, and the learning information acquisition unit 12 may select one or more explanatory variables from the two or more candidate explanatory variables whose importance satisfies a selection condition. Importance refers to the degree of influence on the target variable, and may also be referred to as influence or contribution. The selection condition refers to a condition for the explanatory variables selected to acquire the training information, and is a condition related to importance. Examples of selection conditions include "importance equal to or greater than a threshold," "importance greater than a threshold," and "top N importance (N is a natural number greater than 1)." The selection conditions are stored in the storage unit 11. Training data in which explanatory variables whose importance does not satisfy the selection condition are removed from the two or more candidate explanatory variables contained in the training data is called reduced training data. Reduced training data may also be referred to as dimensionally reduced training data.

The learning information acquisition unit 12, for example, calculates a correlation coefficient between the values of each objective variable contained in two or more pieces of training data and the values of candidate explanatory variables, and acquires an importance that is the correlation coefficient, or an importance that increases as the correlation coefficient increases.

The learning information acquisition unit 12, for example, performs a learning process using a random forest algorithm on two or more pieces of training data to acquire a learning device. Next, the learning information acquisition unit 12, for example, uses the learning device to acquire the importance of each explanatory variable.

The method by which the learning information acquisition unit 12 acquires the importance of each candidate explanatory variable is not important.

The learning information may be, for example, an arithmetic expression, a learning device, or a correspondence table. Below, an example of processing by the learning information acquisition unit 12 for each of the three types of learning information will be described.
(1) When the learning information is an arithmetic formula

The learning information acquisition unit 12 acquires two or more pieces of training data from the training data storage unit 111, performs regression analysis (e.g., multiple regression analysis) on the two or more pieces of training data, and acquires an arithmetic formula. The arithmetic formula typically takes the values of two or more explanatory variables as input and outputs the value of one objective variable. Note that regression analysis itself is a well-known technique, so a detailed explanation will be omitted.

The calculation formula, for example, is "Other types of test values for latest weight = f(latest weight, last year's weight, ..., weight N years ago, test value last year, ..., test value N years ago)" (for example, N is a natural number greater than or equal to 2). The calculation formula, for example, is "Other types of test values for latest weight = f(latest weight, last year's weight, ..., weight N years ago)" (for example, N is a natural number greater than or equal to 2). The calculation formula, for example, is "Other types of test values for latest weight = f(latest weight)".

Note that the "latest weight" in the above arithmetic formula can be substituted with a future target weight value. The input parameters in the arithmetic formula can be only the future target weight value, a set of two or more time-series weight values, a future target weight value and the most recent test value, or a future target weight value and a set of two or more time-series test values. The test values that are input parameters in the arithmetic formula can be the same type of test value as the output test value, or they can be a different type of test value.

The learning information acquisition unit 12, for example, acquires two or more pieces of teacher data from the teacher data storage unit 111, performs linear regression analysis on the two or more pieces of teacher data, and acquires a linear arithmetic formula.

The learning information acquisition unit 12, for example, acquires two or more pieces of teacher data from the teacher data storage unit 111, performs nonlinear regression analysis on the two or more pieces of teacher data, and acquires a nonlinear arithmetic formula.

When the training data contains only one explanatory variable value, the learning information acquisition unit 12 may perform a simple regression analysis on two or more training data to obtain an arithmetic formula. In such a case, the arithmetic formula may be, for example, "other type of test value for latest weight = f (latest weight)." Note that the target weight value may be substituted for the latest weight.

It is preferable that the learning information acquisition unit 12 acquires an arithmetic formula using only the values of one or more explanatory variables selected from two or more sets of teacher data using a selection condition and the value of the objective variable. In other words, it is preferable that the learning information acquisition unit 12 reduces the data contained in the teacher data (dimensionality compression), and then performs regression analysis on the dimensionally compressed two or more sets of teacher data (appropriately referred to as "compressed teacher data") to acquire an arithmetic formula.
(2) When the learning information is a learning module

The learning information acquisition unit 12 acquires two or more pieces of training data from the training data storage unit 111, performs a learning process on the two or more pieces of training data using a machine learning algorithm, and acquires a learning device. The learning device may also be called a classifier, model, etc. The algorithm for the machine learning learning process is not important. Examples of machine learning include deep learning, random forest, decision tree, SVR, and SVM. The algorithm for the machine learning prediction process described below is also not important, as is the learning process.

The learning information acquisition unit 12 provides the acquired two or more pieces of training data to a machine learning learning processing module and executes the module to acquire a learner. Examples of machine learning learning processing modules include the TensorFlow library, fastText, tinySVM, and the random forest module of the R language. The machine learning prediction processing module, which will be described later, is also the same as the machine learning learning processing module.

It is preferable that the learning information acquisition unit 12 acquires an arithmetic expression using only the values of one or more explanatory variables selected from two or more sets of teacher data using a selection condition and the value of the objective variable. In other words, it is preferable that the learning information acquisition unit 12 reduces the data contained in the teacher data (dimensionality compression), and then performs a learning process using a machine learning algorithm on the two or more compressed sets of teacher data to acquire a learner.
(3) When the learning information is a correspondence table

The learning information acquisition unit 12 acquires two or more pieces of teacher data from the teacher data storage unit 111 and creates a correspondence table containing correspondence information indicating the correspondence between the values of one or more explanatory variables and the values of one or more objective variables from each of the two or more pieces of teacher data. The correspondence information includes, for example, a vector having the values of two or more explanatory variables as elements and the value of one objective variable. The correspondence information is, for example, link information that identifies the link between a vector having the values of two or more explanatory variables as elements and the value of one objective variable.

It is preferable that the learning information acquisition unit 12 acquires compressed training data from each of two or more sets of training data, and constructs a correspondence table from the two or more compressed training data that contains correspondence information indicating the correspondence between the values of one or more explanatory variables and the values of one or more objective variables.

The learning information accumulation unit 13 accumulates the learning information acquired by the learning information acquisition unit 12. When the learning information acquisition unit 12 acquires learning information for each of two or more attribute value conditions, the learning information accumulation unit 13 accumulates the learning information acquired by the learning information acquisition unit 12 in association with the attribute value conditions. When the learning information acquisition unit 12 acquires learning information for each of two or more objective variables, the learning information accumulation unit 13 accumulates the learning information acquired by the learning information acquisition unit 12 in association with the objective variables.

The learning information accumulation unit 13 accumulates the learning information, for example, in the storage unit 11. The learning information accumulation unit 13 accumulates the learning information, for example, in an external device (not shown).

The storage unit 11 and the teacher data storage unit 111 are preferably non-volatile recording media, but can also be implemented as volatile recording media.

The process by which information is stored in the storage unit 11, etc. is not important. For example, information may be stored in the storage unit 11, etc. via a recording medium, information transmitted via a communication line, etc. may be stored in the storage unit 11, etc., or information input via an input device may be stored in the storage unit 11, etc.

The learning information acquisition unit 12 and learning information storage unit 13 are typically implemented using a processor, memory, etc. The processing procedures of the learning information acquisition unit 12, etc. are typically implemented using software, and this software is recorded on a recording medium such as a ROM. However, they may also be implemented using hardware (dedicated circuits). The processor may be, for example, a CPU, MPU, GPU, etc., and its type is not important.

Next, an example of the operation of the learning device 1 will be explained using the flowchart in Figure 3. In this example, learning information is constructed for each objective variable.

(Step S301) The learning information acquisition unit 12 assigns 1 to counter i.

(Step S302) The learning information acquisition unit 12 determines whether the i-th objective variable constituting the learning information exists. If the i-th objective variable exists, the process proceeds to step S303; if the i-th objective variable does not exist, the process ends.

(Step S303) The learning information acquisition unit 12 selects explanatory variables corresponding to the i-th objective variable. An example of such explanatory variable selection processing will be described using the flowchart in Figure 4. Note that the processing of step S303 does not necessarily have to be performed.

(Step S304) The learning information acquisition unit 12 acquires two or more pieces of teacher data from the teacher data storage unit 111. The learning information acquisition unit 12 acquires compressed teacher data having values corresponding to one or more explanatory variables selected in step S303 and a value corresponding to the i-th objective variable from among the elements of each acquired teacher data. Note that if the processing of step S303 does not exist, in this step the learning information acquisition unit 12 simply acquires two or more pieces of teacher data from the teacher data storage unit 111.

(Step S305) The learning information acquisition unit 12 acquires learning information using the two or more compressed training data sets acquired in step S304. Note that the learning information may be, for example, an arithmetic formula, a learning device, or a correspondence table. Details of how to acquire each of these have been described above, so further explanation will be omitted here.

(Step S306) The learning information accumulation unit 13 accumulates the learning information acquired in step S305 in association with the identifier of the i-th objective variable.

(Step S307) The learning information acquisition unit 12 increments counter i by 1. Return to step S302.

Next, an example of the explanatory variable selection process in step S303 will be explained using the flowchart in Figure 4.

(Step S401) The learning information acquisition unit 12 acquires the value of the objective variable from each of two or more pieces of teacher data stored in the teacher data storage unit 111. The learning information acquisition unit 12 constructs a vector whose elements are the values of the objective variables acquired from each of the two or more pieces of teacher data. Note that the value of the objective variable is the value of the i-th objective variable in the flowchart of Figure 3.

(Step S402) The learning information acquisition unit 12 assigns 1 to counter j.

(Step S403) The learning information acquisition unit 12 determines whether a candidate for the jth explanatory variable exists. If a candidate for the jth explanatory variable exists, the process proceeds to step S404; if not, the process proceeds to step S408.

(Step S404) The learning information acquisition unit 12 acquires candidate values for the jth explanatory variable from each of two or more pieces of teacher data stored in the teacher data storage unit 111. The learning information acquisition unit 12 constructs a vector whose elements are the candidate values for the jth explanatory variable acquired from each of the two or more pieces of teacher data.

(Step S405) The learning information acquisition unit 12 acquires the correlation coefficient between the vector constructed in step S401 and the vector constructed in step S404.

(Step S406) The learning information acquisition unit 12 temporarily stores the importance in a buffer (not shown) in association with the identifier of the jth explanatory variable. Note that the importance is a correlation coefficient or information based on the correlation coefficient.

(Step S407) The learning information acquisition unit 12 increments counter j by 1. Return to step S403.

(Step S408) The learning information acquisition unit 12 determines whether the corresponding importance level satisfies the selection condition for each of the identifiers of two or more explanatory variables. Then, it acquires the identifiers of one or more explanatory variables paired with the importance level that satisfies the selection condition. It then returns to the higher-level processing. Note that the explanatory variables identified by these one or more identifiers are the selected explanatory variables.

The following describes a specific example of the operation of the learning device 1 in this embodiment.

Currently, the teacher data management table shown in Figure 5 is stored in the teacher data storage unit 111 of the learning device 1. The teacher data management table stores two or more pieces of teacher data. The teacher data management table manages two or more records each having a "user ID," "test subject," "health checkup test values for each year from this year up to N years ago," and "attribute values." In this case, the "attribute values" include "gender," "age," and "smoking habits."

In such a situation, the learning device 1 operates as follows. That is, the learning device 1 acquires learning information using the training data in the training data management table. The learning information here is information for accepting future (here, next year) weight and outputting a predicted test value from one or more test values other than weight (e.g., abdominal circumference, systolic blood pressure, diastolic blood pressure, triglycerides, HDL cholesterol, LDL cholesterol, fasting blood glucose, HbA1C, urine sugar). In this case, the predicted test value is a predicted value for next year.

Here, we will explain the case where the objective variable is "systolic blood pressure," for example. In other words, we will explain the case where the learning device 1 acquires learning information for outputting next year's systolic blood pressure when next year's target weight is input.

First, the learning information acquisition unit 12 of the learning device 1 acquires the importance of each weight from one year ago to N years ago, which are candidates for explanatory variables, and each test value from one year ago to N years ago, using the method described above. Note that each test value is a type of test value other than weight, and in this case, is abdominal circumference, systolic blood pressure, diastolic blood pressure, triglycerides, HDL cholesterol, LDL cholesterol, fasting blood glucose, HbA1C, and urinary sugar.

Next, the learning information acquisition unit 12 acquires the selection condition (here, for example, "importance >= threshold") from the storage unit 11. Next, it is assumed that the learning information acquisition unit 12 selects the systolic blood pressure for the past three years and the weight for the past three years. In other words, the compressed training data has a structure, for example, of (weight this year, weight one year ago, weight two years ago, weight three years ago, systolic blood pressure this year, systolic blood pressure one year ago, systolic blood pressure two years ago, systolic blood pressure three years ago).

Next, the learning information acquisition unit 12 acquires compressed teacher data from each record in the teacher data management table of FIG. 5. Then, the learning information acquisition unit 12 acquires teacher data to be used for linear regression analysis from the compressed teacher data. That is, for example, the learning information acquisition unit 12 acquires ( _WE11 , HB11), (WE12, _HB12 ), ( _WE13 , _HB13 ), and ( _WE14 , _{HB14 )} from the record for user ID "U001." The learning information acquisition unit 12 also acquires ( _WE21 , _HB21 ), (WE22, _HB22 ), ( _WE23 , _HB23 ) _{, and (WE24 , HB24 )} from the record for user ID "U002." Then, the learning information acquisition unit 12 acquires four compressed teacher data from each of all records in the teacher data management table.

Next, the learning information acquiring unit 12 performs linear regression analysis on all the acquired compressed training data, and acquires the arithmetic formula "systolic blood pressure = α ₁ × weight + C ₁ " (α ₁ and C ₁ are positive constants).

Next, the learning information accumulation unit 13 associates the calculation formula "systolic blood pressure = α × body weight + C" with the test value type identifier "systolic blood pressure" and accumulates it in the storage unit 11.

Similarly, the learning information acquisition unit 12 of the learning device 1 acquires two or more compressed teacher data for each of one or more other types of test value, performs linear regression analysis on the two or more compressed teacher data, and acquires the arithmetic formula "test value = α _n × weight + C _n " (α _n and C _n are positive constants).

Next, the learning information accumulation unit 13 associates the calculation formula "systolic blood pressure = α × body weight + C" with each of the other types of identifiers and accumulates it in the storage unit 11.

Through the above processing, the learning information of the learning device 1 was obtained. Note that in the above example, the learning device 1 obtained the learning information through regression analysis, but as mentioned above, it may also obtain learning information that is an arithmetic formula using another algorithm, or it may obtain learning information that is a learner through machine learning learning processing, or it may obtain learning information that is a correspondence table.

When the learning device 1 performs machine learning learning processing, for example, the learning information acquisition unit 12 acquires (weight this year, weight one year ago, ..., weight N years ago, test value one year ago, ..., test value N years ago) as a set of explanatory variable values from each training data. The learning information acquisition unit 12 also acquires (test value this year) as the value of the objective variable. Then, a learning process is performed on two or more compressed training data consisting of a set of explanatory variable values and the value of the objective variable, and a learning device is acquired. Note that the test value may be various, such as systolic blood pressure or diastolic blood pressure. Furthermore, while N is preferably set to 3, other values may be used.

The learning device 1 may also acquire learning information for each of one or more attribute value conditions of the user. Attribute value conditions may be, for example, "Male & Smoking Yes," "Male & No Smoking," "Female & Smoking Yes," and "Female & No Smoking." In such a case, four pieces of learning information are acquired for each type of test value of the objective variable.

As described above, according to this embodiment, learning information can be acquired to obtain predicted test values for two or more target weight values.

The processing in this embodiment may be implemented using software. This software may be distributed by software download, etc. The software may also be recorded on a recording medium such as a CD-ROM and distributed. This also applies to other embodiments in this specification. The software that implements the learning device 1 in this embodiment is the following program. In other words, this program causes a computer that can access a training data storage unit that stores two or more pieces of training data that associate weight measurements obtained over the same period for a single person with one or more other types of test values to function as a training information acquisition unit that acquires training information using the two or more pieces of training data, and a training information storage unit that stores the training information.

(Embodiment 2)
In this embodiment, a prediction device will be described that applies learning information acquired by the learning device 1 to a plurality of future target weight values of one user, and acquires and outputs test values (predicted values) of one or more objective variables corresponding to each of the future target weight values of the user.

In addition, this embodiment describes a prediction device that automatically obtains two or more target weight values using a user's most recent weight measurement, and obtains and outputs a predicted value for each of the two or more target weight values.

In addition, in this embodiment, we will describe a prediction device that uses learning information corresponding to two or more types of objective variables to obtain and output two or more types of predicted values.

In this embodiment, we also describe a prediction device that selects learning information that matches the user's attribute values, and uses the selected learning information to obtain and output a predicted value.

In this embodiment, we also describe a prediction device that constructs and outputs a table that associates predicted values of one or more objective variables with each target weight value.

Furthermore, in this embodiment, a prediction device is described that outputs results based on the results of comparing predicted values for one or more objective variables with a predetermined threshold.

Figure 6 is a block diagram of the prediction device 2 in this embodiment. The prediction device 2 includes a storage unit 21, a processing unit 22, and an output unit 23. The storage unit 21 includes a learning information storage unit 211 and a reference value storage unit 212. The processing unit 22 includes a target acquisition unit 221, a prediction unit 222, and a configuration unit 223. The output unit 23 includes a prediction output unit 231.

The storage unit 21 stores various types of information. This information includes, for example, learning information and reference values, which will be described later.

The learning information storage unit 211 stores one or more pieces of learning information. The learning information may be associated with a type identifier that identifies the type of each of the two or more predicted values to be output. The learning information may also be associated with an attribute value condition that is a condition using one or more attribute values. The learning information is information acquired by the learning device 1, and may be, for example, an arithmetic formula, a learner, or a correspondence table.

The reference value storage unit 212 stores reference values for one or more other types of test values. It is preferable that the reference values correspond to type identifiers. Type identifiers include, for example, "systolic blood pressure," "diastolic blood pressure," "triglycerides," "HDL cholesterol," "LDL cholesterol," "fasting blood glucose level," "HbA1C," and "presence or absence of urinary sugar." The reference value is, for example, a threshold value between normal and abnormal values.

The processing unit 22 performs various types of processing. These types of processing are, for example, processing performed by the target acquisition unit 221, prediction unit 222, and configuration unit 223.

The goal acquisition unit 221 acquires two or more future target weight values for a single user. The goal acquisition unit 221 acquires, for example, two or more target weight values input by the user.

The goal acquisition unit 221 acquires two or more target weight values, for example, using the most recent weight measurement value of a single user. Note that the most recent weight measurement value of a single user is stored, for example, in the storage unit 21. Furthermore, the most recent weight measurement value of a single user is, for example, information input by the user and acquired by the goal acquisition unit 221.

The target acquisition unit 221, for example, acquires a value obtained by subtracting two or more predetermined values from a user's most recent weight measurement. For example, the target acquisition unit 221 acquires a user's most recent weight measurement (X) and acquires target weight values "X," "X-2 kg," "X-4 kg," "X-6 kg," "X-8 kg," and "X-10 kg." It is preferable that the target acquisition unit 221 acquires two or more target weight values by successively subtracting a certain amount (for example, 2 kg).

It is preferable that the target acquisition unit 221 acquires a different number of target weight values depending on, for example, a user's most recent weight measurement (X). In other words, it is preferable that the target acquisition unit 221 acquires a larger number of target weight values based on the difference between a user's most recent weight measurement (X) and their ideal weight (Y). The ideal weight (Y) is pre-stored in the storage unit 21, for example. In addition, in a situation where the storage unit 21 stores information indicating the correspondence between height and ideal weight (Y) (e.g., "ideal weight (Y) = height (cm) - 110", a correspondence table between height ranges and ideal weight (Y)), the target acquisition unit 221 acquires the ideal weight (Y) corresponding to the received user's height, calculates the difference from the received most recent weight measurement (X), and acquires N target weight values (N is a natural number greater than or equal to 1) between the most recent weight measurement (X) and the ideal weight (Y).

The prediction unit 222 uses the learning information to obtain predicted values for each type of one or more future test values corresponding to each of the two or more target weight values obtained by the goal acquisition unit 221.

For example, the prediction unit 222 obtains learning information corresponding to each of the two or more types of predicted values from the storage unit 21, and uses the learning information to obtain predicted values corresponding to each of the two or more target weight values.

The prediction unit 222, for example, acquires learning information from the storage unit 21 corresponding to attribute value conditions that are matched by one or more attribute values of a user, and uses the learning information to acquire one or more predicted values corresponding to two or more target weight values.

The processing of the prediction unit 222 according to the type of learning information will be described below.
(1) When the learning information is an arithmetic formula

The prediction unit 222 substitutes one or more target weight values acquired by the target acquisition unit 221 into an arithmetic formula, executes the arithmetic formula, and acquires a predicted value.

When obtaining multiple types of predicted values, the prediction unit 222 obtains the calculation formula corresponding to each type from the storage unit 21, substitutes one or more target weight values for each type into the obtained calculation formula, executes the calculation formula, and obtains the predicted value.

If the calculation formula uses explanatory variables other than the target weight value as input parameters, the prediction unit 222 obtains the values of the explanatory variables other than the target weight value, substitutes the values of the explanatory variables and the target weight value into the calculation formula, executes the calculation formula, and obtains a predicted value.

Note that, among the set of values of one or more explanatory variables, information excluding the target weight value is, for example, information stored in the storage unit 21 or information received from the user. The same applies hereinafter.
(2) When the learning information is a learning module

The prediction unit 222 acquires a set of one or more explanatory variable values including the target weight value for each of one or more target weight values acquired by the goal acquisition unit 221. Next, the prediction unit 222 applies the set of one or more explanatory variable values to a learning device for each of one or more target weight values, performs machine learning prediction processing, and acquires a predicted value.

When obtaining multiple types of predicted values, the prediction unit 222 obtains a learning device corresponding to each type from the storage unit 21 and obtains, for each type, a set of one or more explanatory variable values including one or more target body weight values. Next, the prediction unit 222 applies the set of one or more explanatory variable values to the learning device for each of the multiple types and for one or more target body weight values, performs machine learning prediction processing, and obtains predicted values.
(3) When the learning information is a correspondence table

The prediction unit 222, for example, acquires a set of values for one or more explanatory variables that includes one or more target weight values acquired by the target acquisition unit 221. Next, the prediction unit 222 searches the correspondence table for a vector that most closely resembles the set (vector) of values for one or more explanatory variables for each of the one or more target weight values. Next, the prediction unit 222 acquires from the correspondence table a test value (predicted value) that pairs with the most closely resembled vector.

The prediction unit 222 acquires, for example, a set of values for one or more explanatory variables that includes one or more target weight values acquired by the target acquisition unit 221. Next, for each of one or more target weight values, the prediction unit 222 searches a correspondence table for one or more vectors that match the set (vector) of values for one or more explanatory variables and the approximation conditions (e.g., distance within a threshold, distance within the top N). Next, the prediction unit 222 acquires from the correspondence table test values (predicted values) that pair with each of the one or more vectors. Next, the prediction unit 222 calculates a weighted average of the one or more test values (predicted values) using the distance from the set (vector) of values for one or more explanatory variables to calculate a single predicted value. Note that the prediction unit 222 typically calculates a weighted average with a larger weight the shorter the distance.

When obtaining multiple types of predicted values, the prediction unit 222 uses a correspondence table corresponding to each type to obtain predicted values for each type and for one or more target weight values using the above-mentioned process.

The composition unit 223 composes output information that associates two or more target weight values with one or more types of predicted values acquired by the prediction unit 222. The output information may be, for example, a table or graph, and its format is not important.

The composition unit 223 composes output information according to the comparison result between the predicted value acquired by the prediction unit 222 and the reference value corresponding to the predicted value. It is preferable that the composition unit 223 composes output information in different output formats when the predicted value is favorable relative to the reference value (for example, when "predicted value <= reference value") and when the predicted value is not favorable relative to the reference value (for example, when "predicted value > reference value"). A different output format means, for example, different display attribute values for the predicted value, such as the font, font color, and background color of the predicted value. A different output format means that the target weight value and predicted value are not included in the output information when the predicted value is not favorable relative to the reference value, and the target weight value and predicted value are included in the output information when the predicted value is favorable relative to the reference value.

The output unit 23 outputs various types of information. Examples of such information include predicted values, target weight values, and output information. Here, output is a concept that includes displaying on a display, projecting using a projector, printing using a printer, outputting sound, transmitting to an external device, storing on a recording medium, and passing processing results to other processing devices or other programs. Note that transmission to an external device refers to transmission to a user's end-of-line device (not shown).

The prediction output unit 231 outputs one or more predicted values obtained by the prediction unit 222. For example, the prediction output unit 231 outputs a predicted value for each type of each of the two or more predicted values obtained by the prediction unit 222. For example, the prediction output unit 231 outputs output information constructed by the construction unit 223.

The storage unit 21, learning information storage unit 211, and reference value storage unit 212 are preferably non-volatile recording media, but can also be realized using volatile recording media.

The process by which information is stored in the storage unit 21, etc. is not important. For example, information may be stored in the storage unit 21, etc. via a recording medium, information transmitted via a communication line, etc. may be stored in the storage unit 21, etc., or information input via an input device may be stored in the storage unit 21, etc.

The processing unit 22, target acquisition unit 221, prediction unit 222, and configuration unit 223 can typically be realized using a processor, memory, etc. The processing procedures of the processing unit 22, etc. are typically realized using software, and this software is recorded on a recording medium such as a ROM. However, they may also be realized using hardware (dedicated circuitry). The processor may be, for example, a CPU, MPU, GPU, etc., and its type does not matter.

The output unit 23 and the prediction output unit 231 may or may not include output devices such as a display or speaker. The output unit 23, etc. may be realized by driver software for an output device, or by a combination of driver software for an output device and an output device, etc.

The output unit 23 and the prediction output unit 231 may be implemented using wireless or wired communication means. In such cases, the prediction output unit 231 and the like transmit information to the user's terminal device (not shown).

Next, an example of the operation of the prediction device 2 will be explained using the flowchart in Figure 7.

(Step S701) The goal acquisition unit 221 acquires the most recent weight of a user.

(Step S702) The goal acquisition unit 221 acquires one or more goal weights using the user's most recent weight. An example of this target weight acquisition process is described using the flowchart in Figure 8.

(Step S703) When obtaining a predicted value, the prediction unit 222 determines whether or not to use the user's attribute value. If the attribute value is to be used, the process proceeds to step S704; if the attribute value is not to be used, the process proceeds to step S705. Note that whether or not to use the attribute value may be determined in advance, or may be determined based on whether or not the user's attribute value has been accepted.

(Step S704) The prediction unit 222 obtains one or more attribute values of the user.

(Step S705) The prediction unit 222 assigns 1 to counter i.

(Step S706) The prediction unit 222 determines whether or not to acquire a predicted value for the i-th type of test value. If a predicted value for the i-th type of test value is to be acquired, the process proceeds to step S707; if not, the process proceeds to step S714. Note that, for example, the type for which a predicted value is to be acquired is determined in advance.

(Step S707) The prediction unit 222 acquires learning information corresponding to the i-th type from the learning information storage unit 211. Note that the prediction unit 222 may also acquire learning information corresponding to the i-th type and one or more attribute values of the user from the learning information storage unit 211.

(Step S708) The prediction unit 222 assigns 1 to counter j.

(Step S709) The prediction unit 222 determines whether the jth target weight exists among the target weights obtained in step S702. If the jth target weight exists, the process proceeds to step S710; if the jth target weight does not exist, the process proceeds to step S713.

(Step S710) The prediction unit 222 acquires the jth target weight and uses this target weight and the learning information acquired in step S707 to acquire a predicted value corresponding to the ith type and the jth target weight. Note that if the value of an explanatory variable other than the target weight is required when acquiring the predicted value, the prediction unit 222 acquires the value of that explanatory variable and also uses that value to acquire the predicted value.

(Step S711) The prediction unit 222 stores the predicted value acquired in step S710 in a buffer (not shown) in association with the i-th type and the j-th target weight. Note that here, the configuration unit 223 may acquire a reference value corresponding to the i-th type from the reference value storage unit 212, compare the reference value with the acquired predicted value, acquire the comparison result, and store it in association with the predicted value in a buffer (not shown).

(Step S712) The prediction unit 222 increments counter j by 1. Return to step S709.

(Step S713) The prediction unit 222 increments counter i by 1. Return to step S706.

(Step S714) The composition unit 223 composes output information using predicted values stored in a buffer (not shown). Note that the composition unit 223 may also compose output information using comparison results stored in a buffer (not shown).

(Step S715) The prediction output unit 231 outputs the output information constructed in step S714. Processing ends.

Next, an example of the target weight acquisition process in step S702 will be explained using the flowchart in Figure 7.

(Step S801) The target acquisition unit 221 acquires the user's height.

(Step S802) The goal acquisition unit 221 acquires the ideal weight corresponding to the height acquired in step S801.

(Step S803) The goal acquisition unit 221 determines whether "latest weight > ideal weight". If "latest weight > ideal weight", proceed to step S804; if not, proceed to step S805.

(Step S804) The target acquisition unit 221 subtracts "x kilograms (kg)" from the latest weight (X) until it reaches the ideal weight (Y), and acquires n target weights (n is a natural number greater than or equal to 1). The target acquisition unit 221 acquires, for example, "X," "X - x kilograms," "X - 2x kilograms," ... "X - nx kilograms," and "ideal weight (Y)" as the target weight.

(Step S805) The goal acquisition unit 221 acquires, for example, an ideal weight (Y) or the like as the target weight. Note that the ideal weight (Y) or the like may be just the ideal weight (Y), or it may be "X + x kilograms," "X + 2x kilograms," ... "ideal weight (Y)."

The following describes a specific example of the operation of the prediction device 2 in this embodiment.

Now, let's assume that the storage unit 21 of the prediction device 2 stores the results of a health checkup for 2019, including the user's most recent weight of "74.1". It is also preferable that the storage unit 21 also stores the results of the user's health checkups from 2018 and earlier.

The learning information storage unit 211 also stores learning information for each type of test value (test item). The test items here are "systolic blood pressure," "diastolic blood pressure," "triglycerides," "HDL cholesterol," "LDL cholesterol," "fasting blood glucose level," "HbA1C," and "presence or absence of urinary sugar."

Furthermore, it is assumed that the storage unit 21 stores reference values for each test item. The reference values here are of three types: "reference value," "health insurance guidance judgment value," and "visit recommendation judgment value." The "reference value" is a value within the normal range. The "health insurance guidance judgment value" is a value within a range in which guidance is required. The "visit recommendation judgment value" is a value within a range in which a visit to a medical institution is recommended.

Then, the goal acquisition unit 221 of the prediction device 2 first acquires one user's latest weight of "74.1". Next, the goal acquisition unit 221 acquires the latest weight of "74.1" and the weight loss values for each 2 kg up to a 10 kg loss: "74.1", "72.1", "70.1", "68.1", "66.1", and "64.1".

Next, using the learning information for each type of test value (test item) acquired by the learning device 1, the target acquisition unit 221 applies the weight value to the learning information for each of the six weights acquired for each test item to acquire a predicted value.

Next, the composition unit 223 composes output information that associates two or more target weight values with one or more types of predicted values acquired by the prediction unit 222. Furthermore, when composing the output information, the composition unit 223 acquires the results of a comparison between the predicted values acquired by the prediction unit 222 and the reference values corresponding to the predicted values (here, the "health insurance guidance judgment value" and the "visit recommendation judgment value"). Then, the composition unit 223 composes output information in a manner that makes the predicted values within the range of the "health insurance guidance judgment value" or "visit recommendation judgment value" and the 2019 measurement values stand out compared to other values.

Next, the prediction output unit 231 outputs the output information constructed by the construction unit 223. An example of such output is shown in Figure 9.

By looking at the table in Figure 9, users can be sure of their weight loss goals for the next year (in this case, 2020).

As described above, this embodiment makes it possible to present predicted test values for one or more target weight values.

Furthermore, according to this embodiment, predicted test values for two or more target weight values can be presented using information on the test value history.

Furthermore, this embodiment makes it possible to present predicted test values for two or more appropriate target weight values.

Furthermore, according to this embodiment, by using learning information corresponding to the acquired type of predicted value, it is possible to present two or more types of highly accurate predicted values.

Furthermore, according to this embodiment, highly accurate predictions can be presented by using learning information based on the user's attribute values.

Furthermore, according to this embodiment, the relationship between two or more target weight values and predicted values can be presented in an easy-to-understand manner.

Note that the prediction device 2 in this embodiment may also have the functions of the learning device 1. A block diagram of the prediction device 3 in this case is shown in Figure 10. The prediction device 3 includes a storage unit 31, a processing unit 32, and an output unit 23. The storage unit 31 includes a teacher data storage unit 111, a learning information storage unit 211, and a reference value storage unit 212. The processing unit 32 includes a learning information acquisition unit 12, a learning information accumulation unit 13, a target acquisition unit 221, a prediction unit 222, and a configuration unit 223.

The prediction device 2 may also function as a server, communicating with one or more terminal devices 4 and transmitting output information, etc., to the terminal devices 4. Figure 11 shows a block diagram of a prediction system A having a prediction device 2 and one or more terminal devices 4 in such a case.

The terminal device 4 comprises a terminal storage unit 41 that can store information, a terminal reception unit 42 that receives instructions and information, a terminal processing unit 43 that performs various processes, a terminal transmission unit 44 that transmits instructions and information to the prediction device 2, a terminal reception unit 45 that receives information, and a terminal output unit 46 that outputs information. Note that the terminal device 4 can be realized using a known device, so a detailed description will be omitted.

Also, for example, the terminal device 4 sends an instruction to the prediction device 2. The prediction device 2 then receives the instruction and executes the above-described processing in accordance with the instruction. The output unit 23 of the prediction device 2 then sends the output information, etc. acquired by the processing unit 22 to the terminal device 4. Next, the terminal device 4 receives and outputs the output information, etc. An example of such output is shown in Figure 9.

In FIG. 11, the prediction device 2 is a server, such as a cloud server or an ASP server, but the type does not matter. Furthermore, the terminal device 4 is a device used by a user, such as a PC, tablet terminal, or smartphone, but the type does not matter.

The processing in this embodiment may also be implemented using software. This software may be distributed by software download, etc. The software may also be recorded on a recording medium such as a CD-ROM and distributed. This also applies to other embodiments in this specification. The software that implements the prediction device 2 in this embodiment is the following program. In other words, this program causes a computer that can access a learning information storage unit that stores learning information acquired using two or more pieces of training data that associate a person's measured weight with one or more other types of test values to function as: a target acquisition unit that acquires two or more future target weight values for a user; a prediction unit that uses the learning information to acquire predicted values for each type of one or more future test values corresponding to each of the two or more target weight values; and a prediction output unit that outputs the one or more predicted values.

Furthermore, Figure 12 shows the external appearance of a computer that executes the programs described herein to realize the prediction devices and the like of the various embodiments described above. The above-described embodiments can be realized by computer hardware and computer programs executed on the computer hardware. Figure 12 is an overview diagram of this computer system 300, and Figure 13 is a block diagram of system 300.

In FIG. 12, computer system 300 includes computer 301, which includes a CD-ROM drive, keyboard 302, mouse 303, and monitor 304.

In FIG. 13, computer 301 includes, in addition to CD-ROM drive 3012, MPU 3013, bus 3014 connected to CD-ROM drive 3012 etc., ROM 3015 for storing programs such as boot-up programs, RAM 3016 connected to MPU 3013 for temporarily storing application program instructions and providing temporary storage space, and hard disk 3017 for storing application programs, system programs, and data. Although not shown here, computer 301 may also include a network card that provides connection to a LAN.

A program that causes the computer system 300 to execute the functions of the prediction device 2 and the like of the above-described embodiment may be stored on CD-ROM 3101, inserted into CD-ROM drive 3012, and then transferred to hard disk 3017. Alternatively, the program may be sent to computer 301 via a network (not shown) and stored on hard disk 3017. The program is loaded into RAM 3016 when executed. The program may also be loaded directly from CD-ROM 3101 or the network.

The program does not necessarily have to include an operating system (OS) or third-party programs that cause the computer 301 to execute functions such as the prediction device 2 of the above-described embodiment. The program need only include instructions that call appropriate functions (modules) in a controlled manner to achieve the desired results. How the computer system 300 operates is well known, and a detailed description will be omitted.

Note that in the above program, the steps of transmitting information and receiving information do not include processing performed by hardware, such as processing performed by a modem or interface card in the transmission step (processing that can only be performed by hardware).

Furthermore, the computer that executes the above program may be a single computer or multiple computers. In other words, it may perform centralized processing or distributed processing.

It goes without saying that in each of the above embodiments, two or more communication means present in one device may be physically realized using a single medium.

Furthermore, in each of the above embodiments, each process may be realized by centralized processing on a single device, or may be realized by distributed processing on multiple devices.

The present invention is not limited to the above-described embodiments, and various modifications are possible, and it goes without saying that these modifications are also included within the scope of the present invention.

As described above, the prediction device 2 of the present invention has the advantage of being able to present predicted test values for two or more target weight values, and is useful as a device to support health promotion, etc.

REFERENCE SIGNS LIST 1 Learning device 2, 3 Prediction device 4 Terminal device 11, 21, 31 Storage unit 12 Learning information acquisition unit 13 Learning information accumulation unit 22, 32 Processing unit 23 Output unit 111 Teacher data storage unit 211 Learning information storage unit 212 Reference value storage unit 221 Target acquisition unit 222 Prediction unit 223 Configuration unit 231 Prediction output unit

Claims (12)

a learning information storage unit that stores learning information acquired using two or more pieces of training data in which a weight measurement value of one person is associated with two or more other types of test values;
a goal acquisition unit that acquires two or more future goal weight values of one user;
a prediction unit that obtains predicted values of each type of one or more future test values corresponding to each of the two or more target weight values using the learning information;
a prediction output unit that outputs the one or more predicted values,
the importance of each of the other two or more types of test values is stored in association with an identifier of the other two or more types of test values;
The learning information is information obtained using two or more degenerate teacher data sets that are pairs of one or more test values corresponding to one or more types whose importance levels for each of the other two or more test value types satisfy the selection conditions and the weight measurement values. a learning information acquisition unit that acquires the importance of each type of the other two or more test values and stores the importance in association with each type's identifier;
The learning information acquisition unit
The prediction device of claim 1 uses the two or more teacher data to calculate, for each type of the other two or more test values, a correlation coefficient between the weight measurement value contained in each of the two or more teacher data and the two or more other types of test values contained in each of the two or more teacher data, and obtains an importance based on the correlation coefficient, or performs a learning process on the two or more teacher data using a random forest algorithm to obtain a learning device, and uses the learning device to obtain the importance of each type of the other two or more test values. A prediction device as described in claim 1 or claim 2, wherein the training data includes test values of the same type at two or more different time periods and weight measurements at each of the two or more time periods. The prediction device described in any one of claims 1 to 3, wherein the learning information is an arithmetic formula obtained by performing linear regression analysis on two or more training data. The target acquisition unit
The prediction device according to claim 1 , wherein the two or more target weight values are obtained using a most recent weight measurement value of the one user. The target acquisition unit
The prediction device according to claim 5 , wherein a value obtained by subtracting two or more predetermined values from the most recent weight measurement value of the one user is obtained. The teacher data is associated with one or more attribute values of the person,
the learning information storage unit stores learning information for each of two or more conditions using one or more attribute values;
The prediction unit
obtaining one or more predicted values corresponding to the two or more target weight values using learning information corresponding to a condition that one or more attribute values of the one user match;
The prediction device according to claim 1 , wherein the one or more attribute values include information related to lifestyle habits. The prediction device of claim 7, wherein the information regarding lifestyle habits is whether or not the person has a smoking habit, the level of smoking habit, whether or not the person has an exercise habit, the amount of exercise during a specified period, the level of exercise habit, whether or not the person has a drinking habit, or the level of drinking habit . a configuration unit that configures output information that associates each of the two or more target weight values with one or more types of predicted values acquired by the prediction unit;
The prediction output unit
The prediction device according to claim 1 , wherein the prediction device outputs the output information. further comprising a reference value storage unit in which reference values of the one or more other types of test values are stored;
a configuration unit that configures output information according to a comparison result between the predicted value acquired by the prediction unit and a reference value corresponding to the predicted value;
The prediction output unit
The prediction device according to claim 1 , wherein the prediction device outputs the output information. A prediction method for causing a computer to execute the process performed by the prediction device according to claim 1 . Computer,
A program for causing the prediction device according to any one of claims 1 to 10 to function.