WO2018101462A1 - Pregnancy time period forecasting device, pregnancy time period forecasting method, and pregnancy time period forecasting program - Google Patents

Abstract

The purpose of the present invention is to provide a pregnancy time period forecasting device, pregnancy time period forecasting method, and pregnancy time period forecasting program, whereby a time period to a pregnancy is forecast with a user's individual situation being accounted for. Provided is a pregnancy time period forecasting device, comprising: a feature vector computation unit 11 which, on the basis of first survey data which includes results of a survey which is carried out on a plurality of persons who have experienced pregnancy, computes a plurality of feature vectors which correspond respectively to the plurality of persons who have experienced pregnancy; a user feature vector computation unit 13 which, on the basis of second survey data which includes results of a survey which is carried out on a user, computes a user feature vector which corresponds to the user; an extraction unit 14 which extracts, from the plurality of feature vectors, one or a plurality of similar vectors which are similar to the user feature vector; and a forecasting unit 15 which, on the basis of a time period which one or a plurality of the persons who have experienced pregnancy who correspond to the one or the plurality of similar vectors required to become pregnant, computes a forecast value of the time period until the user becomes pregnant.

Description

Pregnancy period prediction device, pregnancy period prediction method, and pregnancy period prediction program Cross-reference of related applications

This application is based on Japanese Patent Application No. 2016-234468 filed on Dec. 1, 2016, the contents of which are incorporated herein by reference.

The present invention relates to a pregnancy period prediction device, a pregnancy period prediction method, and a pregnancy period prediction program.

In recent years, so-called “pregnancy activities” in which people who want to become pregnant actively work to become pregnant have been performed. Pregnancy may include review of meal contents and lifestyle, formation of exercise habits, and the like.
Regarding the prediction of the ovulation date that is important in pregnancy, Patent Document 1 describes the relationship between the interval between the menstrual date and the ovulation date and the average menstrual cycle estimated based on the data of a plurality of persons acquired in advance. An ovulation day prediction program for calculating predicted ovulation day data corresponding to a specific menstrual cycle is disclosed.

Patent No. 5998307

Conventionally, it has been said that the probability of pregnancy depends on the content of meals, lifestyle habits, and the existence of exercise habits. In some cases, a person who is pregnant is performing a review of meal contents and lifestyle, forming exercise habits, etc., according to a conventionally known rule of thumb.

However, the conventionally known rule of thumb is a general theory that seems to hold many people and does not reflect the circumstances of individuals. In addition, conventionally known rules of thumb qualitatively indicate behavior that seems to shorten the time required to become pregnant, and may not specifically or quantitatively indicate the time required to become pregnant. . For this reason, it may be difficult for a person who is pregnant to establish a prospect of pregnancy.

Therefore, an object of the present invention is to provide a pregnancy period prediction device, a pregnancy period prediction method, and a pregnancy period prediction program that predict a period until pregnancy is considered in consideration of the personal circumstances of a user.

The pregnancy period prediction device according to an aspect of the present invention calculates a plurality of feature vectors corresponding to each of a plurality of pregnant persons based on first questionnaire data including a result of a questionnaire conducted to a plurality of pregnant persons. A feature vector calculation unit, a user feature vector calculation unit that calculates a user feature vector corresponding to the user based on the second questionnaire data including a result of conducting a questionnaire to the user, and a plurality of feature vectors into a user feature vector A period until a user becomes pregnant based on an extraction unit that extracts one or more similar vectors and a period required until one or more pregnant persons corresponding to one or more similar vectors become pregnant A predicting unit that calculates a predicted value.

According to this aspect, it is possible to extract a pregnancy experienced person whose personal situation is similar to the user from a plurality of pregnancy experienced persons, and the user becomes pregnant based on the pregnancy period of the extracted pregnancy experienced person The predicted value for the period up to can be calculated. Therefore, the period until becoming pregnant can be predicted in consideration of the user's personal circumstances. By obtaining a quantitative prediction value, the user can easily set the outlook for pregnancy.

In the above aspect, when the content of the second questionnaire data is changed, a change calculation unit that calculates a change in a predicted value for a period until the user becomes pregnant, and a change in the predicted value calculated by the change calculation unit In addition, a determination unit that determines one or more questionnaire items to be improved among the plurality of questionnaire items included in the questionnaire may be further provided so that the predicted value becomes shorter.

According to this aspect, it is possible to clarify the points to be improved in order to shorten the period until pregnancy, and it is possible to give specific specific advice to the user.

In the above aspect, the first questionnaire data and the second questionnaire data include evaluation values respectively corresponding to a plurality of questionnaire items included in the questionnaire, and the determination unit includes a distribution of the evaluation values included in the first questionnaire data, Two or more questionnaire items to be improved may be determined based on the evaluation value included in the two questionnaire data.

According to this aspect, it is possible to extract items that are considered easy to improve for the user in view of personal circumstances, and to present an effort goal that can produce many results with little effort.

In the above aspect, the predicting unit is a cumulative distribution that indicates a probability that the user will become pregnant within an arbitrary period based on a period of time required for one or more pregnant persons corresponding to one or more similar vectors to become pregnant. The function may be calculated, and the change calculation unit may calculate a change in the cumulative distribution function when the content of the second questionnaire data is changed.

According to this aspect, the period until becoming pregnant can be stochastically shown, and the result when the effort goal is achieved can be quantitatively shown in the form of an increase in probability or a shortening of the period.

In the above aspect, the extraction unit extracts one or a plurality of similar vectors similar to the user feature vector from the plurality of feature vectors based on a norm of a difference vector between the user feature vector and any one of the plurality of feature vectors. May be.

According to this aspect, it is possible to extract a pregnancy experienced person whose personal situation is similar to that of the user by an arithmetic processing with a relatively light load.

In the above aspect, the dimension of each of the plurality of feature vectors and the dimension of the user feature vector may each be smaller than the number of items of the plurality of questionnaire items included in the questionnaire.

According to this aspect, it is possible to reduce the influence of the blurring of answers in the questionnaire, and to better capture the personal circumstances.

In the above aspect, the plurality of feature vectors each include one or more elements related to objective information of a plurality of pregnant persons and one or more elements related to subjective information of the plurality of pregnant persons. One or more elements related to the objective information of the user and one or more elements related to the subjective information of the user may be included.

According to this aspect, it is possible to capture the features of pregnant persons and users from both an objective viewpoint and a subjective viewpoint, and it is possible to better understand personal circumstances.

The pregnancy period prediction method according to an aspect of the present invention calculates a plurality of feature vectors corresponding to each of a plurality of pregnant persons based on first questionnaire data including a result of a questionnaire conducted to a plurality of pregnant persons. A step of calculating a user feature vector corresponding to the user based on second questionnaire data including a result of conducting a questionnaire to the user, and one or a plurality of similarities similar to the user feature vector from the plurality of feature vectors Extracting a vector; calculating a predicted value of a period until the user becomes pregnant based on a period required until one or more pregnant persons corresponding to one or more similar vectors become pregnant; and ,including.

According to this aspect, it is possible to extract a pregnancy experienced person whose personal situation is similar to the user from a plurality of pregnancy experienced persons, and the user becomes pregnant based on the pregnancy period of the extracted pregnancy experienced person The predicted value for the period up to can be calculated. Therefore, the period until becoming pregnant can be predicted in consideration of the user's personal circumstances.

The gestation period prediction program according to one aspect of the present invention is based on first questionnaire data including a result of a questionnaire conducted on a plurality of pregnancy experienced persons, and a plurality of feature vectors corresponding to each of the plurality of pregnancy experienced persons. A feature vector calculation unit that calculates a user feature vector, a user feature vector calculation unit that calculates a user feature vector corresponding to the user based on second questionnaire data including a result of a questionnaire conducted to the user, and a plurality of feature vectors. The user becomes pregnant based on an extraction unit that extracts one or a plurality of similar vectors similar to the feature vector and a period of time required for one or more pregnant persons corresponding to the one or more similar vectors to become pregnant And function as a prediction unit that calculates a predicted value for the period up to.

According to this aspect, it is possible to extract a pregnancy experienced person whose personal situation is similar to the user from a plurality of pregnancy experienced persons, and the user becomes pregnant based on the pregnancy period of the extracted pregnancy experienced person The predicted value for the period up to can be calculated. Therefore, the period until becoming pregnant can be predicted in consideration of the user's personal circumstances.

It is a system configuration diagram of a pregnancy period prediction device according to an embodiment of the present invention. It is a functional block diagram of the pregnancy period prediction apparatus concerning the embodiment of the present invention. It is a figure which shows the content of the 1st questionnaire data stored in a 1st questionnaire data database. It is a flowchart which shows the feature vector calculation process performed by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a flowchart which shows the user feature vector calculation process performed by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a flowchart which shows the similar vector extraction process performed by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a flowchart which shows the cumulative distribution function calculation process performed by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a figure which shows the pregnancy period prediction result output by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a flowchart which shows the improvement item determination process performed by the pregnancy period prediction apparatus which concerns on embodiment of this invention. It is a figure which shows the improvement advice output by the pregnancy period prediction apparatus which concerns on embodiment of this invention.

Embodiments of the present invention will be described with reference to the accompanying drawings. In addition, in each figure, what attached | subjected the same code | symbol has the same or similar structure.

FIG. 1 is a system configuration diagram of a pregnancy period prediction apparatus 10 according to an embodiment of the present invention. The pregnancy period prediction device 10 according to the present embodiment is connected to the user terminal device 20 and the first questionnaire data database DB by a communication network NW. The user terminal device 20 is a device used by a user of the pregnancy period prediction device 10, and is, for example, a personal computer or a smartphone. The communication network NW is a network that connects the pregnancy period prediction device 10, the user terminal device 20, and the first questionnaire data database DB, and is, for example, the Internet. The first questionnaire data database DB stores first questionnaire data including results obtained by conducting a questionnaire on a plurality of pregnant persons. The contents of the questionnaire and the contents of the first questionnaire data will be described in detail with reference to FIG. A user who tries to become pregnant operates the user terminal device 20 to transmit his / her questionnaire response result to the pregnancy period prediction apparatus 10 via the communication network NW, and predicts the pregnancy period output from the pregnancy period prediction apparatus 10. The value is confirmed by the user terminal device 20.

In this specification, a person who has experienced pregnancy means a person who has experienced natural pregnancy in the past. However, those who have experienced pregnancy may include those who have experienced pregnancy through artificial or in vitro insemination. Moreover, you may classify a pregnancy experience person into a natural pregnancy experience person, a pregnancy experience person by artificial insemination, or a pregnancy experience person by in vitro insemination. Those who desire natural pregnancy but have not become pregnant in natural pregnancy may be classified as having experienced pregnancy after an infinite period of time required to become pregnant.

FIG. 2 is a functional block diagram of the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. The pregnancy period prediction device 10 includes a feature vector calculation unit 11, a second questionnaire data acquisition unit 12, a user feature vector calculation unit 13, an extraction unit 14, a prediction unit 15, a change calculation unit 16, and a determination unit 17. And comprising.

The feature vector calculation unit 11 calculates a plurality of feature vectors corresponding to each of the plurality of pregnant pregnant persons based on the first questionnaire data including the result of the questionnaire conducted on the plurality of pregnant pregnant persons. The feature vector calculation process will be described in detail with reference to FIG.

The second questionnaire data acquisition unit 12 acquires second questionnaire data including a result of questionnaire to the user from the user terminal device 20. Here, the questionnaire conducted with respect to the user includes the same content as the questionnaire conducted with respect to a plurality of pregnant persons, or at least includes the superimposed content. The user feature vector calculation unit 13 calculates a user feature vector corresponding to the user based on the second questionnaire data including the result of conducting a questionnaire to the user. The user feature vector calculation process will be described in detail with reference to FIG.

The extraction unit 14 extracts one or a plurality of similar vectors similar to the user feature vector from the plurality of feature vectors. The similar vector extraction process will be described in detail with reference to FIG.

The prediction unit 15 calculates a predicted value of a period until the user becomes pregnant based on a period required until one or more pregnant persons corresponding to one or more similar vectors become pregnant. More specifically, the predicting unit 15 calculates the probability that the user will become pregnant within an arbitrary period based on the period required until one or more pregnant persons corresponding to one or more similar vectors become pregnant. The cumulative distribution function shown is calculated. The cumulative distribution function calculation process will be described in detail with reference to FIG.

In this specification, the period required for a person who has become pregnant to become pregnant means the period required for becoming pregnant after being aware of pregnancy. In addition, awareness of pregnancy means conscious of basal body temperature and ovulation date and consciously trying to become pregnant. However, the definition of each term is not intended to be limited to these, and any definition can be adopted as long as no contradiction arises. Alternatively, the time required for pregnancy of a person who desires natural pregnancy but does not reach natural pregnancy may be added as infinite.

The change calculation unit 16 calculates the change in the predicted value of the period until the user becomes pregnant when the content of the second questionnaire data is changed. The determining unit 17 determines one or more questionnaire items to be improved among a plurality of questionnaire items included in the questionnaire so that the predicted value is shortened based on the change in the predicted value calculated by the change calculating unit 16. To do. The improvement item determination process will be described in detail with reference to FIG.

FIG. 3 is a diagram showing the contents of the first questionnaire data QD stored in the first questionnaire data database DB. The 1st questionnaire data QD contains the information regarding the questionnaire result performed with respect to the pregnancy experienced person. The first questionnaire data QD includes answer information corresponding to each of a plurality of questionnaire items included in the questionnaire. The answer information may be represented by characters and symbols, but the first questionnaire data of this example includes evaluation values corresponding to a plurality of questionnaire items included in the questionnaire as the answer information. Note that the first questionnaire data QD may include information related to the period of time required for pregnant persons to become pregnant.

A plurality of questionnaire items included in the questionnaire are classified into physical information QD1, exercise category QD2, meal category QD3, couple category QD4, stress category QD5, and life category QD6. The first questionnaire data QD includes three evaluation values for the physical information QD1, and includes 20 evaluation values for the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6. The physical information QD1, the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6 included in the first questionnaire data QD of this example are examples, and the first questionnaire data QD is other than these. Items may be included, or some of these may not be included. A plurality of questionnaire items included in the first questionnaire data QD can be arbitrarily set.

The physical information QD1 includes evaluation values for height, weight, and age. In this example, the height is “T [cm]”, the weight is “W [kg]”, and the age is “A” ( A year old). The body information QD1 is objective information regarding a person who has experienced pregnancy. Note that the weight and age in this example are values at the time of pregnancy, but if the values at the time of pregnancy are uncertain, values estimated by a pregnant person can also be used. In the first questionnaire data QD of this example, the case where three items are included in the physical information QD1 is shown, but any number of items may be included in the physical information QD1.

The exercise category QD2 includes evaluation values of the first item to the twentieth item. In this example, the evaluation value of the first item is “a1” and the evaluation value of the twentieth item is “a20”. Note that the evaluation values of the second item to the nineteenth item are not shown. The meal category QD3 includes evaluation values of the 21st item to the 40th item. In this example, the evaluation value of the 21st item is “a21”, and the evaluation value of the 40th item is “a40”. The couple category QD4 includes evaluation values of the 41st item to the 60th item. In this example, the evaluation value of the 41st item is “a41”, and the evaluation value of the 60th item is “a60”. The stress category QD5 includes evaluation values of the 61st item to the 80th item. In this example, the evaluation value of the 61st item is “a61”, and the evaluation value of the 80th item is “a80”. The life category QD6 includes evaluation values of the 81st item to the 100th item. In this example, the evaluation value of the 81st item is “a81”, and the evaluation value of the 100th item is “a100”. In the first questionnaire data QD of this example, 20 categories are shown in each category of exercise category QD2, meal category QD3, couple category QD4, stress category QD5 and life category QD6. An arbitrary number of items may be included, and the number of items included in each category may be different.

In this example, the evaluation values a1 to a100 are any integers from 1 to 5, respectively. That is, these evaluation values are values obtained by evaluating the habit and way of thinking of a person who has been pregnant for five levels with respect to the n-th question included in the questionnaire. For example, for a question of the nth item, an evaluation value of “5” is given if it is exactly the same, and an evaluation value of “4” is given if it is considered to be the same. If there is not, an evaluation value of “3” is given. If not, an evaluation value of “2” is given. If not, an evaluation value of “1” is given. . The method of giving the evaluation value is arbitrary, and may be assigned when “1” is considered to be exactly as it is, and may be assigned when “5” is not considered at all. An integer may be used, or a real number from 0 to 1 may be used.

FIG. 4 is a flowchart showing a feature vector calculation process executed by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. First, the feature vector calculation unit 11 acquires first questionnaire data QD from the first questionnaire data database DB (S10). And an average and dispersion | distribution are each calculated about the evaluation value contained in the physical information QD1 (S11). When the first questionnaire data QD of N pregnant women is stored in the first questionnaire data database DB, the average height is a value obtained by dividing the total height of N people by N, and the variance of the height is The square of the difference between the height of each person and the average height is summed for all and divided by N. It should be noted that as the variance of the height, a value (unbiased variance) obtained by summing up the square of the difference between the height of each person and the average height for all members and dividing by N−1 may be used. For weight and age, the average and variance are calculated as in the case of height.

The feature vector calculation unit 11 normalizes the evaluation value of the body information Q1 by subtracting the average from the evaluation value of each person and dividing by the square root of the variance, and obtains a three-dimensional vector w1 having elements relating to height, weight, and age. It calculates for every experienced person (S12). For example, if the average _height is μ _height [cm], the variance of _height is σ ² _height [cm ² ], and the height of a certain pregnant person is T [cm], the normalized height evaluation value is It is a dimensionless quantity of (T−μ _height ) / σ _height . Similarly, when the weight average is μ _weight [kg], the weight variance is σ ² _weight [kg ² ], and the weight of a certain pregnant woman is W [kg], the normalized weight evaluation value Is a dimensionless quantity of (W−μ _weight ) / σ _weight . In addition, when the average of _age is μ _age and the variance of _age is σ ² _age , when the _age of a certain pregnant person is A, the normalized evaluation value of age is (A−μ _age ) / σ _age is there. That is, the three-dimensional vector w1 relating to a pregnant person having a height T [cm], a weight W [kg], and an age A is w1 = ((T−μ _height ) / σ _height , (W−μ _weight ) / σ _weight , (A−μ _age ) / σ _age ).

The feature vector calculation unit 11 calculates the average and variance of the sum of the evaluation values for each category included in the first questionnaire data (S13). For example, when the value (a1 + a2 +... + A19 + a20) obtained by adding the evaluation value a20 of the first item to the evaluation value a20 of the twentieth item included in the exercise category QD2 is expressed as α _category1 , the average μ _category1 is the value of N pregnant women The total of α _category1 is divided by N, and the variance σ ² _category1 is a value obtained by summing the square of the difference between α _{category1 of} each person and the average μ _category1 and dividing by N. As the variance σ ² _category 1, a value (unbiased variance) obtained by summing the squares of the differences between the α _{category 1 of} each person and the average μ _category 1 for all and dividing by N−1 may be used. For the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6, the average and variance are calculated in the same manner as the exercise category QD2.

The feature vector calculation unit 11 normalizes the evaluation value of each category by subtracting the average from the evaluation value of each person and dividing by the square root of the variance, and obtains a five-dimensional vector w2 having elements relating to each category for each person who has experienced pregnancy. Calculate (S14). For example, for a certain pregnancy experienced person, the value obtained by adding the evaluation value a20 of the first item to the evaluation value a20 of the twentieth item included in the exercise category QD2 is α _category1 , the average of N people is μ _category1 , and the variance is σ ^2. _In the case of _category1 , the normalized evaluation value is (α _category1 −μ _category1 ) / σ _category1 . Similarly, for a certain pregnant person, the value obtained by adding the evaluation value a40 of the 21st item to the evaluation value a40 of the 40th item included in the meal category QD3 is α _category2 , the average of N people is μ _category2 , and the variance is σ ^In the case of ² _category ² , the normalized evaluation value is (α _{category 1} −μ _{category 1} ) / σ _{category 1} . Similarly, for a certain pregnant person, the value obtained by adding the evaluation value a60 of the 41st item to the evaluation value a60 of the 60th item included in the marital category QD4 is α _category3 , the average of N people is μ _category3 , and the variance is σ ^In the case of ² _category3 , the normalized evaluation value is (α _category3 −μ _category3 ) / σ _category3 . Similarly, for a certain pregnant person, the value obtained by adding the evaluation value a61 of the 61st item to the evaluation value a80 of the 80th item included in the stress category QD5 is α _category4 , the average of N people is μ _category4 , and the variance is σ ^In the case of ² _{category 4} , the normalized evaluation value is (α _{category 4} −μ _{category 4} ) / σ _{category 4} . Similarly, the value obtained by adding the evaluation value a100 of the 81st item to the evaluation value a100 of the 100th item included in the life category QD6 is α _category5 , the average of N people is μ _category5 , and the variance is σ. ^In the case of ² _category5 , the normalized evaluation value is (α _category5 −μ _category5 ) / σ _category5 . Then, 5-dimensional vector w2 related pregnancies experienced _{person, w2 = ((α category1 -μ} category1) / σ category1, (α category2 -μ category2) / σ category2, (α category3 -μ category3) / σ category3, ( α _category4 −μ _category4 ) / σ _category4 , (α _category5 −μ _category5 ) / σ _category5 ).

The feature vector calculation unit 11 combines the three-dimensional vector w1 and the five-dimensional vector w2 calculated for each pregnancy experienced person, and calculates the eight-dimensional feature vector v for each pregnancy experienced person (S15). Specifically, an eight-dimensional feature vector v is calculated by the direct sum of the three-dimensional vector w1 and the five-dimensional vector w2. The three-dimensional vector related to the physical information QD1 calculated for a certain pregnant person is w1 = ((T−μ _height ) / σ _height , (W−μ _weight ) / σ _weight , (A−μ _age ) / σ _age ) Yes, the five-dimensional vectors calculated for the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5 and the life category QD6 are w2 = ((α _category1 −μ _category1 ) / σ _category1 , (α _category2 −μ _category2 ) _{/ σ category2, (α category3 -μ} category3) / σ category3, (α category4 -μ category4) / σ category4, (α category5 -μ cate _gory5 ) / σ _category5 ), the feature vector v is v = ((T−μ _height ) / σ _height , (W−μ _weight ) / σ _weight , (A−μ _age ) / σ _age , (α _{category1 -μ category1) / σ category1,} (α category2 -μ category2) / σ category2, (α category3 -μ category3) / σ category3, (α category4 -μ category4) / σ category4, (α category5 -μ category5) / σ _category5 ).

In the above description, the feature vector calculation unit 11 calculates the average and variance of the evaluation values for all the pregnant persons and calculates the feature vector v. The feature vector calculation unit 11 includes a plurality of pregnancy persons. Classification may be made into groups, and the average or variance of evaluation values may be calculated for each group, and the feature vector v may be calculated for each group. For example, the feature vector calculation unit 11 classifies those who have experienced pregnancy into a first group whose period required to become pregnant is M months or less and a second group whose period required to become pregnant is longer than M months. Thus, the feature vector v may be calculated respectively.

FIG. 5 is a flowchart showing a user feature vector calculation process executed by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. First, the second questionnaire data acquisition unit 12 acquires second questionnaire data from the user terminal device 20 (S20). The second questionnaire data is acquired by the user using the user terminal device 20 from the evaluation values related to the height, weight and age included in the physical information QD1, and from the first item to the 20th item included in the exercise category QD2. Evaluation values of stages, evaluation values of 5 stages from 21st item to 40th item included in meal category QD3, evaluation values of 5 stages of 41st item to 60th item included in couple category QD4, and stress category 5 levels of evaluation values from 61st item to 80th item included in QD5 and 5 levels of evaluation values from 81st item to 100th item included in life category QD6 are included.

The user feature vector calculator 13 calculates the average _height and variance σ ² _{height of the height} included in the body information QD1 of the first questionnaire data, the average µ _height and variance σ ² _weight of the _weight , the average _age μ _age and variance of the _age Reference is made to σ ² _age (S21). The user feature vector calculation unit 13 normalizes the evaluation value of the physical information Q1 by subtracting the reference average from the user evaluation value and dividing by the square root of the reference variance, and has a three-dimensional element having elements related to height, weight, and age Vector q1 is calculated (S22). When the height of the _user is T _user , the weight is W _user , and the age is A _user , the three-dimensional vector q1 is q1 = ((T _user −μ _height ) / σ _height , (W _user −μ _weight ) / σ _weight , (A _user −μ _age ) / σ _age ).

Next, the user feature vector calculator 13 calculates the average μ _category1 and variance σ ² _category1 of the total evaluation values included in the exercise category QD2 of the first questionnaire data, and the average μ of the total evaluation values included in the meal category QD3. and _Category2 and variance sigma ² _Category2, the mean mu _Category3 and variance sigma ² _Category3 of the sum of evaluation values included in the couple category QD4, the mean mu _Category4 and variance sigma ² _Category4 of the sum of evaluation values included in the stress category QD5, Reference is made to the average μ _category5 and the variance σ ² _category5 of the sum of evaluation values included in the life category QD6 (S23). The user feature vector calculation unit 13 normalizes the evaluation value of each category by subtracting the reference average from the evaluation value of the user and dividing by the square root of the reference variance, and calculates a five-dimensional vector q2 having elements relating to each category. (S24). The sum of the evaluation value of the user of the motion category QD2 is the α _category1 ^user, a sum _α category2 ^user of the evaluation value of the meal category QD3, the sum of the evaluation value of the couple category QD4 is _α category3 ^user, stress category QD5 The sum of the evaluation values of α _category4 ^user and the sum of the evaluation values of life category QD6 is α _category6 ^user , the five-dimensional vector q2 is q2 = ((α _category1 ^user −μ _category1 ) / σ _category1 , ( ^{_{α category2 user -μ category2) / σ}} category2, (α category3 user -μ category3) / σ category3, (α category4 u ^{_er -μ category4)} / ^{σ category4,} a ^{_{(α category5 user -μ category5) /}} σ category5).

The user feature vector calculator 13 combines the three-dimensional vector q1 and the five-dimensional vector q2 calculated for the user to calculate an eight-dimensional user feature vector q (S25). Specifically, an 8-dimensional user feature vector q is calculated from the direct sum of the 3-dimensional vector q1 and the 5-dimensional vector q2. The user feature vector q is q = ((T _user −μ _height ) / σ _height , (W _user −μ _weight ) / σ _weight , (A _user −μ _age ) / σ _age , (α _category1 ^user −μ _cate ) 1 ^{_{/ σ category1, (α category2 user}} -μ category2) / σ category2, (α category3 user -μ category3) / σ category3, (α category4 user -μ category4) / σ category4, (α category5 user -μ category5) / σ _category5 ).

The dimension of each feature vector v corresponding to a plurality of pregnant persons and the dimension of a user feature vector corresponding to a user are smaller than the number of items of a plurality of questionnaire items included in each questionnaire. In the case of this example, the dimension of the feature vector v and the user feature vector q is 8 respectively, which is smaller than the number 103 of items of a plurality of questionnaire items included in the questionnaire. The exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6 each include 20 questionnaire items. The feature vector v and the user feature vector q are the exercise category QD2, the meal category QD3, and the couple category. It has one element each corresponding to QD4, stress category QD5, and life category QD6. In this way, by adopting representative values (total values in this example) for multiple evaluation values included in the same category, it is possible to reduce the impact of response blurring in the questionnaire, and to improve personal circumstances. I can catch it well.

The feature vector v corresponding to a plurality of pregnancy experienced persons includes one or more elements related to objective information of a plurality of pregnancy experienced persons and one or more elements related to subjective information of the plurality of pregnant experience persons. The user feature vector q includes one or more elements related to the objective information of the user and one or more elements related to the subjective information of the user. In the case of this example, the feature vector v and the user feature vector q include three elements of the three-dimensional vector w1 or q1 corresponding to the body information QD1 as one or a plurality of elements related to objective information. The feature vector v and the user feature vector q are five-dimensional vectors w2 or q2 corresponding to the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6 as one or more elements related to subjective information. The five elements are included. Thereby, the characteristics of pregnant persons and users can be grasped from both an objective viewpoint and a subjective viewpoint, and personal circumstances can be better understood. Note that the feature vector v exemplified in this specification is an 8-dimensional vector v obtained by synthesizing the 3-dimensional vector w1 and the 5-dimensional vector w2. However, the dimension of the feature vector v is not limited to 8, and the physical information QD1 is related to it. The dimension of the vector w1 may be increased or decreased according to the number of items included in the physical information QD1, and the dimension of the vector w2 regarding each category may be increased or decreased according to the number of categories. Similarly, the user feature vector q exemplified in the present specification is an eight-dimensional vector q obtained by combining the three-dimensional vector q1 and the five-dimensional vector q2. However, the dimension of the user feature vector q is not limited to eight, The dimension of the vector q1 related to the information QD1 may be increased or decreased according to the number of items included in the physical information QD1, and the dimension of the vector q2 related to each category may be increased or decreased according to the number of categories.

In addition, when the pregnancy pregnant persons are classified into a plurality of groups, the average or variance of evaluation values is calculated for each group, and the feature vector v is calculated for each group, the user feature vector calculation unit 13 The user feature vector q may be calculated for each group. For example, the user feature vector calculation unit 13 calculates the first user feature vector using the average and variance of the evaluation values of the first group whose period required to become pregnant is M months or less, and was required before becoming pregnant. The second user feature vector may be calculated using the average and variance of evaluation values of the second group whose period is longer than M months. Thereby, the predicted value of the pregnancy period when the user belongs to the first group and the predicted value when the user belongs to the second group are obtained, and the user makes a pregnancy plan by comparing a plurality of scenarios. be able to.

FIG. 6 is a flowchart showing similar vector extraction processing executed by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. The extraction unit 14 calculates a difference vector q−v between the user feature vector q and the feature vector v corresponding to a certain pregnancy experienced person (S30). Then, the norm || q−v || of the difference vector is calculated (S31). The symbol “||. ||” represents the norm of the vector. Here, the norm of the vector may be defined by the L ² norm, that is, the square root of the sum of squares of each element of the vector, but the L ^p norm (p ≧ 0) may be used, and a different weight for each element. A norm that gives may be used.

The extraction unit 14 calculates the norm || q−v || of the difference vector from the feature vector v corresponding to all pregnant persons with respect to the user feature vector q, and the K values in ascending order of the norm value. The vectors are extracted as similar vectors v1, v2,... VK (S32). Here, the value of K is an integer of 1 or more and is arbitrary, but may be determined so that the prediction accuracy of the predicted value of the period until pregnancy, which will be described in detail later, is increased. For example, when N feature vectors related to N pregnant women are obtained, the extraction unit 14 uses the top 20 to 40% feature vectors having a small norm of the difference vector from the user feature vector q as similar vectors. It may be extracted. In this case, the number K of similar vectors is K = 0.2N to 0.4N. As described above, by extracting a similar vector based on a vector norm, it is possible to extract a pregnancy experienced person whose personal situation is similar to that of the user by a relatively light-weight calculation process.

Note that the group to which the user belongs can also be estimated using the similar vector extracted by the extraction unit 14. For example, the feature vector v is calculated on the basis of the average and variance of the evaluation values of all pregnant persons, and the pregnant person is required to become pregnant with the first group whose period required to become pregnant is M months or less. The second period is longer than M months. Then, the user feature vector q is calculated based on the average and variance of the evaluation values of the whole pregnancy experienced person, the similarity between the feature vector v belonging to the first group and the user feature vector q, and the feature vector belonging to the second group By comparing v and the similarity of the user feature vector q, the group having the highest similarity can be estimated as the group to which the user belongs. Here, the comparison of the degrees of similarity, for example, by calculating the norm of the difference vector between the user feature vector q and the feature vector v, respectively, and the average value of the norm related to the first group and the average value of the norm related to the second group And a group having a small norm average value can be estimated as a group to which the user belongs. In addition, for example, a person who has experienced pregnancy is classified into a first group that has reached natural pregnancy and a second group that has not reached natural pregnancy but has become pregnant by artificial or in vitro insemination, and the user selects either group. Can also be estimated. Naturally, even when pregnant women are classified into three or more groups, the group to which the user belongs can be estimated by the same method.

FIG. 7 is a flowchart showing a cumulative distribution function calculation process executed by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. The prediction unit 15 acquires a period required until K pregnant women corresponding to the similar vectors v1, v2,... VK become pregnant (S40). The prediction unit 15 according to the present embodiment acquires a period required until the pregnant person becomes pregnant from the first questionnaire data database DB.

The prediction unit 15 generates a histogram of the period required for the K pregnant persons corresponding to the similar vector to become pregnant, and calculates the probability distribution of the pregnancy period from the histogram by kernel density estimation (S41). Here, the kernel function used for kernel density estimation may be a normal distribution, but other functions may be adopted. The bandwidth used for kernel density estimation may be a width corresponding to several months. Note that the prediction unit 15 may calculate the probability distribution of the pregnancy period from a period required until K pregnant persons become pregnant by a method other than the kernel density estimation.

The prediction unit 15 calculates a cumulative distribution function indicating the probability that the user will become pregnant within an arbitrary period from the probability distribution of the pregnancy period (S42). The prediction unit 15 can give the probability that the user will become pregnant within half a year, for example, by the calculated cumulative distribution function. In addition, the prediction unit 15 can give a predicted value of a period until the user becomes pregnant, for example, by obtaining a period in which the pregnancy probability is 80% from the cumulative distribution function.

According to the pregnancy period prediction apparatus 10 according to the present embodiment, by providing the extraction unit 14, it is possible to extract a pregnancy experienced person whose personal situation is similar to the user from a plurality of pregnancy experienced persons. In addition, the prediction unit 15 can calculate a predicted value of a period until the user becomes pregnant based on the extracted pregnancy period of the experienced pregnancy person. Therefore, the period until becoming pregnant can be predicted in consideration of the user's personal circumstances. By obtaining a quantitative prediction value, the user can easily set the outlook for pregnancy. For example, the user may review lifestyle habits for pregnancy, use predicted values as a material for determining whether or not to see a specialist, or set a period of concentration on pregnancy based on predicted values be able to.

FIG. 8 is a diagram illustrating a pregnancy period prediction result output by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. In the figure, an example in which a result display screen DP including a pregnancy period prediction result is displayed on the user terminal device 20 is shown.

The result display screen DP includes a prediction result DP1, an advice display DP2, and a first graph DP3. The prediction result DP1 is a prediction result of the pregnancy period, the first message “the probability of becoming pregnant within 6 months is 10%”, the second message “the probability of becoming pregnant within one year is 45%”, and And a third message that “the probability of becoming pregnant within one year and six months is 90%”. The first message, the second message, and the third message indicate the relationship between the period and the pregnancy probability calculated based on the cumulative distribution function calculated by the prediction unit 15. The prediction result DP1 quantitatively shows the relationship between the period required to become pregnant and the pregnancy probability, and the user can make a prediction of pregnancy life with reference to the prediction result DP1. The contents of the first message, the second message, and the third message shown in FIG. According to the pregnancy period prediction device 10, it is possible to indicate the probability of becoming pregnant within an arbitrary period, and it is also possible to indicate the period during which a specific pregnancy probability is achieved.

The advice display DP2 is a button for displaying improvement points related to pregnancy. The user can obtain advice on how much the pregnancy probability is improved by changing the user's lifestyle etc. by pressing the advice display DP2. The contents of the advice will be described in detail with reference to FIG.

The first graph DP3 indicates the cumulative distribution function DF1 corresponding to the user, with the horizontal axis indicating the period and the vertical axis indicating the pregnancy probability. It can be seen from the first graph DP3 that the pregnancy probability within 6 months is 10%, the pregnancy probability within 1 year is 45%, and the pregnancy probability within 1 year 6 months is 90%. The first graph DP3 allows the user to visually grasp the relationship between the period and the pregnancy probability.

FIG. 9 is a flowchart showing improvement item determination processing executed by the pregnancy period prediction device 10 according to the embodiment of the present invention. The change calculation unit 16 calculates a change in the cumulative distribution function when the content of the second questionnaire data is changed. Specifically, the change calculation unit 16 changes the five-dimensional vector q2 out of the three-dimensional vector q1 and the five-dimensional vector q2 constituting the user feature vector q to generate an eight-dimensional trial vector t (S50). That is, the change calculation unit 16 increases or decreases the elements of the user feature vector q corresponding to the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6 among the plurality of questionnaire items included in the questionnaire. To generate a trial vector t. But the change calculation part 16 may produce | generate the trial vector t by increasing / decreasing the element of the user characteristic vector q corresponding to the weight and age contained in the physical information QD1. Thus, increasing or decreasing the element of the user feature vector q corresponds to changing the contents of the second questionnaire data.

The change calculation unit 16 calculates a cumulative distribution function based on the 8-dimensional trial vector t (S51). More specifically, K similar vectors similar to the 8-dimensional trial vector t are extracted from a plurality of feature vectors v, and based on a period required for pregnancy of a pregnant person corresponding to the K similar vectors. A histogram is created, a probability distribution is calculated from the histogram by kernel density estimation, and a cumulative distribution function is calculated from the probability distribution.

The change calculation unit 16 calculates a cumulative distribution function for a plurality of trial vectors t, and the determination unit 17 selects a trial vector t that minimizes the predicted value of the pregnancy period (S52). For example, the change calculation unit 16 increases (or decreases) the elements of the user feature vector q corresponding to the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6, respectively, and increases (or decreases) 5) are generated so that the sum of the amount) becomes a predetermined value, a cumulative distribution function is calculated, and the determination unit 17 has the highest probability of becoming pregnant within a predetermined period (for example, one year). A trial vector t may be selected. Further, for example, the change calculation unit 16 increases two elements of the user feature vector q corresponding to two categories selected from the exercise category QD2, the meal category QD3, the couple category QD4, the stress category QD5, and the life category QD6 ( Or 10) to generate a trial distribution vector t so that the sum of the increase amounts (or decrease amounts) becomes a predetermined value, calculate a cumulative distribution function, and the determination unit 17 determines a predetermined period (for example, 1 The trial vector t having the highest probability of becoming pregnant within a year may be selected.

The determination unit 17 determines one or a plurality of questionnaire items to be improved in order to realize the selected trial vector t among a plurality of questionnaire items included in the questionnaire. For example, when the cumulative distribution function is best improved by the trial vector t in which the element of the user feature vector q corresponding to the meal category QD3 is changed, the determination unit 17 changes the 21st to 40th items included in the meal category QD3. Decide which of the items should be improved.

The determination unit 17 determines one or more questionnaire items to be improved based on the distribution of the evaluation values included in the first questionnaire data and the evaluation values included in the second questionnaire data (S53). Specifically, for each questionnaire item that is a candidate for improvement, the determination unit 17 determines the variance σ ² _n of the evaluation values of the questionnaire items included in the first questionnaire data and the questionnaire items included in the second questionnaire data. The one or more questionnaire items to be improved are determined in descending order of the value of (max−an) σ _n based on the evaluation points an and the maximum score max of the evaluation points. By selecting items to be improved in this way, it is possible to select items that have sufficient margin for growth and have large variations among those who have experienced pregnancy as improvement items, and it is easy for users to improve. Items can be extracted in consideration of personal circumstances, and it is possible to present effort targets that can produce many results with little effort.

The change calculation unit 16 generates a trial vector t for all combinations of changes (or representative combinations of changes) in which the total amount of change of the elements of the user feature vector q is a predetermined value, and calculates the cumulative distribution function. After calculating, the determination unit 17 may select the trial vector t that provides the shortest predicted value of the pregnancy period. In general, when the number of elements whose values are increased or decreased by the change calculation unit 16 among the elements of the user feature vector q is n, and the elements of the user feature vector q are allowed to overlap, and the total p points are allocated one point at a time, The number of all combinations of changes is the number of overlapping combinations that select p out of _n , and is _{(n + p−1)} C _p . The case of this example will be described in detail. The number of elements whose values are increased or decreased by the change calculation unit 16 among the elements of the user feature vector q is 5 which is the number of categories, and the elements of the user feature vector q Assuming that the total change amount is 3 points, ₇ C ₃ = 35 trial vectors t are generated.

FIG. 10 is a diagram showing improvement advice output by the pregnancy period prediction apparatus 10 according to the embodiment of the present invention. In the figure, an example in which a result display screen DP including improvement advice is displayed on the user terminal device 20 is shown.

The result display screen DP includes an advice DP4, an improvement example DP5, and a second graph DP6. The advice DP4 is an improvement advice obtained based on the trial vector t in which the cumulative distribution function is most improved, and “the improvement in the diet category and the stress category increases the probability of becoming pregnant within one year by 5%.” Contains a message. In this example, the determination unit 17 selects a trial vector t in which two elements of the user feature vector q corresponding to the meal category QD3 and the stress category QD5 are changed, and the cumulative distribution calculated based on the trial vector t. The case where the function indicates that the probability of becoming pregnant within one year increases by 5% is illustrated. However, the pregnancy period prediction apparatus 10 selects a trial vector t in which one element of the user feature vector q corresponding to one category is changed, and shows a cumulative distribution function calculated based on the trial vector t. Alternatively, a trial vector t in which three or more elements of the user feature vector q corresponding to three or more categories are changed is selected, and a cumulative distribution function calculated based on the trial vector t is shown. Good. The pregnancy period prediction apparatus 10 may calculate a cumulative distribution function for a plurality of trial vectors t generated by a plurality of methods, and may select a trial vector t having the highest pregnancy probability when compared in the same period. In this example, the message indicates that the probability of becoming pregnant within one year increases, but the content of the message is not limited to this. For example, it may indicate that the period until a certain pregnancy probability is achieved is shortened, such as “the period during which the pregnancy probability is 50% is shortened from one year to one month”.

The improvement example DP5 shows an example of which of the questionnaire items included in the meal category QD3 and the stress category QD5 should be improved. The first evaluation value “the evaluation value of the 30th item of the meal category is +1”. An improvement example, a second improvement example “the evaluation value of the 35th item of the meal category is +2”, and a third improvement example of “the evaluation value of the 70th item of the stress category is +1” are included. In addition, in this example, when all the 1st improvement example, the 2nd improvement example, and the 3rd improvement example are achieved, it has shown that the raise of the pregnancy probability shown by advice DP4 is anticipated. The first improvement example of “+1 for the evaluation value of the 30th item of the meal category” represents that the meal content should be reviewed so that the evaluation score of the 5-level evaluation of the 30th item is increased by one. The second improvement example in which the evaluation value of the 35th item is +2 ”indicates that the meal contents should be reviewed so that the evaluation score of the 5-step evaluation of the 35th item is increased by two points. In addition, the third improvement example of “+1 for the evaluation value of the 70th item in the stress category” indicates that the stress situation should be reviewed so that the evaluation score for the 5-level evaluation of the 70th item is increased by one.

According to the pregnancy period prediction apparatus 10 according to the present embodiment, the user characteristic vector q is changed to search for a direction in which the pregnancy probability is most improved, and should be improved in order to shorten the period until pregnancy. The point can be clarified, and individual specific advice can be given to the user.

In addition, according to the pregnancy period prediction device 10 according to the present embodiment, by calculating the cumulative probability distribution from the result of the questionnaire to the user, the period until pregnancy can be stochastically shown, and the effort target can be determined. The outcome when achieved can be shown quantitatively in the form of increased probability or reduced duration. Thereby, the goal in pregnancy can be clarified, and the motivation to continue the pregnancy can be given by being aware of the expected result quantitatively.

The second graph DP6 shows the period on the horizontal axis, the pregnancy probability on the vertical axis, the cumulative distribution function DF1 corresponding to the user, the improved cumulative distribution function DF2 corresponding to the case where improvement advice is executed, Is shown. From the second graph DP6, it can be seen that the pregnancy probability within one year in the current state is 45%, but the pregnancy probability within one year after improvement is 50%. The second graph DP6 allows the user to visually grasp the effects when the lifestyle habits are reviewed according to the improvement advice.

The pregnancy period prediction device 10 stores the second questionnaire data including the result of the questionnaire for the user in the storage unit, and adds the questionnaire data regarding the user to the first questionnaire data database DB when the user becomes pregnant. May be. By accumulating questionnaire data, the data used for the prediction of pregnancy period becomes more abundant, and a more reliable prediction becomes possible.

By modifying the present invention, various prediction devices based on basic data having a plurality of items can be configured. As an invention obtained by modifying the present invention, for example, a power demand prediction device for predicting power demand can be cited. The power demand prediction apparatus calculates a plurality of feature vectors corresponding to the past date and time based on the environmental data regarding the past date and time. Here, the environmental data includes data such as weather, temperature, humidity, day of the week, and whether or not it falls on a public holiday. In the same manner as described in this specification, the plurality of feature vectors are obtained by calculating the average and variance for each item included in the environmental data related to the past date and time, and dividing the difference between the value of each item and the average by the square root of variance. You may calculate by calculating | requiring a value. Each dimension of the plurality of feature vectors may be equal to or less than the number of items included in the environment data. For example, when the environmental data includes four items of weather, temperature, humidity, and day of the week for the past date and time, the feature vector may be a four-dimensional vector. When the environment data includes N items, the feature vector may have N dimensions or less, and several items included in the environment data may be represented as one element of the feature vector. By combining several items included in the environmental data into one element of the feature vector, a feature vector in which the influence of fluctuations in the environmental data is reduced may be obtained.

Further, the power demand prediction apparatus calculates a feature vector for the current day based on the environmental data for the current day. As described in this specification, the feature vector of the current day is obtained by dividing the difference between the value of each item and the average of the environmental data related to the past date and time by the square root of the variance of the environmental data related to the past date and time. You may calculate by calculating | requiring. The dimension of the feature vector of the day may be equal to or less than the number of items included in the environmental data. The power demand prediction apparatus extracts one or a plurality of similar vectors similar to the current day feature vector from the plurality of feature vectors. The power demand prediction apparatus may extract a similar vector based on a norm of a difference vector between the current day feature vector and any one of a plurality of feature vectors. Similar vector, similar to that described herein, seeking L ² norm of the difference vector between any of the feature vectors and a plurality of feature vectors of the day, a plurality of feature vectors in the order value of the norm is small It may be K vectors extracted from the house. Incidentally, may be used norm other than L ² norm may be used a method of measuring the distance between other vectors.

The power demand prediction apparatus calculates a predicted value of the power demand on the current day based on the actual power demand on one or more dates and times corresponding to one or more similar vectors. The power demand prediction device generates a power demand histogram based on the actual power demand at one or more dates and times corresponding to one or more similar vectors, and calculates a probability distribution of the power demand from the histogram by kernel density estimation. A cumulative distribution function indicating the probability that the power demand falls below an arbitrary value may be calculated. Various parameters in the kernel density estimation can be arbitrarily selected as described in the present specification. According to the power demand prediction device, environmental data having a similar situation to that of the current day can be extracted from the past environmental data, and the power demand is predicted on the current day based on the actual power demand on the extracted date and time. A value can be calculated. Therefore, the power demand can be predicted in consideration of the environment of the day. Moreover, the power demand prediction apparatus can indicate the probability that the power demand on the current day falls below an arbitrary value, and can make a quantitative prediction regarding the power demand on the current day. For this reason, according to the power demand prediction apparatus, it is possible to obtain a quantitative estimate of how much power supply should be.

The power demand prediction device may calculate a change in the predicted value of the power demand for the day when the environmental data for the day changes. For example, with regard to data such as weather, temperature, and humidity included in environmental data, actual weather, temperature, and humidity data were obtained after calculating the predicted value of power demand for the day using values based on the weather forecast In this case, the predicted value of power demand can be updated sequentially. Specifically, in the same manner as described in the present specification, the power demand on the current day is calculated by calculating a trial vector in which the feature vector on the current day is changed and obtaining a cumulative distribution function of the power demand based on the trial vector. The change in the predicted value may be calculated. In addition, the power demand prediction apparatus can have the same configuration as the configuration of the pregnancy period prediction apparatus 10 described in this specification. Moreover, the electric power demand prediction method and electric power demand prediction program which estimate electric power demand can also be comprised by deform | transforming this invention.

By modifying the present invention, it is possible to configure a prediction device that calculates a predicted value at which a predetermined event will occur based on basic data having a plurality of items. The prediction device calculates a plurality of feature vectors having elements corresponding to a plurality of items included in the acquired basic data based on the acquired basic data. As described in this specification, the plurality of feature vectors are obtained by calculating an average and a variance for each item included in the acquired basic data, and dividing the difference between the value of each item and the average by the square root of the variance. You may calculate by calculating | requiring. The dimension of the plurality of feature vectors may be less than or equal to the number of items included in each acquired basic data.

The prediction device calculates a new feature vector based on the new basic data. The new feature vector corresponds to the user feature vector described in this specification, and the value obtained by dividing the difference between the value of each item and the average of the acquired basic data by the square root of the variance of the acquired basic data is obtained. It may be calculated. The dimension of the new feature vector may be equal to or less than the number of items included in the new basic data. The prediction device extracts one or a plurality of similar vectors similar to the new feature vector from the plurality of feature vectors. The prediction device may extract a similar vector based on a norm of a difference vector between the new feature vector and any one of the plurality of feature vectors. Similar vector, similar to that described herein, seeking L ² norm of the difference vector between any of the new feature vector and a plurality of feature vectors, among the plurality of feature vectors in the order value of the norm is small May be K vectors extracted from.

The prediction device calculates a predicted value of a new event based on the results of one or more events corresponding to one or more similar vectors. The prediction device generates a histogram indicating the frequency of occurrence of a new event based on the performance of one or more events corresponding to one or more similar vectors, and obtains a probability distribution of occurrence of new events by kernel density estimation from the histogram. A cumulative distribution function indicating the probability of occurrence of a new event may be calculated. Various parameters in the kernel density estimation can be arbitrarily selected as described in the present specification. According to the prediction device, it is possible to extract data having a similar situation to the new basic data from the acquired basic data, and based on the results of the event corresponding to the extracted data, the predicted value of the new event is obtained. Can be calculated.

The prediction device may calculate a change in the predicted value of the new event when the new basic data changes. The prediction device may calculate a change in the predicted value of the new event by calculating a trial vector obtained by changing the new feature vector and obtaining a cumulative distribution function of the event based on the trial vector. In addition, the prediction device can have the same configuration as the configuration of the pregnancy period prediction device 10 described in this specification. In addition, by modifying the present invention, it is possible to configure a prediction method and a prediction program for calculating a predicted value at which a predetermined event occurs based on basic data having a plurality of items.

The embodiment described above is for facilitating the understanding of the present invention, and is not intended to limit the present invention. Each element included in the embodiment and its arrangement, material, condition, shape, size, and the like are not limited to those illustrated, and can be changed as appropriate. In addition, the structures shown in different embodiments can be partially replaced or combined.

DESCRIPTION OF SYMBOLS 10 ... Pregnancy period prediction apparatus, 11 ... Feature vector calculation part, 12 ... 2nd questionnaire data acquisition part, 13 ... User feature vector calculation part, 14 ... Extraction part, 15 ... Prediction part, 16 ... Change calculation part, 17 ... Determination , 20 ... user terminal device, DB ... first questionnaire data database, DF1 ... cumulative distribution function, DF2 ... cumulative distribution function after improvement, DP ... result display screen, DP1 ... prediction result, DP2 ... advice display, DP3 ... first 1 graph, DP4 ... advice, DP5 ... improvement example, DP6 ... second graph, NW ... communication network, QD ... first questionnaire data, QD1 ... physical information, QD2 ... exercise category, QD3 ... meal category, QD4 ... couple category, QD5 ... stress category, QD6 ... life category.

Claims (9)

1. A feature vector calculation unit for calculating a plurality of feature vectors corresponding to each of the plurality of pregnancy experienced persons based on first questionnaire data including a result of conducting a questionnaire to a plurality of pregnant persons;
   A user feature vector calculation unit that calculates a user feature vector corresponding to the user based on second questionnaire data including a result of performing the questionnaire to the user;
   An extraction unit that extracts one or more similar vectors similar to the user feature vector from the plurality of feature vectors;
   A prediction unit that calculates a predicted value of a period until the user becomes pregnant based on a period required until one or more pregnant persons corresponding to the one or more similar vectors become pregnant;
   A pregnancy period prediction apparatus comprising:
2. A change calculating unit that calculates a change in a predicted value of a period until the user becomes pregnant when the content of the second questionnaire data is changed;
   Determination of determining one or more questionnaire items to be improved among a plurality of questionnaire items included in the questionnaire so that the predicted value is shortened based on a change in the predicted value calculated by the change calculation unit And
   The pregnancy period prediction apparatus according to claim 1, further comprising:
3. The first questionnaire data and the second questionnaire data include evaluation values respectively corresponding to a plurality of questionnaire items included in the questionnaire,
   The determination unit determines the one or more questionnaire items to be improved based on the distribution of the evaluation values included in the first questionnaire data and the evaluation values included in the second questionnaire data.
   The pregnancy period prediction apparatus according to claim 2.
4. The prediction unit is a cumulative distribution function that indicates a probability that the user will become pregnant within an arbitrary period based on a period of time required for one or more pregnant persons corresponding to the one or more similar vectors to become pregnant. To calculate
   The change calculation unit calculates a change in the cumulative distribution function when the content of the second questionnaire data is changed.
   The pregnancy period prediction apparatus according to claim 2 or 3.
5. The extraction unit extracts one or a plurality of similar vectors similar to the user feature vector from the plurality of feature vectors based on a norm of a difference vector between the user feature vector and one of the plurality of feature vectors. To
   The pregnancy period prediction apparatus of any one of Claim 1 to 4.
6. The dimensions of each of the plurality of feature vectors and the dimension of the user feature vector are smaller than the number of items of the plurality of questionnaire items included in the questionnaire,
   The pregnancy period prediction apparatus of any one of Claim 1 to 5.
7. The plurality of feature vectors each include one or more elements relating to objective information of the plurality of pregnant persons and one or more elements relating to subjective information of the plurality of pregnant persons,
   The user feature vector includes one or more elements related to objective information of the user and one or more elements related to subjective information of the user.
   The pregnancy period prediction apparatus of any one of Claim 1 to 6.
8. Calculating a plurality of feature vectors corresponding to each of the plurality of pregnancy experienced persons based on first questionnaire data including a result of conducting a questionnaire to a plurality of pregnancy experienced persons;
   Calculating a user feature vector corresponding to the user based on second questionnaire data including a result of conducting the questionnaire to the user;
   Extracting one or more similar vectors similar to the user feature vector from the plurality of feature vectors;
   Calculating a predicted value of a period until the user becomes pregnant based on a period required until one or more pregnant persons corresponding to the one or more similar vectors become pregnant;
   Pregnancy method including
9. Computer
   A feature vector calculation unit for calculating a plurality of feature vectors corresponding to each of the plurality of pregnancy experienced persons based on first questionnaire data including a result of conducting a questionnaire to a plurality of pregnant persons;
   A user feature vector calculation unit that calculates a user feature vector corresponding to the user based on second questionnaire data including a result of performing the questionnaire to the user;
   An extraction unit that extracts one or more similar vectors similar to the user feature vector from the plurality of feature vectors;
   A prediction unit that calculates a predicted value of a period until the user becomes pregnant based on a period required until one or more pregnant persons corresponding to the one or more similar vectors become pregnant;
   Pregnancy program to function as.

Images