PMS symptom prediction device, PMS symptom prediction method, learning device, learning method, and program

Abstract

To improve the accuracy of predictions regarding PMS symptoms. [Solution] A PMS symptom prediction device according to one aspect of the present invention comprises an acquisition unit that acquires user attribute information and heart rate information, and a processing unit that inputs the attribute information and heart rate information into a PMS symptom prediction model to make predictions about the user's PMS symptoms.

Description

Technical Field

[0001] The present invention relates to a PMS symptom prediction device, a PMS symptom prediction method, a learning device, a learning method, and a program.

Background Art

[0002] Conventionally, premenstrual syndrome (PMS), which is a physical and mental disorder before menstruation, has been known. In Patent Document 1, there is disclosed a comprehensive physical condition management system that acquires information on the start of menstruation and physical disorders, as well as weather data, calculates the probability of developing disorders due to the menstrual cycle, etc., and gives a prior warning against physical disorders.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, it has been difficult to accurately predict each symptom associated with menstruation for each individual only by using an individual's menstrual cycle and weather data. In this regard, as a result of intensive research, the inventor of the present case has found that PMS symptoms show a strong correlation with information regarding a predetermined heartbeat.

[0005] Therefore, an object of the present invention is to improve the accuracy of prediction regarding PMS symptoms.

Means for Solving the Problems

[0006] A PMS symptom prediction device according to an aspect of the present invention includes an acquisition unit that acquires attribute information of a user and heartbeat information regarding the heartbeat, and a processing unit that performs a prediction regarding the PMS symptoms of the user by inputting the attribute information and the heartbeat information into a PMS symptom prediction model.

[0007] According to this embodiment, it becomes possible to predict PMS symptoms based on user attribute information and heart rate information, thereby improving the accuracy of PMS symptom predictions. [Effects of the Invention]

[0008] According to the present invention, the accuracy of predictions regarding PMS symptoms can be improved. [Brief explanation of the drawing]

[0009] [Figure 1] This is a block diagram showing the schematic configuration of the PMS symptom prediction system 1 according to this embodiment. [Figure 2] This is a hardware configuration diagram of the computer 300 according to this embodiment. [Figure 3] This is an operation flow diagram showing an example of the operation process related to the generation of a PMS symptom prediction model according to this embodiment. [Figure 4] This is an operation flow diagram showing an example of the operation process related to the prediction process of PMS symptoms according to this embodiment. [Figure 5] This is a schematic diagram showing an example of a screen displayed on the display device 40 according to this embodiment. [Figure 6] This is a schematic diagram showing another example of the screen displayed on the display device 40 according to this embodiment. [Modes for carrying out the invention]

[0010] The embodiments will be described below with reference to the drawings. Note that the embodiments are illustrative and the present invention is not limited to the configurations described below.

[0011] (1) Functional configuration (1-1) PMS Symptom Prediction System 1 Figure 1 is a block diagram illustrating the schematic configuration of the PMS symptom prediction system 1 according to this embodiment. The PMS symptom prediction system 1 comprises a terminal device 10, a heart rate information measuring device 20, a PMS symptom prediction device 30A, a learning device 30B, and a display device 40.

[0012] In this disclosure, "User" means a person who uses the PMS symptom prediction system to make predictions about PMS symptoms. "Subject" means a person who, after following prescribed procedures and obtaining consent, provides attribute information, heart rate information, and PMS symptom scores as training data used in the PMS symptom prediction system.

[0013] (1-2) Terminal device 10 The terminal device 10 can be any information terminal that allows user attribute information to be input and that can output the input information to the PMS symptom prediction device 30A via a wired or wireless communication network. Examples include tablet devices, smartphones, wearable devices, and other mobile devices, or a PC (Personal Computer). Height and weight may be measured by the heart rate information measurement device 20 described later.

[0014] The terminal device 10 may also be configured to allow input of the user's PMS symptom score (a score that evaluates the severity of PMS symptoms from a predetermined perspective). The input user's PMS symptom score may be added to the training dataset of the training data storage unit 34, which will be described later. This makes it possible to make more personalized and accurate predictions regarding PMS symptoms.

[0015] (1-3) Heart rate information measuring device 20 The electrocardiogram information measuring device 20 can measure the electrocardiogram information of the user, and can be any information terminal as long as it can output the measured electrocardiogram information to the PMS symptom prediction device 30A through a wired or wireless communication network. The electrocardiogram information measuring device 20 may be, for example, a wearable terminal such as a smartwatch. Note that the electrocardiogram information measuring device 20 may be integrated with the terminal device 10. The electrocardiogram information measured by the electrocardiogram information measuring device 20 may be in the same data format as the electrocardiogram information included in the training data set described later (information based on the heart rate, information based on the heart rate variability such as RMSSD and SDNN, etc.). Alternatively, the electrocardiogram information measured by the electrocardiogram information measuring device 20 may be the log data of the electrocardiogram that is the basis of the electrocardiogram information. In this case, the PMS symptom prediction device 30A may process the log data acquired from the electrocardiogram information measuring device 20 into the same data format as the electrocardiogram information included in the training data set.

[0016] The timing at which the electrocardiogram information measuring device 20 acquires the electrocardiogram information of the user is not particularly limited. For example, it may be during sleep or during waking. The period during sleep may be a period divided at a predetermined interval (for example, 15 minutes, etc.), or a period divided at a specific stage according to the quality of sleep, etc. (for example, REM sleep / non-REM sleep, light sleep / deep sleep, etc.). The period during sleep may include a predetermined time before going to bed (for example, within 60 minutes, etc.) or a predetermined time after waking up (for example, within 60 minutes, etc.). Also, the timing of acquiring the electrocardiogram information may be when the subject is at rest. The method for determining whether the subject is at rest is not particularly limited. For example, a value related to the activity state of the subject measured by an acceleration sensor or the like provided in the terminal device 10 or the electrocardiogram information measuring device 20 may be used.

[0017] The timing when the heartbeat information measuring device 20 acquires the user's heartbeat information during the menstrual cycle is not particularly limited. For example, it may include a period belonging to at least any one of the menstrual period, follicular phase, ovulation phase, and luteal phase. The heartbeat information may be calculated based on values for a plurality of periods during the menstrual cycle. In particular, the heartbeat information may be, for example, the difference between a value based on a measurement in the luteal phase (e.g., an average value or median value in the luteal phase) and a value based on a measurement in another period (e.g., an average value or median value in the other period), or it may be the rate of change of the difference (a value obtained by dividing the difference by a predetermined reference value, etc.).

[0018] (1-4)PMS symptom prediction device 30A and learning device 30B The PMS symptom prediction device 30A includes a first acquisition unit 31, a second acquisition unit 32, a user data storage unit 33, a prediction processing unit 37, and a prediction data storage unit 38. The learning device 30B includes a training data storage unit 34, a learning processing unit 35, and a prediction model storage unit 36.

[0019] The first acquisition unit 31 acquires the user's attribute information from the terminal device 10. The attribute information is not particularly limited. For example, it may include the user's name, a predetermined ID, age or date of birth, height, weight, etc. Further, the attribute information may include data related to lifestyle habits. For example, it may include data related to drinking (frequency and amount of drinking, records, etc.) and data related to activities (including exercise) (frequency and amount of activities, records, etc.). Note that the data related to activities (including exercise) may include data measured by a wearable terminal or the like.

[0020] The second acquisition unit 32 acquires the user's heart rate information from the heart rate information measuring device 20. The period for which the user's heart rate information is acquired by the second acquisition unit 32 is not particularly limited, but for example, if the period from the start of menstruation to the start of the next menstruation is defined as the "menstrual cycle," then it may include some or all of the data from the menstrual cycle immediately preceding the current menstrual cycle (from the start date to the end date of that preceding menstrual cycle). Also, the period for which the user's heart rate information is acquired by the second acquisition unit 32 may include some or all of the data from the current menstrual cycle (from the start date of the current menstrual cycle to the present day). In particular, the period for which the user's heart rate information is acquired by the second acquisition unit 32 may include some or all of the data from the menstrual cycle immediately preceding the current menstrual cycle and some or all of the data from the current menstrual cycle. Here, the data for "part" of the menstrual cycle may be data from a predetermined set number of start days to a predetermined number of end days within the menstrual cycle, or it may be data for a specific period within the menstrual cycle (menstrual phase, follicular phase, ovulation phase, and luteal phase, etc.). The second acquisition unit 32 may acquire only the log data of the heart rate that forms the basis of the heart rate information from the heart rate information measuring device 20, and then perform a predetermined processing on the log data to acquire the user's heart rate information (information based on the heart rate as described above, or information based on heart rate variability such as RMSSD and SDNN, etc.).

[0021] The user data storage unit 33 stores the user's attribute information and heart rate information acquired from the first acquisition unit 31 and the second acquisition unit 32.

[0022] The training data storage unit 34 stores a training dataset for generating a prediction model of PMS symptoms. The training dataset will be described further later.

[0023] The learning processing unit 35 retrieves the training dataset stored in the training data storage unit 34 and generates a PMS symptom prediction model using a predetermined machine learning algorithm based on the training dataset.

[0024] The prediction model storage unit 36 ​​stores the PMS symptom prediction model generated by the learning processing unit 35. The prediction model will be described further later.

[0025] The prediction processing unit 37 uses the prediction model generated by the learning processing unit 35 to make predictions about the user's PMS symptoms based on the user's attribute information and heart rate information. Specifically, the prediction processing unit 37 inputs the user's attribute information and heart rate information into the prediction model generated by the learning processing unit 35, and obtains predetermined information about the user's PMS symptoms output from the prediction model.

[0026] The output of the prediction model acquired by the prediction processing unit 37 may be stored in the prediction data storage unit 38. The display device 40 can display the output of the prediction model acquired by the prediction processing unit 37 along with the user's attribute information and heart rate information. This data may also be displayed on the user's terminal device 10 or heart rate information measuring device 20, etc.

[0027] (2) Predictive models A predictive model for predicting PMS symptoms will be described. This predictive model may be generated, for example, using a training dataset stored in the training data storage unit 34 based on a predetermined machine learning algorithm by the learning processing unit 35.

[0028] (2-1) Training dataset The training dataset stored in the training data storage unit 34 may include, for example, attribute information of multiple subjects acquired in advance, heart rate information, and PMS symptom scores. The subjects may also include users. That is, user data (attribute information and / or heart rate information, etc.) stored in the user data storage unit 33 may be transferred to the training data storage unit 34, included in the training dataset, and then used for retraining the predictive model by the learning processing unit 35.

[0029] <Attribute information> Attribute information is not particularly limited, but may include, for example, the subject's name, designated ID, age and date of birth, height, weight, etc. Attribute information may also include data related to lifestyle habits, such as data on alcohol consumption (frequency, amount, and records of drinking) or activity data (frequency, amount, and records of activity). Note that data on activity (including exercise) may include data measured by wearable devices, etc.

[0030] <Heart rate information> Heart rate information may include, for example, information based on heart rate, or information based on heart rate variability. Heart rate information may be acquired, for example, by a wearable device such as a smartwatch.

[0031] Information based on heart rate may, for example, be the number of beats per minute.

[0032] Information based on heart rate variability may also be, for example, time-domain analytical indicators (statistical or geometric indicators). Such indicators may be used for purposes such as predicting prognosis for acute myocardial infarction or screening for autonomic nervous system disorders. Statistical indicators may include, for example, SDNN (standard deviation of NN intervals; ms), SDANN (standard deviation of mean NN intervals every 5 minutes over 24 hours; ms), RMSSD (mean square of the difference between consecutive NN intervals; ms), pNN50 (percentage of consecutive NN intervals with a difference of 50 ms or more; %), or CVRR (coefficient of variation of NN intervals). Geometric indicators may include, for example, HRVTI (triangle index).

[0033] Information based on heart rate variability may also be, for example, a frequency domain index. A frequency domain index may be, for example, the power of the high-frequency (0.15~0.45Hz) component; ms 2 ), LF (Power of low frequency (0.04~0.15Hz) component; ms 2), or LF / HF (the ratio of the power of the LF component to the HF component). These indices may be used for purposes such as evaluating autonomic nervous system function. In addition, indices in the frequency domain may include, for example, VLF (very low frequency (0.0033~0.04Hz) power) and ULF (extremely low frequency (<0.0033Hz) power; ms 2 These indicators may also be used for purposes such as predicting life expectancy in conditions like acute myocardial infarction.

[0034] Information based on heart rate variability may also be, for example, an index for nonlinear analysis. Examples of nonlinear analysis indices may be α1 (short-term (4-11 beats) scaling index by DFA), α2 (long-term (>11 beats) scaling index by DFA), β (spectral index by log-log spectrum), DC (heart rate deceleration ability by RPSA; ms), HRT (heart rate turbulence associated with ventricular premature contractions), or λ (non-Gaussian probability density function (PDF)). These indices may be used, for example, to predict life expectancy in conditions such as acute myocardial infarction.

[0035] Information based on heart rate variability may include, for example, periodic heart rate variability (CVR) indices. These CVR indices may include, for example, Fcv (frequency of CVHR occurrences; times / hour) or Acv (amplitude of CVHR; ms). For example, Fcv may be used for screening sleep apnea, and Acv may be used for predicting the prognosis of cardiovascular diseases.

[0036] The timing for acquiring the subject's heart rate information is not particularly limited, but may be, for example, during sleep or during wakefulness. The sleep period may be divided into periods at predetermined intervals (e.g., 15 minutes) or into periods divided into specific stages according to sleep quality (e.g., REM sleep / non-REM sleep, light sleep / deep sleep). The sleep period may include a predetermined time before going to sleep (e.g., within 60 minutes) or a predetermined time after waking up (e.g., within 60 minutes). Furthermore, the timing for acquiring heart rate information may be when the subject is at rest. The method for determining whether the subject is at rest or not is not particularly limited, but may be, for example, a value related to the subject's activity state measured by an acceleration sensor or the like provided in the terminal device 10 or the heart rate information measuring device 20.

[0037] The timing of acquiring a subject's heart rate information within the menstrual cycle is not particularly limited and may include, for example, periods belonging to at least one of the menstrual phase, follicular phase, ovulation phase, and luteal phase. The heart rate information may be calculated based on values ​​from multiple periods within the menstrual cycle. In particular, the heart rate information may be, for example, the difference between a value based on measurements during the luteal phase (e.g., the mean or median during the luteal phase) and a value based on measurements during other periods (e.g., the mean or median during those other periods), or it may be the rate of change of that difference (e.g., the difference divided by a predetermined reference value).

[0038] The inventor of the present invention has found that changes in heart rate or heart rate variability (the above-mentioned various indices as information based on heart rate variability) during the luteal phase are correlated with PMS symptoms. Therefore, the heart rate information may include, for example, information indicating such changes (changes in heart rate or heart rate variability during the luteal phase). The information indicating such changes may be quantitatively calculated, for example, by comparing the heart rate or heart rate variability belonging to each of a plurality of periods during the luteal phase. The plurality of periods may include, for example, a first period and a second period after the first period during the luteal phase. In this case, the heart rate information may be an index as a value obtained by subtracting the average value in the second period from the average value in the first period for the heart rate or heart rate variability. For example, when the first period is 4 to 6 days before menstruation and the second period is 1 to 3 days before menstruation, the heart rate information may be a value obtained by subtracting the average value 1 to 3 days before menstruation from the average value 4 to 6 days before menstruation for the heart rate or heart rate variability.

[0039] When the PMS symptoms are mild, the heart rate variability gradually increases and the heart rate decreases toward the start of menstruation. On the other hand, when the PMS symptoms are severe, the heart rate variability decreases and the heart rate increases during the period from around 3 days before the start of menstruation to the start of menstruation. Therefore, the value obtained by subtracting the average value of the heart rate 1 to 3 days before menstruation from the average value of the heart rate 4 to 6 days before menstruation tends to be a positive value when the PMS symptoms are mild, while it tends to be a negative value when the PMS symptoms are severe. Also, the value obtained by subtracting the average value of the heart rate variability 1 to 3 days before menstruation from the average value of the heart rate variability 4 to 6 days before menstruation tends to be a negative value when the PMS symptoms are mild, while it tends to be a positive value when the PMS symptoms are severe.

[0040] <PMS symptom score> The PMS symptom score is a score for evaluating the severity of PMS symptoms from a predetermined perspective and may be associated with time information such as date and time.

[0041] The method for obtaining a PMS symptom score is not particularly limited, but it may be obtained by a prescribed questionnaire. The questionnaire is not particularly limited, but may include questions similar to those in the MDQ (Menstrual Distress Questionnaire) (Moos, RH The development of a menstrual distress questionnaire. Psychosom Med. 30(6), 853-867 (1968)), as well as questions about appetite, sleep quality, etc. An example of a questionnaire is shown in the table below. The questionnaire may use, for example, at least some of the categories and answer choices shown in the table below.

[0042] [Table 1]

[0043] <Countermeasure Information> The training dataset stored in the training data storage unit 34 may include countermeasure information, which is information about measures to alleviate PMS symptoms. For example, the countermeasure information may include countermeasure information obtained by subjects in response to a prescribed questionnaire, or it may include data on lifestyle habits included in the subject's attribute information. The countermeasure information may also include information on countermeasures recommended by experts, etc. (which may be associated with the level and content of PMS symptoms).

[0044] The types of information provided regarding countermeasures are not particularly limited, but may include, for example, diet, sleep, activity (including exercise), and other lifestyle habits. Dietary countermeasures may include, for example, the intake of specific nutrients or the intake of foods rich in those nutrients. Specific nutrients may include, for example, calcium, magnesium, vitamin D, and vitamin B6. In particular, if activity levels are high but alcohol consumption is low, intake of tryptophan, magnesium, and calcium is considered desirable. Also, if there is no imbalance in activity levels or alcohol consumption and severe PMS symptoms are present, intake of tryptophan, magnesium, and vitamin B6 is considered desirable. Furthermore, dietary countermeasures may also include, for example, measures regarding alcohol consumption. For example, since PMS symptoms tend to be more severe with higher alcohol consumption, information regarding alcohol consumption may include reducing alcohol consumption (reducing opportunities to drink, or substituting with non-alcoholic beverages or carbonated water). Dietary countermeasures may also include eating small snacks frequently. This is because it is expected to suppress the worsening of PMS symptoms by preventing rapid fluctuations in blood sugar levels. In particular, it is recommended to consume nutritious snacks from time to time, such as nuts, yogurt, and fruit.

[0045] Activity-related measures may include increasing activity levels (exercise). Studies have shown that people with higher activity levels tend to experience milder PMS symptoms. Specifically, activity-related measures may include concrete advice on increasing activity levels in daily life, such as taking the stairs instead of the elevator or taking walks around the neighborhood.

[0046] (2-2) Generation of a predictive model The learning processing unit 35 acquires the training dataset stored in the training data storage unit 34 and generates a PMS symptom prediction model using a predetermined machine learning algorithm based on the training dataset. In particular, the learning processing unit 35 may train the prediction model to output predetermined information regarding the user's PMS symptoms in response to input of user attribute information and / or heart rate information. The attribute information input to the prediction model may include, for example, at least one item as shown in the <Attribute Information> section above. The heart rate information input to the prediction model may include, for example, at least one item as shown in the <Heart Rate Information> section above. The learning processing unit 35, for example, normalizes the training dataset and learns the relationship between attribute information and heart rate information and PMS symptoms obtained from questionnaires using machine learning methods such as neural networks (NN), gradient boosting such as XGBoost, linear regression, logistic regression, support vector machines, random forests, k-NN (k-nearest neighbors), RNN (recurrent neural network), and ensemble learning (Ridge regression). It then generates a predictive model that predicts PMS symptoms from heart rate information and attribute information. Note that the above-mentioned machine learning algorithms are examples and are not limited to these.

[0047] The predictive model may predict the level of a user's PMS symptoms (severity, nature, etc.). The predicted level of PMS symptoms may be associated with any point in time, both present and / or future, or may include the time-series changes in PMS symptoms over a predetermined period, both present and / or future. The dimensions (units) of the predicted level of PMS symptoms are not particularly limited and may be values ​​that quantitatively and / or qualitatively represent the level of PMS symptoms, or may include a specific description of the PMS symptoms.

[0048] The predictive model may predict events related to PMS symptoms. The predicted results of events related to PMS symptoms may include, for example, the start date, end date, and date on which the level of PMS symptoms changes (worsens, etc.). The start date of PMS symptoms may be, for example, the date on which the level of PMS symptoms exceeds a predetermined threshold value (first threshold value). The end date of PMS symptoms may be, for example, the date on which the level of PMS symptoms falls below a predetermined threshold value (second threshold value). Here, the first and second threshold values ​​may be equal or different. The date on which the level of PMS symptoms changes (worsens, etc.) may be, for example, the date on which the level of PMS symptoms exceeds a predetermined threshold value (third threshold value). Here, the third threshold value may be a value greater than the first and / or second threshold values.

[0049] The predictive model may include suggestions for measures to alleviate PMS symptoms. The types of suggested measures are not particularly limited, but may include, for example, diet, sleep, activity, and other activities. Dietary measures may include, for example, the intake of specific nutrients or the intake of foods rich in those nutrients. Activity measures may include increasing the amount of activity. The specific content of the measures may be in line with the content of the measures information included in the training dataset.

[0050] (2-3) Experiment The following shows the results of a predetermined experiment using the PMS symptom prediction model provided by the PMS symptom prediction system 1 according to this embodiment. In this experiment, 119 women participated in a study, and their heart rate information (sleep RMSSD, sleep SDNN, resting heart rate) during sleep and daytime was measured using a wearable device over a certain period. During the same period, a subjective questionnaire regarding PMS symptoms related to menstruation was administered, and a PMS symptom score was calculated. To evaluate the relationship between the acquired heart rate information and the PMS symptom score, a multivariate regression analysis was performed with age as a covariate and each heart rate information as an explanatory variable.

[0051] The overall results of this experiment revealed that RMSSD and SDNN were statistically significantly positively correlated with PMS symptom scores, while heart rate was statistically significantly negatively correlated with PMS symptom scores. As an example of the experimental results, the analysis results of three models are shown below.

[0052] [Table 2]

[0053] In Model 1, the relationship between age and luteal phase sleep RMSSD (mean) was evaluated as explanatory variables. The regression coefficient for age was -0.781, the t-value was -2.869, and the p-value was 0.005, which is statistically significant (p<0.05), suggesting that PMS symptoms tend to be milder with increasing age. The regression coefficient for luteal phase sleep RMSSD was +1.133, the t-value was 3.338, and the p-value was 0.001, which is also statistically significant, suggesting that PMS symptoms tend to be more severe with higher luteal phase sleep RMSSD.

[0054] In Model 2, the relationship between age and sleep-induced SDNN (mean) during the luteal phase was evaluated. The regression coefficient for age was -0.842, the t-value was -3.059, and the p-value was 0.003, which is statistically significant, suggesting that PMS symptoms tend to be milder with increasing age. The regression coefficient for sleep-induced SDNN during the luteal phase was +0.269, the t-value was 2.387, and the p-value was 0.018, which is also statistically significant, suggesting that PMS symptoms tend to be more severe with higher sleep-induced SDNN during the luteal phase.

[0055] In Model 3, the relationship between age and resting heart rate (mean) during the luteal phase was evaluated as explanatory variables. The regression coefficient for age was -1.016, the t-value was -3.712, and the p-value was 0.000, which is statistically significant, suggesting that PMS symptoms tend to be milder with increasing age. Similarly, the regression coefficient for resting heart rate during the luteal phase was -1.014, the t-value was -3.273, and the p-value was 0.001, which is also statistically significant, suggesting that PMS symptoms tend to be milder with higher resting heart rate during the luteal phase.

[0056] (3) Hardware configuration of the device Figure 2 is a hardware configuration diagram of the computer 300 according to this embodiment. The PMS symptom prediction device 30A may be composed of one or more computers 300. The learning device 30B may be composed of one or more computers 300. As shown in Figure 2, the computer 300 may include one or more processors 301, memory 302, storage 303, input / output ports 304, and communication ports 305. The processor 301 performs operation processing according to this embodiment by executing a program. In particular, the processor 301 in the PMS symptom prediction device 30A may perform operation processing related to PMS symptom prediction according to this embodiment by executing a program. The processor 301 in the learning device 30B may perform operation processing related to learning the prediction model according to this embodiment by executing a program. The memory 302 temporarily stores programs and the calculation results of programs. The storage 303 stores programs that perform predetermined operation processing. The storage 303 can be any type of storage medium that is readable by a computer. For example, various recording media such as magnetic disks, optical disks, random access memory, flash memory, and read-only memory can be used. The input / output port 304 receives information from other devices (such as the terminal device 10 and the heart rate information measurement device 20) and outputs information to other devices (display device 40). The communication port 305 transmits and receives data with other information terminals such as computers (not shown). Communication can be conducted via wireless or wired communication.

[0057] (4) Operation processing (4-1) Generation of a predictive model for PMS symptoms using machine learning Figure 3 is an operation flow diagram showing an example of the operation process related to the generation of a PMS symptom prediction model according to this embodiment.

[0058] In step ST101, the learning processing unit 35 performs preprocessing on the input data (for example, the training dataset described above). The learning processing unit 35 may, for example, perform data transformation or normalization depending on the data included in the training dataset. Data transformation may include, for example, transforming each subject's attribute information into a one-hot vector. Data normalization may include, for example, normalizing attribute information, heart rate information, and PMS symptom scores using Yeo-Johnson transformation or Boc-Box transformation. The learning processing unit 35 may also perform noise reduction on heart rate information based on attribute information. For example, since heart rate may show large outliers depending on the amount of alcohol consumed, the learning processing unit 35 may perform noise reduction on heart rate information included in the training dataset based on heart rate. Specifically, if the value or rate of change of heart rate is above a predetermined threshold, the learning processing unit 35 may remove heart rate information (heart rate and heart rate variability) based on that heart rate as noise from the training dataset. Alternatively, the learning processing unit 35 may remove heart rate information (heart rate and heart rate variability) corresponding to the drinking days (including information entered into an information processing device (terminal device 10, etc.)) from the training dataset as noise, based on the drinking day information self-reported by the subject.

[0059] In step ST102, the learning processing unit 35 generates a PMS symptom prediction model by performing machine learning using a pre-processed training dataset. The machine learning may include at least one of the following: neural networks (NN), gradient boosting such as XGBoost, linear regression, logistic regression, support vector machines, random forests, k-NN (k-nearest neighbors), RNN (recurrent neural networks), ensemble learning (Ridge regression), etc. The learning processing unit 35 stores the PMS symptom prediction model generated by the above learning process in the prediction model storage unit 36.

[0060] (4-2) Prediction of PMS symptoms using a PMS symptom prediction model Figure 4 is an operation flow diagram showing an example of the operation process related to the PMS symptom prediction process according to this embodiment.

[0061] In step ST201, the first acquisition unit 31 of the PMS symptom prediction device 30A acquires user attribute information from the terminal device 10. Specifically, the terminal device 10 may acquire user attribute information by displaying questions to the user asking for attribute information (age, date of birth, height, weight, drinking habits, etc.) and receiving answers from the user. The terminal device 10 may then transmit the received attribute information to the PMS symptom prediction device 30A. The first acquisition unit 31 may acquire the attribute information transmitted from the terminal device 10. The acquired attribute information is stored in the user data storage unit 33.

[0062] In step ST202, the second acquisition unit 32 of the PMS symptom prediction device 30A acquires the user's heart rate information for a predetermined period (for example, the menstrual cycle immediately preceding the current menstrual cycle and the current menstrual cycle). The heart rate information may include, for example, information based on heart rate (heart rate, etc.) or information based on heart rate variability (RMSSD or SDNN, etc.). Alternatively, the second acquisition unit 32 may acquire only the log data of the heart rate that forms the basis of the heart rate information from the heart rate information measurement device 20, and then perform a predetermined processing on the log data to acquire the user's heart rate information (information based on heart rate as described above, or information based on heart rate variability such as RMSSD or SDNN, etc.). The acquired heart rate information is stored in the user data storage unit 33.

[0063] In step ST203, the prediction processing unit 37 makes predictions about PMS symptoms using the PMS symptom prediction model stored in the prediction model storage unit 36. Specifically, the prediction processing unit 37 inputs the user's attribute information and heart rate information stored in the user data storage unit 33 into the prediction model, thereby obtaining predetermined information about the user's PMS symptoms output from the prediction model.

[0064] In step ST204, the prediction results for PMS symptoms (predetermined information regarding PMS symptoms output from the prediction model) are output to the prediction data storage unit 38 for storage, and the prediction results are also displayed on an external terminal such as the display device 40. The prediction results may also be displayed on a terminal device 10 or a heart rate information measurement device 20 owned by the user.

[0065] In step ST205, the PMS symptom score entered by the user is recorded in the training dataset stored in the training data storage unit 34. The recorded PMS symptom score may be used to retrain the prediction model by the learning processing unit 35. In addition, user data (attribute information and / or heart rate information, etc.) transferred from the user data storage unit 33 to the training data storage unit 34 may also be used to retrain the prediction model.

[0066] (5) Screen example (5-1) Screen displaying the prediction results Figure 5 is a schematic diagram showing an example of a screen displayed on the display device 40 according to this embodiment. Figure 5 shows screen G1 for displaying the prediction results of PMS symptoms. Screen G1 may be displayed on the display device 40 in step ST204 described above, for example.

[0067] Area G11 displays "Predicted PMS Onset Date," "Predicted Symptom Level Increase Date," and "Predicted PMS End Date" as examples of predicted events related to PMS symptoms. "Predicted PMS Onset Date" is the date on which the prediction model predicts the onset of PMS symptoms. "Predicted Symptom Level Increase Date" is the date on which the prediction model predicts the severity of PMS symptoms will increase. "Predicted PMS End Date" is the date on which the prediction model predicts the end of PMS symptoms.

[0068] In area G12, located below area G11, an example of a predicted PMS symptom level is displayed, showing "Lv2," which is the current predicted level of PMS symptoms. Note that the format of the predicted PMS symptom level shown is just one example, and the predicted results may be expressed in any other format.

[0069] At the bottom of area G12 are the menstruation start button G13 and the symptom recording start button G14. The menstruation start button G13 is for the user to record the start of their menstruation. When the menstruation start button G13 is pressed, the time information of the press may be added to the training dataset stored in the training data storage unit 34. The symptom recording start button G14 is for starting the recording of PMS symptom scores. The recording of PMS symptom scores may be done, for example, by a questionnaire format for answering the PMS symptom scores as described above. As a result, the recorded PMS symptom scores may be added to the training dataset stored in the training data storage unit 34 and used to retrain the prediction model. This enables more accurate predictions tailored to the individual characteristics of each user.

[0070] (5-2) Display screen for proposed countermeasures Figure 6 is a schematic diagram showing another example of a screen displayed on the display device 40 according to this embodiment. Figure 6 shows screen G2 for displaying suggestions for measures to alleviate PMS symptoms. Screen G2 may be displayed on the display device 40 in step ST204 described above, for example. Alternatively, screen G2 may be displayed on a terminal device 10 or heart rate information measuring device 20, etc., held by the user, in step ST204 described above.

[0071] Area G21 is an area for displaying a message indicating the result of a determination regarding the user's activity level, and may display text such as "It appears that your activity level is lower than the standard value." The prediction processing unit 37 may, for example, determine whether the activity level stored in the user data storage unit 33 is lower than a predetermined standard value, and display the determination result, such as the message in area G21, according to the determination. Note that the activity level may be data entered by the user or data measured by a wearable device, etc.

[0072] Area G22 is an area that displays specific advice regarding suggested measures to alleviate PMS symptoms. For example, it might display suggestions such as, "It might be a good idea to increase opportunities for physical activity, such as taking the stairs instead of the elevator!" Note that the suggested measures shown in the illustration are just examples, and other measures to alleviate PMS symptoms may also be included. Such other measures may include, for example, a message encouraging users who drink a lot of alcohol to reduce their alcohol intake, such as, "Reducing alcohol consumption may alleviate PMS symptoms," or dietary advice for users who would benefit from consuming certain nutrients (e.g., magnesium, vitamin B6), such as, "Try incorporating nuts and yogurt in moderation."

[0073] The embodiments described above are provided to facilitate understanding of the present invention and are not intended to limit its interpretation. The elements, arrangement, materials, conditions, shapes, and sizes of the embodiments are not limited to those exemplified and can be modified as appropriate. Furthermore, it is possible to partially substitute or combine the configurations shown in different embodiments. [Explanation of symbols]

[0074] 1…PMS symptom prediction system, 10…Terminal device, 20…Heart rate information measurement device, 30A…PMS symptom prediction device, 30B…Learning device, 31…First acquisition unit, 32…Second acquisition unit, 33…User data storage unit, 34…Training data storage unit, 35…Learning processing unit, 36…Prediction model storage unit, 37…Prediction processing unit, 38…Prediction data storage unit, 40…Display device, 300…Computer, 301…Processor, 302…Memory, 303…Storage, 304…Input / output port, 305…Communication port

Claims

1. An acquisition unit that acquires user attribute information and heart rate information related to heart rate during a predetermined period, A PMS symptom prediction device comprising: a processing unit that inputs the attribute information and heart rate information during the predetermined period into a PMS symptom prediction model to make a prediction regarding the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period; The PMS symptom prediction model is a PMS symptom prediction device generated by machine learning using a training dataset that includes attribute information of a subject for at least one menstrual cycle, heart rate information relating to the subject's heart rate for at least one menstrual cycle, and a score relating to the subject's PMS symptoms for at least one menstrual cycle, and accepts input of attribute information and heart rate information for a predetermined period, and outputs information relating to the level of PMS symptoms at a predetermined future point in time.

2. The PMS symptom prediction device according to claim 1, wherein the subject includes the user.

3. The PMS symptom prediction device according to claim 1, wherein the heart rate information includes information based on heart rate and / or information based on heart rate variability.

4. The PMS symptom prediction device according to claim 3, wherein the information based on the heart rate variability includes RMSSD and / or SDNN.

5. The PMS symptom prediction device according to claim 1, wherein the PMS symptom prediction model is further generated by machine learning using the training dataset to predict the date on which the user's PMS symptom level exceeds or falls below a predetermined threshold.

6. The training dataset further includes information on measures to alleviate PMS symptoms, The PMS symptom prediction model is generated using machine learning with the training dataset to further output suggestions for measures to alleviate PMS symptoms. The PMS symptom prediction device according to claim 5, wherein the processing unit inputs the attribute information and heart rate information during the predetermined period into the PMS symptom prediction model to further obtain suggestions for measures to alleviate PMS symptoms for the user.

7. One or more computers, To acquire user attribute information and heart rate information related to heart rate during a predetermined period, A PMS symptom prediction method that performs the following: inputting attribute information and heart rate information during the predetermined period into a PMS symptom prediction model to make a prediction about the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period, The PMS symptom prediction model is generated by machine learning using a training dataset that includes attribute information of a subject for at least one menstrual cycle, heart rate information relating to the subject's heart rate for at least one menstrual cycle, and a score relating to the subject's PMS symptoms for at least one menstrual cycle, and accepts input of attribute information and heart rate information for a predetermined period of time to output information relating to the level of PMS symptoms at a predetermined future point in time.

8. One or more computers, An acquisition unit that acquires user attribute information and heart rate information related to heart rate during a predetermined period, A program for functioning as a processing unit that inputs the attribute information and heart rate information for the predetermined period into a PMS symptom prediction model, thereby making a prediction about the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period, The PMS symptom prediction model is a program generated by machine learning using a training dataset that includes attribute information of a subject for at least one menstrual cycle, heart rate information of the subject for at least one menstrual cycle, and a score regarding the subject's PMS symptoms for at least one menstrual cycle, to accept input of attribute information and heart rate information for a predetermined period and output information regarding the level of PMS symptoms at a predetermined future point in time.

9. A learning device comprising: a learning unit that generates a PMS symptom prediction model by machine learning using a training dataset which includes attribute information of a subject for at least one menstrual cycle, heart rate information relating to the subject's heart rate for at least one menstrual cycle, and a score relating to the subject's PMS symptoms for at least one menstrual cycle, to receive input of attribute information of a user and heart rate information relating to their heart rate over a predetermined period, and to output information relating to the level of the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period.

10. The learning apparatus according to claim 9, wherein the subject includes the user.

11. The learning device according to claim 9, wherein the heart rate information includes information based on heart rate and / or information based on heart rate variability.

12. The learning device according to claim 11, wherein the information based on the heart rate variability includes RMSSD and / or SDNN.

13. The learning device according to claim 9, wherein the PMS symptom prediction model predicts the level of the user's PMS symptoms and / or events related to the user's PMS symptoms in response to the input of attribute information and heart rate information.

14. The learning device according to claim 13, wherein the PMS symptom prediction model further outputs suggestions for measures to alleviate PMS symptoms for the user in response to the input of attribute information and heart rate information.

15. One or more computers, A learning method that performs machine learning using a training dataset which includes attribute information of a subject for at least one menstrual cycle, heart rate information relating to the subject's heart rate for at least one menstrual cycle, and a score relating to the subject's PMS symptoms for at least one menstrual cycle, to generate a PMS symptom prediction model that accepts input of attribute information and heart rate information relating to a user's heart rate for a predetermined period and outputs information relating to the level of the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period.

16. One or more computers, A program to function as a learning unit that generates a PMS symptom prediction model by machine learning using a training dataset that includes attribute information of a subject for at least one menstrual cycle, heart rate information of the subject for at least one menstrual cycle, and a score regarding the subject's PMS symptoms for at least one menstrual cycle, and accepts input of attribute information and heart rate information of a user for a predetermined period, and outputs information regarding the level of the user's PMS symptoms at a predetermined point in time that is further in the future than the predetermined period.

Images