JP5091595B2 - Health index value presentation system and health index value presentation method - Google Patents

Description

  The present invention collects health data that requires input or measurement by a user, calculates a health index value from the accumulated data and the collected data, and presents the health index value presentation system to the user. The present invention relates to an index value presentation method.

  In order to perform appropriate health management, it is first necessary to collect sufficient health data necessary to grasp the health status. If there is a large amount of high-quality health data, it is possible to appropriately know the current health status using appropriate health index values, and by creating a prediction model, health indicators such as disease risk indicating the future health status It is possible to calculate the value and give appropriate health guidance.

  In addition, if there is data at an appropriate time, the timing of health guidance can be determined. For example, as a conventional technique for determining the timing of health guidance, there is known a system capable of adjusting the timing of health guidance by regularly monitoring health data obtained from a biological information measuring device or a terminal. (For example, Patent Document 1).

  Generally, health information necessary for grasping a health condition is collected by using a measuring device or answering an inquiry question. Therefore, in order to obtain necessary and sufficient health information, if there are many items that need to be measured or items that need to be answered, or if measurement and responses are frequently required, the burden on the user is heavy. There was a problem.

As a conventional technique for reducing such a burden on the user, a system is known in which diagnosis is performed with a small number of inquiry items by using an inference process based on an occurrence frequency and an answer pattern (for example, Patent Document 2). In addition, there is known an apparatus that increases the frequency of information acquisition when the measured value or the result of the inquiry answer is not good (for example, Patent Document 3).
JP 2006-235938 A JP 2003-242248 A JP 2005-11329 A

  However, in the prior art, since the relationship between a plurality of health information items is not considered, it is difficult to collect only an appropriate amount of health information having a complicated relationship.

  Therefore, in view of the problems of the prior art, the present invention takes into account the relationship between a plurality of health information items, reduces the burden on the user, and presents a health index value presentation system with high accuracy. It is another object of the present invention to provide a health index value presentation method.

A health index value presentation system according to the present invention includes a health data storage unit that stores health data, a plurality of questions and answers related to the health data, and a selection criterion for the stored questions. A question selection criterion input unit that inputs a question, a presentation question selection unit that selects a question to be presented to a user terminal based on the input selection criterion, and the question selected by the presentation question selection unit Obtained from the question / answer storage unit, a question output unit for outputting to the user terminal, and an answer input from the user terminal for the output question, and obtained in the question / answer storage unit an answer input section for storing the health data storage unit to the stored health data, questions previously stored Symbol question storage unit, based on the strength of relationship and the relationship between the plurality of questions Has a graph structure associated between the plurality of model components in impact, and to generate a model for calculating the health index value indicating a health condition, the calculated health index value or to other model components A model component having a degree of influence of 0 and a contracted model obtained by reducing the relationship between the model components from the model are generated together, and the answer input in the answer input unit is the model or the contracted model. a health index calculation unit for calculating the health index value fitted to, characterized in that the calculated health index value obtaining Bei and a health index value presenting unit for presenting to the user terminal.

In the health index value presentation method according to the present invention, a server computer calculates a health index value indicating a health state according to input information from a plurality of user terminals connected via a network, and A health index value presentation method for presenting, a health data storage step for storing health data in a health data storage device, and a question / answer storage for storing a plurality of questions and answers related to the health data in a question / answer storage device A step of inputting a selection criterion for the stored question from the user terminal, and a presentation question for selecting a question to be presented to the user terminal based on the input selection criterion A selection step; a question output step of acquiring the selected question from the question / answer storage device and outputting the same to the user terminal; and the output Get the answer inputted from the user terminal to the interrogation, the answers entered storing the question storage device, the health data Health data stored in the storage step, the question selection reference input Based on the question selection criteria input in the step, the health index value, the relationship between the plurality of questions and the strength of the relationship, a plurality of model components are associated with each other by the degree of influence, And a model generation step for generating a model for calculating the health index value, a model component whose influence on the calculated health index value or other model components is 0 from the generated model, and a reduced model generating step of generating a reduced model with a reduced relationship of the model components, inputted in the answer input step It has a health index value calculation step of calculating the health index value answer fit to the model or the reduced model, and a health index value presenting step for presenting the calculated health index value to said user terminal and wherein a call.

  According to the present invention, there are provided a health index value presentation system and a health index value presentation method that take into account the relationship between a plurality of health information items, reduce the burden on the user, and present a highly accurate health index value. Provided.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an overall configuration of a health index value presentation system 1 according to an embodiment of the present invention. The health index value presentation system 1 is a computer system composed of a server 10 and user terminals 20, and a plurality of user terminals 20 (20-1 to n) are connected to the server 10 via a network 30. Yes. Here, the “health index value” refers to an estimated value indicating a health condition. For example, the probability of being diagnosed with hypertension within 3 years is one example of a health index value. In addition, an index value indicating a mental state or a temporary physical condition is an example of a health index value.

  As shown in FIG. 1, the server 10 includes a health data storage unit 101, a question / answer storage unit 102, a question selection criterion input unit 103, a question presentation opportunity input unit 104, a presentation question selection unit 105, and a question output unit. 106, an answer input unit 107, a health index value calculation unit 108, and a health index value presentation unit 109.

  The health data storage unit 101 is a device that stores health data. “Health data” refers to any data relating to health. For example, measurement data such as body weight and blood pressure obtained in regular health examinations, response data in interviews with medical personnel, questionnaire response data related to health, measurement data by a measuring device such as a pedometer, and the like. In addition, basic data such as age and gender, and work-related data such as work status are also related to health because they are indirectly related to health.

  The question / answer storage unit 102 is a device that stores question / answer data. “Question / answer data” is data related to a question presented to the user and a corresponding answer. For example, a combination of a question sentence “Is it easy to get tired” and answer sentences “Yes”, “sometimes”, and “No” is question / answer data. The answer may be a numerical value depending on the degree. When an identifier is given to a question sentence or an answer sentence for facilitating data processing, the identifier is also included in the question / answer data.

  In addition, when questions are classified by category or the like, information on the category is also question / answer data. The questions presented do not necessarily have to be presented in text. For example, when the question “Is it easy to get tired” is associated with a specific alarm sound or LED flashing, the alarm sound or LED flashing is a question of question / answer data. Also, the answer corresponding to the question does not necessarily have to be answered as a sentence, and an answer number, an inspection value, an output of a measuring device, or the like can be used. For example, when a question “how many blood pressures” is given, the output value of the sphygmomanometer can be used as an answer. If there are many forms of answers to the question text and it is difficult to generate a model, preprocessing such as data normalization and discretization is appropriately performed.

  This question data becomes an explanatory variable of the model generated in the health index value calculation unit 108 described later. FIG. 2 is a diagram showing a specific example of question / answer data compiled in a table format. Here, a question identifier, a question, an abbreviation, and an answer format are shown as questions / answer data items.

  The question selection criterion input unit 103 is a program for inputting a question selection criterion when selecting a question to be output to the user terminal 20 from a question output unit 106 described later. The “question selection criterion” is a criterion used for selecting a question to be presented to the user. For example, it is the minimum accuracy that the health index value must satisfy, or the maximum number of questions that the user can answer without thinking about the burden.

  The question presentation opportunity input unit 104 is a program that accepts a question presentation opportunity input from the user terminal 20. The “question presentation opportunity” is a condition related to the timing and frequency of questions. For example, immediately before the health index value becomes minimum when the health index value has periodicity is one of the question presentation opportunities. Similarly, the timing when the accuracy of the health index value does not fall below a specific reference value is also a question presentation opportunity. The input question presentation opportunity is stored in the question / answer data storage unit 102.

  The presentation question selection unit 105 uses the model generated by the health index value calculation unit 108, the calculated health index value, and the accuracy of the health index value to determine the question selection criterion input by the question selection criterion input unit 103. The question presented in the question input / output unit 106 is selected so as to satisfy.

  The question output unit 106 is a program that outputs an appropriate question selected by the presentation question selection unit 105 to the user terminal 20 at an appropriate timing and frequency. The contents of the question to be output are determined using data stored in the question / answer storage unit 102.

  The answer input unit 107 is a program for inputting an answer to a question from the user terminal 20. The input answer is stored in the question / answer storage unit 102.

  The health index value calculation unit 108 generates a model for calculating the health index value from the data stored in the health data storage unit 101 and the question / answer storage unit 102, and when an answer to the question is input, It is a program that calculates health index values according to answers.

  The health index value calculation unit 108 also calculates the accuracy of the health index value calculated from the model when generating the model. Furthermore, a reduced model from which the question items having a small influence on the calculation of the health index value and the relationship between the question items are removed as necessary is generated. In addition, since it may take a lot of time to generate an excellent model, it is preferable to prepare a storage unit (not shown) for storing a model generated in advance. For example, it may be created using an algorithm or the like used in graphical modeling as described in “Graphical Modeling” (written by Masami Miyagawa) published by Asakura Shoten in 1997.

  Hereinafter, when simply described as “model”, it refers to a mathematical model for calculating a health index value using the relationship with the question item data. Examples of models include multiple regression models, logistic regression models, path analysis models, structural equation models, Bayesian networks, and the like. Since the structural equation model is a generalized model that includes many of the models listed above, in this embodiment, the structural equation model will be described as an example. Since a model based on Bayesian statistics such as a Bayesian network does not have an error variable, the reliability of a question item or the like is used instead of the error variable.

  Further, when “generate a model” is used, it means that a relationship structure between questions that are model components is defined and the strength of the relationship between model components is calculated. However, when the relationship structure has already been determined, calculating the strength of the relationship is called model generation. As a method for calculating the strength of the model relationship, for example, the method described in “Covariance Structure Analysis [Introduction] -Structural Equation Modeling-” (written by Hideki Toyoda) published by Asakura Shoten in 1998 is used.

  In the present embodiment, only a model having no loop in the direction of the relationship is used. Also, the items that do not directly affect the health index value are considered in the model so that they can be presented as new questions when unanswered or rejected.

FIG. 3 is a diagram showing a specific example of a model based on the question / answer data shown in FIG. In FIG. 3, “physical condition index value” and “mental index value” are health index values, and the others are question items. Arrows indicate relationships between items, and values attached to the arrows indicate the strength of relationships between items. For example, the “physical condition index value” is affected by “headache”, “fatigue”, and “gastrointestinal”, and the strengths of the influence are 2.1, 1.2, and 3.3, respectively. This can be expressed as a formula:
“Physical condition index value” = 2.1 × “Headache” + 1.2 × “Fatigue” + 3.3 × “Gastrointestinal”
It becomes. For example, when the answer of “headache” is 2, the answer of “fatigue” is 3, and the answer of “gastrointestinal” is 1, the “condition index value” is 2.1 × 2 + 1.2 × 3 + 3.3 × 1 = 11.1.

  For the sake of brevity, reference numerals are given to the relationship between health index values and items and the relationship between items, and numbers (hereinafter referred to as “item numbers”) are assigned to health index values and items. FIG. 4 is a diagram showing the relationship of the model of FIG. The numbers in parentheses indicate item numbers.

  The model can also be expressed using a matrix representation. Here, a description will be given using a matrix expression obtained by transposing a general adjacency matrix. When a certain item affects another item, the strength of the effect is written at the position where the row indicated by the item number of the affected item and the column indicated by the item number of the affected item intersect. By referring to the line of the item number corresponding to the item, it can be understood which item is affected. On the other hand, by referring to the column, it can be understood which item is affected. Hereinafter, a matrix representing this model is referred to as a model matrix.

  5 and 6 are diagrams showing specific examples when the models of FIGS. 3 and 4 are represented by matrices. Here, item numbers are given for easy reference. For example, the fact that “headache” in item number 3 has an influence of a1 on “health index value” in item number 1 is indicated by a1 in row 1 and column 3. Also, the fact that item number 3 “headache” does not affect the item number 4 “fatigue” is indicated by 0 in 4 rows and 3 columns. An advantage of using a matrix is that the processing can be performed mechanically with a good visibility. In addition, in recent years, multi-core CPUs continue to develop, and it is also one of the advantages that further increase in the speed of matrix operations is expected in the future. In particular, it is efficient when the relationship between items is close. On the other hand, when the relationship between items is sparse, the matrix is also sparse, which is not efficient. It is preferable to decide whether to perform matrix processing or graph processing according to the complexity of the model.

  “Health index value accuracy” is a value indicating the goodness of the health index value, or a performance index of a model used to calculate the health index value, and is used properly according to the application. For example, there are information criterion such as least square error, recall, precision, ROC (Receiver Operating Characteristic), AUC (Area Under Curve) of curve, AIC (Akaike Information Criterion).

  When a complicated model is used, a plurality of accuracy calculation methods can be considered depending on the question item used for calculating the health index value even if the model is the same. For example, in order to calculate the “health index value” in the model of FIG. 3, a method using three items of “headache”, “fatigue”, “gastrointestinal”, “headache”, “gastrointestinal”, “work load” A plurality of calculation methods such as a method using three items can be considered. In such a case, for example, the average value of the accuracy calculated for all cases is used as the accuracy of the health index value.

  The health index value presentation unit 109 is a program that presents the health index value calculated by the health index value calculation unit 108 to the user terminal 20.

  Hereinafter, the operation of the health index value presentation system 1 will be described with reference to FIGS.

<Question presentation process>
FIG. 7 is a flowchart showing a specific example of the question presentation process performed in the health index value presentation system 1. This question presentation processing is executed when the system structure is activated or when there is a possibility that the model structure is changed by changing health data or question data due to additional correction of a question sentence or answer.

  In step S <b> 701, the presentation question selection unit 105 determines whether the question selection criterion has been set. If it is determined that the question selection criterion is not set, the process proceeds to S702. On the other hand, if it is determined that the question selection criterion has been set, the process proceeds to S703.

  In step S702, the question selection criterion input unit 103 inputs a question selection criterion. Question selection criteria are related to accuracy and the number of questions. An appropriate one is selected according to the user and usage, and an appropriate criterion is input. It is also possible to input question selection criteria for both accuracy and number of questions. In this case, an appropriate question that satisfies the standard is made for both the accuracy and the number of questions.

  In step S703, the health index value calculation unit 108 determines whether a model has been generated. If it is determined that the model has been generated, the process proceeds to S705. On the other hand, if it is determined that the model has not been generated, the process proceeds to S704.

  In S704, the health index value calculation unit 108 generates a model.

  In step S705, the health index value calculation unit 108 determines whether to generate a contracted model. If it is determined that a contracted model is to be generated, the process proceeds to S706. For example, when there are a large number of question items, it takes time to select a presentation question or update a question presentation opportunity. In order to avoid this, it is possible to generate a reduced model by deleting the model component or the relationship between the model components. When the number of question items included in the model is small, a contracted model need not be generated. The generation of the contracted model or how much the model is contracted may be determined in advance, or may be input by the user when S705 is executed. If it is determined in S705 that a reduced model is not generated, the process proceeds to S707.

  In S706, the health index value calculation unit 108 generates a contracted model. The details of the contracted model generation process will be described later.

  In step S <b> 707, the presentation question selection unit 105 determines whether a presentation question has been selected in advance. If it is determined that the presentation question has been selected in advance, the process proceeds to S709. On the other hand, when it is determined that the presentation question is not selected in advance, the presentation question is selected (S708), and the process proceeds to S709. Details of the presentation question selection process will be described later.

  In step S <b> 709, the presentation question selection unit 105 determines whether it corresponds to a question presentation opportunity. The process waits until it corresponds to the question presentation opportunity, and if it corresponds, the process proceeds to S710.

The timing for checking whether it corresponds to the question presentation opportunity should not miss the time designated by the question presentation opportunity.

  In step S <b> 710, the question output unit 106 presents the question selected by the presentation question selection unit 105. Specifically, only the question that directly affects the health index value among the items selected as the presentation question in S708 is presented. Then, when the user does not answer or refuses to answer the presented question, another question is asked. For example, in the model of FIG. 3, when both “headache” and “work load” are selected as presentation questions, only “headache” is presented as the first question, but “headache” is not answered yet. If there is, ask a question about "work load". The question presentation opportunity after answering and the presented question update process will be described later.

<Reduced model generation process>
Next, the contracted model generation process in S705 of FIG. 7 described above will be described in detail. In the reduced model generation process, items are gradually deleted from the model that includes all items, and a reduced model is generated, and items are added gradually to models that do not include items other than health index values. There are two ways to generate a reduced model. In the following, the reduced model generation process by item deletion will be described, and then the reduced model generation process by item addition will be described.

(A) Reduced model generation process by item deletion FIG. 8 is a flowchart showing a specific example of the reduced model generation process by item deletion. A processing method using matrix representation will be described at the same time. Contracting the model corresponds to reducing the size of the matrix.

  In step S <b> 801, the health index value calculation unit 108 calculates a health index value and a ratio of the influence of each question item from other question items. Specifically, paying attention to items that have a direct influence, the ratio of influence is calculated for each target item in the overall influence. For example, “physical index value” in FIG. 3 is affected by 2.1 from “headache”, 1.2 from “fatigue”, and 3.3 from “gastrointestinal”. The ratio of the influence is 2.1 / (2.1 + 1.2 + 3.3) ≈0.32 for “headache” and 1.2 / (2.1 + 1.2 + 3.3) for “feeling of fatigue” ≈0.18, and “Gastrointestinal” is 3.3 / (2.1 + 1.2 + 3.3) ≈0.5. In addition, when using a matrix, it corresponds to calculating a ratio for every column. FIG. 9 is a diagram illustrating a matrix after the processing of S801 is performed on the matrix of FIG.

  In step S <b> 802, the health index value calculation unit 108 determines the presence / absence of an item whose output order is 0. Here, if it is determined that there is an item whose outgoing order is 0, the process proceeds to S803. “Order degree” is the number of items or health index values that an item affects. The “headache” in the model of FIG. 1 affects the two health index values of “physical index value” and “mental index value”, so the order of “headache” is 2. When a matrix is used, this corresponds to searching for a column in which all values are zero.

  If it is determined in S802 that there is no item whose outgoing order is 0, the process proceeds to S804.

  In step S803, the health index value calculation unit 108 deletes the item whose outgoing order is 0, and proceeds to step S804. It should be noted that the deletion method is short circuit removal, and all relationships connected to items whose output order is 0 are deleted. For example, in the model shown in FIG. 3, when “headache” is removed by a short circuit, “headache” is deleted, and relationships a1, a2, and a10 are also deleted at the same time. This corresponds to deleting a row and a column corresponding to an item number of a column in which all are 0 when using a matrix. For example, in the model of FIG. 3, when the values of the “headache” column are all 0, the row and column of item number 3 corresponding to “headache” are deleted. FIG. 10 is a diagram illustrating a matrix from which “headache” is deleted.

  In step S <b> 804, the health index value calculation unit 108 determines the presence / absence of an item whose incoming order is 1. Here, if it is determined that there is an item whose entry order is 1, the process proceeds to S805. “Order number” is the number of items or health index values that affect a certain item. Accordingly, since the item whose entry order is 1 affects only one item, it can be integrated with that item. Since the “headache” in the model of FIG. 3 is influenced only by “work load”, the entry order of “headache” is 1. When a matrix is used, this corresponds to searching for a row having only one non-zero value.

  In S804, if it is determined that there is no item with an incoming order of 1, the process proceeds to S806.

  In step S805, the health index value calculation unit 108 deletes the item whose incoming order is 1, and proceeds to step S806. The deletion method is degeneracy, and the item with the input order of 1 is deleted by integrating the item with the input order of 1 and the item affecting the item into one item. In item deletion by degeneration, the relationship between the item to be deleted and the item to be deleted and the item that affected the item to be deleted is deleted, but the relationship that the item to be deleted had is retained . For example, when “headache” is removed by degeneration in FIG. 3, “headache” and the relation a <b> 10 are deleted, but relations a <b> 1 and a <b> 2 are held by being connected to “work load”.

  In addition, when the relationship which mutually influences by the removal by degeneracy, the original relationship is left. As an example, a case where “fatigue” affects “headache” in FIG. 3 will be described. At this time, if the “headache” is deleted by removal due to degeneracy, a relationship that affects the “work amount” from the “feeling of fatigue” newly arises. In such a case, the relationship that affects the “work amount” from the newly generated “feeling of fatigue” is deleted, and only the relationship a11 that originally existed is left. If a plurality of relationships remain after removal by degeneration, the larger one is left. This corresponds to adding a column of the reduced item to a column having a value other than 0 in the row related to the reduced item and then deleting the row and the column related to the reduced item in the case of using a matrix. FIG. 11 is a diagram illustrating a matrix obtained by performing the processing of S805 on the matrix of FIG.

  In S806, the health index value calculation unit 108 regenerates the model. If the model structure has not changed, there is no need to regenerate the model. The regeneration of the model here refers to recalculating the relationship between the question and the item and the strength of the relationship between the items.

  In step S807, the health index value calculation unit 108 determines whether the generated model corresponds to an appropriate contraction model. If it is determined that the contracted model is appropriate, the process proceeds to S810. The reference for determining whether or not the model is appropriate is determined in advance, and the optimal one is used according to the application. For example, the number of items, the number of relationships, and the accuracy can be cited as criteria. When the number of items is used as a criterion, it is determined that the model is an appropriate contraction model when the number of items included in the model is equal to or less than the number of specified items. Similarly, when the number of relationships is used as a determination criterion, it is determined that the model is an appropriate contraction model when the number of relationships included in the model is equal to or less than the number of specified relationships. When accuracy is used as a determination criterion, a model in which the accuracy of the model exceeds the specified accuracy and the number of items and the number of relations are reduced to the maximum is determined to be an appropriate reduced model. If it is determined in S807 that the model is not an appropriate contraction model, the process proceeds to S808.

  In step S <b> 808, the health index value calculation unit 108 determines whether there is a relationship with a small influence. Here, if it is determined that there is a relationship having a small influence, the process proceeds to S809. Specifically, if the relationship between the health index value and the item or the relationship between the items calculated in S801 is smaller than the reference value for determining that the effect is small, the relationship is expressed as a relationship having a small effect. Judge that there is. A reference value for determining that the influence is small is determined in advance. For example, in the model of FIG. 3, when the reference value for determining that the influence is small is 0.2, a case is examined in which there is an item having a small influence in the relationship between the “physical condition index value” and the item. In this case, the effect on the “health index value” is 0.32 for “headache”, 0.18 for “feeling of fatigue”, and 0.5 for “gastrointestinal”, and the effect of “feeling of fatigue” is 0. Since it is smaller than 2, the relationship a3 is determined to be a relationship having a small influence. Similarly, for each health index value and each item, the relationship with a small influence is examined. This corresponds to searching for a row having a small influence for each column when using a matrix. If it is determined in S808 that there is no relationship with a small influence, the process returns to S801.

  In step S809, the health index value calculation unit 108 deletes the relationship having a small influence, and the process returns to step S801. A series of processing is repeated until an appropriate reduced model is generated. It should be noted that the number to be deleted may be singular or a plurality may be deleted in ascending order of influence. The smaller the number to be deleted at a time, the more times the model is regenerated. Therefore, an appropriate contracted model is likely to be generated, but processing time is required. The number to be deleted may be determined in advance or may be determined automatically. When the number to be automatically deleted is determined, adjustments such as decreasing the number to be deleted are performed as the number of model regeneration increases. This corresponds to setting the value of a row having a small influence among the columns to 0 when using a matrix. FIG. 12 is a diagram illustrating a matrix obtained by deleting the relationship of small influence from the matrix of FIG. 9 when the determination reference value for determining that the influence is small is 0.2.

  In step S810, the health index value calculation unit 108 determines the reduced model as a model to be used when selecting the question item, and ends the series of reduced model generation processing.

(B) Reduction model generation process by item addition FIG. 13 is a flowchart showing a specific example of reduction model generation process by item addition in the health index value presentation system 1. In the contracted model generation process by adding items, the model at the start of the process is only the health index value.

  In step S1301, the health index value calculation unit 108 determines an item whose distance from the health index value is 1 as an additional candidate item. The “distance from the health index value” is the number of relationships that appear when items that affect the health index value are traced in reverse. For example, in FIG. 3, “headache” is a distance 1 and “work amount” is a distance 2 with respect to “physical condition index value”.

  In S <b> 1302, the health index value calculation unit 108 calculates the health index value and the ratio of the influence of each question item from other question items. The calculation method is the same as that of the reduced model generation method by item deletion.

  In S1303, the health index value calculation unit 108 adds additional candidate items having a strong influence to the model. The number to be added may be singular or a plurality may be added in order from the strongest influence. The smaller the number to be added at a time, the more times the model is regenerated, so that an appropriate contracted model is likely to be generated, but processing time is required. The number to be added may be determined in advance or may be automatically determined. When the number to be automatically added is determined, adjustments such as decreasing the number to be added are performed as the number of model regeneration increases.

  In S1304, the health index value calculation unit 108 regenerates the model. However, if the model structure has not changed, there is no need to regenerate the model. Here, “regeneration of the model” means recalculating the relationship between the question and the item and the strength of the relationship between the items. In the case of using a matrix, since the strength of the relationship between items is calculated by generating a model, a value is added to a corresponding position in the matrix.

  In step S1305, the health index value calculation unit 108 determines whether the generated model is an appropriate contraction model. Note that the criterion for determining an appropriate contracted model is the same as the contracted model generation process by item deletion. If it is determined that the model is not an appropriate contraction model, the process proceeds to S1306. If it is determined that the model is an appropriate contraction model, the process proceeds to S1307.

  In S1306, the health index value calculation unit 108 determines an item further away from the health index value as an additional candidate, and returns to S1302. Note that the processing of S1302 to S1306 is repeated until an appropriate contracted model is generated.

  In step S <b> 1307, the health index value calculation unit 108 determines the contracted model as a model to be used when selecting the question item, and ends the series of contracted model generation processing.

<Presentation question selection process>
Next, the presentation question selection process performed in S708 of FIG. 7 described above will be described in detail. The presented question selection process includes a method based on accuracy and a method based on the number of questions. Hereinafter, a method based on accuracy will be described, and then a method based on the number of questions will be described. In addition, as a criterion for selecting an item as a presentation question, a criterion suitable for a problem can be adopted. Here, the case where the strength of the influence on the health index value calculated from the average value of each item with respect to the average value of the health index value is used as a reference will be described as an example.

(A) Method based on accuracy FIG. 14 is a flowchart illustrating a specific example of a presentation question selection process based on accuracy in the health index value presentation system 1.

  In step S <b> 1401, the presentation question selection unit 105 calculates the average value of the influence strengths by considering the reliability related to the influence strength between the items as a weight. An appropriate index may be used depending on the problem to represent the reliability. Here, standard deviation is used as an index representing reliability. In this case, if the value obtained by subtracting the corresponding standard deviation from the absolute value of the average value of each item is 0 or more, that value is used, and if it is 0 or less, 0 is used as the weight. For example, the effect of “headache” on “physical index value” in FIG. 3 is that the average value of “headache” is 1.8, and the standard deviation of the relationship between “headache” and “physical index value” is 0.2. At some point, it is calculated as 1.8−0.2 = 1.6. Also, if the average value is 0.5 and the standard deviation of the relationship is 0.6, it is calculated as 0 because 0.5-0.6 = -0.1. The case where there is an error term can be handled similarly. FIG. 15 is a diagram for explaining a relationship between items constituting the model, and shows an average value calculated in consideration of reliability as a weight.

  In S1402, the presentation question selection unit 105 calculates the degree of influence on each health index value for each item variable. Depending on the problem, the impact on each health index value may be calculated using the coefficient of the item as it is, or using the health index value used in the reduced model generation process and the ratio of the impact on the item. May be.

  As an example, a method will be described in which the method of using the coefficient of the item as it is and when the weighted average value is FIG. 15, the effect on the “physical index value” in the model of FIG. 3 is calculated. Tracing the relationship that affects them in reverse, the items affecting the "physical index value" are "headache", "fatigue", "gastrointestinal", "work volume", and "satisfaction with work". Therefore, the impact is calculated for each of these items. The influence of the item is calculated by multiplying the coefficient of the relationship from the target item to the health index value and the weight of the target item. For example, the influence of each item directly connected to “physical condition index value” is 2.1 × 1.6 = 3.36 for “headache” and 1.2 × 5.7 = 6. 84, “Gastrointestinal” is 3.3 × 2.3 = 7.59. “Gastrointestinal” has an influence of “1.5 × 1.2 × 2.4 = 4.32” on “physical condition index value” even through “fatigue”, and therefore 7.59 + 4.32 = The impact is 11.91. Similarly, the “work amount” is 0.8 × 2.1 × 2 + 2.3 × 1.2 × 2.1 + 1.2 × 3.3 × 2 + 1.2 × 1.5 × 1.2 × 2 = 21.1. 4. “Satisfaction with work” is 0.7 × 1.2 × 1.7 = 1.43. It should be noted that the weight is considered only for the relationship between the target item and the adjacent item.

When a matrix is used, the degree of influence is calculated by calculating the sum of the matrix obtained by multiplying each component of the model matrix by the weighted average value and the model matrix. There are two methods for calculating the sum of model matrices: a method for calculating the sum as it is and a method using the sum formula. The former is adopted when there is no item far from the health index value, and the latter is adopted when it exists. When the unit matrix is I, the model matrix is A, and the matrix calculated from the multiplication of each component of the model matrix and the weighted average value is B, in the method of calculating the sum of the model matrix, (I + A + A ² +... + A ^n-1 ) B is calculated. n is the maximum distance between the health index value and the item. For example, in the model of FIG. On the other hand, when using the sum formula, (IA) ⁻¹ B is calculated. Note that the inverse matrix of the matrix is represented as A- ¹ . As an inverse matrix calculation method, an efficient algorithm is used for a sparse matrix.

  16 is a matrix obtained by multiplying each component of the model matrix of FIG. 5 and the weighted average value of FIG. 15, and FIG. 17 is a matrix showing the degree of influence. The influence on each health index value is represented in the row of the corresponding item number. For example, in the matrix of FIG. 17, the influence on “physical condition index value” is represented in the row of item number 1 corresponding to “physical condition index value”. Further, the influence of each item on each health index value is represented in the column of the item number column corresponding to each item in the row of health index values. For example, since the item number of “work amount” is 8, the influence of “work amount” on “physical condition index value” is shown as 21.4 in 1 row × 8 column.

  In S1403, the presentation question selection unit 105 selects a presentation question candidate according to the degree of influence calculated in S1402. In the selection method, all items are unselected in the initial state, and items are added in order from the strongest items. In the initial state, all items are selected, and items are deleted in order from the weakest items. There is. The number of items to be added or deleted may be singular or plural depending on the strength of the influence. Also, the smaller the number of objects to be processed at a time, the greater the number of accuracy calculation processes, so that it is easy to select an appropriate question, but processing time is required. Therefore, the number of processing targets may be determined in advance or may be determined automatically. When automatically determining the number of processing targets, it is preferable to make adjustments such as decreasing the number of processing targets as the number of accuracy calculation processes increases.

  In S1404, the presentation question selection unit 105 calculates the accuracy of the health index value calculated from the model configured from the question items selected in S1403.

  In step S1405, the presentation question selection unit 105 determines whether the accuracy calculated in step S1404 satisfies the question selection criterion. If it is determined that the question selection criteria are satisfied, the presented question determination process is terminated. On the other hand, if it is determined that the question selection criterion is not satisfied, the process proceeds to S1406.

  In S1406, when the presentation question selection unit 105 uses an additional method as the selection method, if the accuracy of the health index value calculated in S1404 is higher than the accuracy specified by the question selection criterion, The presentation question candidate selected at that time is determined as the final presentation question, and the presentation question determination process with a series of accuracy is finished. Conversely, if the accuracy specified by the question selection criterion is higher, the number selected as the presentation question candidates is increased and S1403 is redone.

  On the other hand, when the deletion method is used as the selection method, if the accuracy of the health index value calculated in S1404 is higher than the accuracy specified by the question selection criterion, it is selected as a presentation question candidate. S1403 is reduced and the number is reduced. On the other hand, if the accuracy specified by the question selection criterion is higher, in the loop processing from S1403 to S1405, the presentation question candidates selected in the previous loop are confirmed as the final presentation questions and a series. The presenting question determination process with the accuracy of is terminated.

(B) Method based on the number of questions FIG. 18 is a flowchart showing a specific example of the presentation question selection process based on the number of questions in the health index value presentation system 1. Since the processing up to S1802 is the same as the presentation question selection processing based on accuracy, the description thereof is omitted.

  In S <b> 1803, the presentation question selection unit 105 determines the number of items having a strong influence as the number of presentation questions specified by the question selection criterion, and ends the series of presentation question selection processing. To improve the accuracy of health index values, calculate the accuracy of health index values when using fewer questions than the number specified in the question selection criteria, and select the question with the highest accuracy. May be used as a presentation question. For example, when 10 is specified as the question selection criterion, the accuracy of the health index value is calculated not only when the number of questions is 10, but also when 9 or 8 questions are used, and the highest accuracy is obtained. The question selected at the time of confirmation is confirmed as a presentation question.

<Question presentation opportunity update process>
Next, the question presentation opportunity update process performed in S710 of the question presentation process of FIG. 7 will be described in detail. The question presentation opportunity update process is a process executed when an answer is input from the user, and the updated question presentation opportunity is reflected in the process of S809. As for the question presentation opportunity update process, there are a method based on a cycle and a method based on accuracy. Hereinafter, a method based on a cycle will be described, and then a method based on accuracy will be described.

(A) Method based on cycle FIG. 19 is a flowchart showing a specific example of the question presentation opportunity update process by cycle in the health index value presentation system 1.

  In step S1901, the answer input unit 107 acquires an answer input from the user terminal 20. Note that the question presentation opportunity update process is performed according to the answer from the user terminal 20, so if no answer is obtained from the user terminal 20, the process proceeds to the next step until the answer is obtained. Absent.

  In S1902, the health index value calculation unit 108 applies the answer input in S1901 to the model, and calculates a health index value.

  In S1903, the health index value calculation unit 108 determines whether the health index value has periodicity using the health index value obtained in S1902 and the health index value obtained in the past. As a method for examining periodicity, an appropriate method may be used depending on the problem. For example, use the Fourier transform to check for the existence of a specific period with a large amplitude, or use the autoregressive moving average model, which is one of the time series models, to check for the existence of a term with a large value for the moving average. There are methods. The time series model is described in, for example, “Introduction to Time Series Analysis” (authored by Genshiro Kitagawa) published by Asakura Shoten in 1996. If it is determined that there is periodicity, the process proceeds to S1904. If it is determined that there is no periodicity, the series of question presentation opportunity update processing ends.

  In S1904, the health index value calculation unit 108 identifies the cycle. An appropriate method for specifying the period may be used depending on the problem. When Fourier transform is used, a period with a large amplitude is specified. When an autoregressive moving average model is used, a moving average order having a large coefficient is specified.

  In step S1905, the health index value calculation unit 108 updates the question presentation opportunity stored in the question / answer storage unit 102 according to the period specified in step S1904. The question presentation timing in the period variation is designated in advance, and the question presentation opportunity is set so that the question is made at an appropriate time based on the designation. As the timing designation of the question presentation, for example, the timing at which the health index value starts to rise from the lowest state of the fluctuation cycle and the timing at which the health index value starts to fall from the highest state are designated as the phase and moving average order. When the question presentation opportunity is updated, the series of question presentation opportunity update processing ends.

(B) Method Based on Accuracy FIG. 20 is a flowchart showing a specific example of the question presentation opportunity update process based on accuracy in the health index value presentation system 1.

  In step S2001, the answer input unit 107 acquires an answer input from the user terminal 20. Note that the question presentation opportunity update process is performed according to the answer from the user terminal 20, so if no answer is obtained from the user terminal 20, the process proceeds to the next step until the answer is obtained. Absent.

  In S2002, the health index value calculation unit 108 applies the answer input in S2001 to the model, and calculates the health index value and the accuracy of the health index value.

  In step S2003, the health index value calculation unit 108 determines whether to increase the number of questions. The setting related to the increase in the number of questions may be specified in advance in the question presentation opportunity input unit 104, or the possibility of increasing the number of questions may be automatically determined. In the case of automatic determination, for example, an increase in the number of questions is made effective when the accuracy of the health index value is not sufficiently improved even if the question presentation interval is shorter than a certain period. If the increase in the number of questions is valid, the process proceeds to S2004, and processing is performed in consideration of both the number of questions and the question presentation interval. On the other hand, when the increase in the number of questions is invalid, the process proceeds to S2008, and processing is performed in consideration of only the question presentation opportunity.

  In S2004, the health index value calculation unit 108 generates a prediction model regarding the accuracy of the health index value based on the number of questions and the question presentation interval. Any prediction model can be used as long as the accuracy of the health index value can be estimated from the number of questions and the question presentation interval. For example, there is a multiple regression model in which the number of questions and the question presentation interval are explanatory variables and the accuracy of health index values is an objective variable. When the multiple regression model is used, it can be predicted that the question selection criterion regarding the accuracy of the health index value is satisfied if the value of the objective variable exceeds the threshold regarding the accuracy. Note that since the prediction model is difficult to be generated properly if the number of data is not sufficient, a prediction model may be prepared in advance, and the model may be used at the beginning of system operation.

  In S2005, the health index value calculation unit 108 updates the number of questions and the question presentation interval according to the prediction result obtained from the prediction model generated in S2004. The accuracy of the current health index value is calculated, and if the accuracy satisfies the question selection criteria, the number of questions is reduced or the question presentation interval is widened. On the other hand, if not satisfied, the number of questions is increased or the question presentation interval is narrowed. Whether to preferentially change the number of questions or the question presentation interval may be determined in advance or may be changed automatically. For example, in the case of automatically changing, a combination with the lowest accuracy among the combinations of the number of questions and the question presentation interval whose accuracy of the health index value satisfies the question selection criterion is adopted.

  In step S2006, the health index value calculation unit 108 determines whether to increase the number of questions. If the number of questions is to be increased, the process proceeds to S2007. On the other hand, when the number of questions is not increased, the series of question presentation opportunity update processing ends.

  In S2007, the health index value calculation unit 108 performs a presentation question selection process for the number of questions set in S2005. Then, after selecting the presentation question, the series of question presentation opportunity update processing is terminated.

  If it is determined in S2003 that the number of questions is not increased, the process proceeds to S2008.

  In S2008, the health index value calculation unit 108 generates a prediction model regarding the accuracy of the health index value based on the question presentation interval. As for the prediction model, a model similar to S2004 may be used, or a time series model such as an autoregressive moving average model may be used.

  In S2009, the health index value calculation unit 108 updates the question presentation interval according to the prediction result obtained from the generated prediction model. As in S2005, the accuracy of the current health index value is calculated. If the accuracy satisfies the question selection criteria, the question presentation interval is widened, and if not, the question presentation interval is narrowed. When a time series model is used in S2008, the question presentation interval may be set so as to present a question at a time when the accuracy of the health index value satisfies the question selection criterion at a minimum. When the question presentation interval is set, the series of question presentation opportunity update processing ends.

  In this way, an appropriate number of questions are asked according to the set question selection criteria, and answers to these questions are applied to the internally generated model, thereby reducing the burden on the user and providing a highly accurate health index. A value can be presented.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. For example, each storage unit may be a database, text data, data in a format specialized for each application, or the like, and the storage format and storage medium can be optimally selected depending on the application. Each input unit includes a keyboard, a mouse, a pen-type input device, a portable terminal, a measuring device, and the like, and an input format and an input device can be optimally selected according to the application. Similarly, the health index value presenting unit 109 includes a computer display, a printing machine, a portable terminal, and the like, and it is possible to optimally select a presentation format and a presentation device according to the application. In addition, if it is associated with a question, it may be presented as an alarm sound, music, LED blinking, or the like. Furthermore, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment.

The block diagram which shows the example of whole structure of the health index value presentation system 1 which concerns on one Embodiment of this invention. The figure which shows the specific example of the question and answer data which concern on one Embodiment of this invention. The figure which shows the specific example of the model based on the question and answer data concerning one Embodiment of this invention. The figure which shows the specific example of the model based on the question and answer data concerning one Embodiment of this invention. The figure which represented the model which concerns on one Embodiment of this invention with the matrix. The figure which represented the model which concerns on one Embodiment of this invention with the matrix. The flowchart which shows the specific example of the question presentation process in the health index value presentation system 1 which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the reduction | restoration model production | generation process by the item deletion in the health index value presentation system 1 which concerns on one Embodiment of this invention. The figure explaining the ratio of the influence between the items which comprise the model which concerns on one Embodiment of this invention with a matrix. The figure which shows the specific example which performed the item deletion from the model matrix which concerns on one Embodiment of this invention. The figure which shows the specific example which performed the item deletion from the model matrix which concerns on one Embodiment of this invention. The figure which shows the specific example which performed the item deletion from the model matrix which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the reduction | restoration model production | generation process by the item addition in the health index value presentation system 1 which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the presentation question selection process by the precision in the health index value presentation system 1 which concerns on one Embodiment of this invention. The figure explaining the relationship between the items which comprise the model which concerns on one Embodiment of this invention. The figure which shows the model matrix in the health index value presentation system 1 which concerns on one Embodiment of this invention. The figure which shows the specific example of the matrix explaining the degree of the influence between the items which comprise the model which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the presentation question selection process by the number of questions in the health index value presentation system 1 which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the question presentation opportunity update process by the period in the health index value presentation system 1 which concerns on one Embodiment of this invention. The flowchart which shows the specific example of the question presentation opportunity update process by the precision in the health index value presentation system 1 which concerns on one Embodiment of this invention.

Explanation of symbols

1 ... Health index value presentation system,
10 ... server, 20 ... user terminal, 30 ... network,
101 ... health data storage unit, 102 ... question / answer storage unit,
103 ... Question selection criterion input unit, 104 ... Question presentation opportunity input unit,
105 ... presented question selection unit, 106 ... question output unit, 107 ... answer input unit,
108... Health index value calculation unit, health index value presentation unit 109.

Claims (4)

A health data storage unit for storing health data;
A plurality of questions regarding the health data and a question / answer storage unit for storing the answers;
A question selection criterion input unit for inputting a selection criterion of the stored question;
A presentation question selection unit that selects a question to be presented to the user terminal based on the input selection criteria;
A question output unit that acquires the question selected by the presentation question selection unit from the question / answer storage unit and outputs the question to the user terminal;
An answer input unit that obtains an answer input from the user terminal with respect to the output question and stores the answer in the question / answer storage unit;
The health data storage unit to the stored health data, questions previously stored Symbol question storage unit, based on the strength of relationship and the relationship between the plurality of the question, among a plurality of model components has impact A model structure that has a graph structure associated with each other and that generates a model for calculating a health index value indicating a health condition , and has a zero degree of influence on the calculated health index value or other model components A reduced model in which the relationship between the elements and the model components is reduced from the model is generated together, and the health index value is calculated by applying the answer input in the answer input unit to the model or the reduced model A health index value calculation unit to
A health index value presentation unit for presenting the calculated health index value to the user terminal;
Health index value presentation system, characterized in that to obtain Bei the. A health data storage unit for storing health data;
A plurality of questions regarding the health data and a question / answer storage unit for storing the answers;
A question selection criterion input unit for inputting a selection criterion of the stored question;
A presentation question selection unit that selects a question to be presented to the user terminal based on the input selection criteria;
A question output unit that acquires the question selected by the presentation question selection unit from the question / answer storage unit and outputs the question to the user terminal;
An answer input unit that obtains an answer input from the user terminal with respect to the output question and stores the answer in the question / answer storage unit;
A model for calculating a health index value indicating a health condition based on health data stored in the health data storage unit, questions stored in the question / answer storage unit, relationships between the plurality of questions and the strength of the relationship A health index value calculation unit that calculates the health index value by applying the answer input in the answer input unit to the model,
A health index value presentation unit for presenting the calculated health index value to the user terminal;
And a question presentation opportunities input unit for inputting a time interval for presenting a question presenting opportunities the question at the user terminal,
The presentation question selection unit selects a question to be presented to the user terminal based on the question presentation opportunity input by the question presentation opportunity input unit and the periodicity of the health index value calculated by the health index value calculation unit health index value presentation system that is characterized in that. The health index value presentation system according to claim 1 or 2 , wherein the selection criterion of the question is an accuracy or a maximum number of questions to be presented that the health index value should satisfy at a minimum. A server computer calculates a health index value indicating a health state according to input information from a plurality of user terminals connected via a network, and presents the health index value presenting method to the user terminal,
A health data storage step for storing the health data in a health data storage device;
A question / answer storage step for storing a plurality of questions relating to the health data and their answers in a question / answer storage device;
A question selection criterion input step of inputting the stored question selection criterion from the user terminal;
A presentation question selection step of selecting a question to be presented to the user terminal based on the input selection criteria;
Obtaining the selected question from the question / answer storage device and outputting it to the user terminal, a question output step;
An answer input step of obtaining an answer input from the user terminal with respect to the output question and storing it in the question / answer storage device;
Based on the health data stored in the health data storage step, the question selection criteria input in the question selection criteria input step, the health index value, the relationship between the questions, and the strength of the relationship, a plurality of A model generation step of generating a model having a graph structure in which model components are associated with each other by the degree of influence, and calculating the health index value ;
From the generated model, a reduced model that generates a reduced model in which the calculated health index value or other model component has a degree of influence of 0 and the relationship between the model components is reduced. Generation step;
A health index value calculating step of calculating the health index value by applying the answer input in the answer input step to the model or the contracted model ;
A health index value presentation step of presenting the calculated health index value to the user terminal;
Health index value presentation method according to claim that you have a.

Images