Subhealth population prevention scheme health management system based on artificial intelligence

Abstract

The invention discloses a sub-health population prevention scheme health management system based on artificial intelligence, and relates to the technical field of health management, a continuous expectation model is constructed, user behavior data missing is dynamically monitored, health information silence caused by work pressure or habitual neglect is effectively captured, and the health management efficiency is improved. The defect that a traditional health management system only depends on initiative data submission of the user is overcome; the system optimizes model parameters by using a least square method, adjusts an anomaly judgment standard through a dynamic threshold factor, synthesizes an anomaly indication signal and multi-dimensional features of a user in health risk assessment, and finally generates a personalized intervention strategy; in addition, a feedback updating mechanism is adopted, decision parameters are continuously and adaptively corrected according to the intervention effect, the overall early warning and intervention precision is improved, and therefore the disadvantages that an existing system lags in response to hidden risks and intervention means are templated are overcome, and sub-health risks can be found and intervened in the early stage.

Description

Technical Field

[0001] The present invention relates to the technical field of health management, and in particular to an artificial intelligence-based health management system for preventing sub-healthy people. Background Art

[0002] Traditional health management systems rely on periodic physical examinations, active check-in records, and scheduled questionnaires, and are suitable for users who actively pay attention to their own health. However, in actual applications, many people, especially young and middle-aged people with fast-paced work, often do not proactively report or provide feedback when their health begins to deteriorate. We regard this as healthy silence.

[0003] In the existing system, the platform regularly pushes reminders, but this mechanism lacks flexibility and personalization. For users who reduce input due to work pressure, habitual neglect, or psychological resistance, the reminder effect seems formal and easily overlooked. Enterprises and institutions use manual follow-up to make up for data gaps, but due to limited coverage and low efficiency, the system will simply assume that this is a brief omission rather than an anomaly. In fact, this silence may reflect the user's neglect of their health status, or even changes in sub-health status caused by endocrine or psychological stress, but traditional solutions have difficulty in identifying the risk of this absence in a timely manner. Existing interventions are mostly based on unified reminders and fixed-period follow-up, lacking active identification and in-depth analysis of individual behavioral continuity. Therefore, the health management system urgently needs to introduce new thinking, from focusing on users' explicit behavior to focusing on the implicit signal of users' silence, so as to automatically capture potential risks when users fail to update or express their health records, and provide timely feedback through personalized intervention measures. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention provides an artificial intelligence-based health management system for preventing sub-healthy people, which solves the problem that existing interventions are mainly based on unified reminders and fixed-period follow-up, but lack active identification and in-depth analysis of individual behavioral continuity.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The embodiment of the present invention provides a health management system for sub-healthy population prevention program based on artificial intelligence, which includes:

[0008] A data collection unit that acquires user behavior data and health data information from health management processes, including smart wearable devices, mobile terminals, and regular physical examinations;

[0009] a data integration unit for fusing the health data information to form a user health information database;

[0010] A user behavior profile generating unit, which generates a behavior profile within a predetermined period based on a user health information database;

[0011] An anomaly detection unit analyzes the behavior profile based on a preset behavior pattern model, determines whether the user's records are missing or below an expected level within a predetermined period, and generates an anomaly indication signal;

[0012] a risk assessment unit, combining the abnormal indication signal with the user's historical health data to determine the user's health risk level by comparison;

[0013] The personalized intervention decision unit outputs an intervention strategy that matches the user's actual situation based on the risk level and the user's personalized information.

[0014] As a preferred solution of the artificial intelligence-based health management system for sub-healthy population prevention program described in the present invention, the behavior pattern model includes a continuity expectation model constructed based on statistics and machine learning methods, which is used to dynamically predict and judge the missing user behavior data.

[0015] As a preferred solution of the artificial intelligence-based health management system for sub-health population prevention program described in the present invention, a preset behavior pattern model is established in the abnormality detection unit to predict the records that the user should have at a specific moment, thereby determining whether there is a record missing phenomenon. The model expression is:

[0016]

[0017] in, represents the expected record value at time t, i represents the index of the historical record, d i represents the weight of the i-th historical record, x t-i represents the actual recorded value at time ti, e represents the bias parameter of the model, and n represents the number of historical records involved in the model calculation;

[0018] The least squares method is used to train the model parameters, and the objective function used is:

[0019]

[0020] Among them, x t Indicates the actual recorded value at time t, t represents the time index, n represents the number of historical records involved in the model calculation, d i represents the weight of the i-th historical record, x t-iRepresents the actual recorded value at time ti, e represents the bias parameter of the model, and T represents the end time of the record during the training period.

[0021] As a preferred solution of the health management system for sub-health population prevention program based on artificial intelligence described in the present invention, wherein: in the abnormality detection unit, in the abnormality detection stage, an abnormality indication signal is generated by comparing the actual record with the expected record, which is mathematically expressed as: Among them, I t Indicates the abnormal indication signal at time t. If the condition is met, the value is 1, otherwise it is 0. t represents the actual recorded value at time t, c represents the preset abnormality judgment threshold factor, represents the expected record value at time t, 1{·} represents an indicator function, which takes 1 if the conditions in the braces are met and 0 otherwise.

[0022] As a preferred solution of the health management system for sub-health population prevention program based on artificial intelligence described in the present invention, wherein: in the abnormality detection stage, the threshold factor is dynamically adapted, and its calculation formula is: c = 1-δ / σ x , where c represents the preset abnormality judgment threshold factor, δ represents the allowable deviation range, σ x Indicates the standard deviation of historical data.

[0023] As a preferred solution of the health management system for sub-healthy population prevention program based on artificial intelligence described in the present invention, the personalized intervention decision-making unit further integrates the data in the health management link, and actively warns based on user silence data and lack of behavioral continuity; the data in the health management link also includes health self-test, risk prediction and health assessment data.

[0024] As a preferred solution of the health management system for sub-healthy population prevention program based on artificial intelligence described in the present invention, wherein: in the personalized intervention decision unit, an intervention strategy is generated by constructing a decision function according to the abnormal indication results of the abnormality detection unit and the user's personality characteristics, and the decision parameters are dynamically adjusted based on feedback information.

[0025] As a preferred solution of the artificial intelligence-based health management system for sub-healthy population prevention program described in the present invention, the process of determining the intervention strategy in the personalized intervention decision unit includes:

[0026] The user's personality characteristics are weighted and summarized to generate an overall characteristic score, the mathematical expression of which is:

[0027]

[0028] Among them, θ represents the user's overall personality feature score, j is the index of the user feature, and w j represents the weight of the jth user feature, u j represents the jth user feature value, and m represents the total number of user features.

[0029] As a preferred solution of the artificial intelligence-based health management system for sub-health population prevention program described in the present invention, the process of determining the intervention strategy in the personalized intervention decision unit further includes:

[0030] The anomaly detection results are combined with the overall feature scores, and the intervention strategy is generated through decision mapping. The decision function adopts the form of a logical function, and its expression is:

[0031]

[0032] Among them, S t represents the personalized intervention strategy generated at time t, I t represents the abnormality indication signal at time t, refers to the detection result, α represents the weight of the abnormal signal in the decision, β represents the weight of the user's overall feature score in the decision, γ represents the decision bias, and exp(·) represents the natural exponential function, which is used here to implement the logistic mapping.

[0033] As a preferred solution of the health management system for sub-health population prevention program based on artificial intelligence described in the present invention, wherein: the personalized intervention decision unit also establishes a feedback update mechanism, setting r t is the actual feedback value of the intervention effect, η is the learning rate, and the parameter update rule is:

[0034] α new =α old +η(r t -S t ),

[0035] β new =β old +η(r t -S t )θ,

[0036] γ new =γ old +η(r t -S t ),

[0037] w j,new =w j,old +η(r t -S t )u j ,

[0038] Among them, αold and α new Represent the abnormal signal weights before and after adjustment, β old and β new Represent the user feature weights before and after adjustment, γ old and γ new Respectively represent the decision bias before and after adjustment, w j,old and w j,new Respectively represent the values of the j-th user feature weight before and after adjustment, r t It represents the actual feedback of the intervention at time t, reflecting the effect of the intervention strategy. η represents the learning rate, which controls the step size of the feedback adjustment.

[0039] The beneficial effects of the present invention are: the present invention constructs a continuous expectation model, dynamically monitors the missing user behavior data, and effectively captures the silence of health information caused by work pressure or habitual neglect, thereby making up for the deficiency of traditional health management systems that only rely on users to actively submit data; the system uses the least squares method to optimize model parameters, and adjusts the abnormal judgment criteria through dynamic threshold factors, and integrates abnormal indication signals and user multi-dimensional characteristics in health risk assessment, and finally generates personalized intervention strategies; in addition, a feedback update mechanism is adopted to enable decision parameters to be continuously adaptively corrected according to the intervention effect, thereby improving the overall warning and intervention accuracy, thereby solving the disadvantages of the existing system's delayed response to hidden risks and templated intervention methods, and helping to detect and intervene in sub-health risks at an early stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is a schematic diagram of the framework of the health management system for the sub-healthy population prevention program in Example 1. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0043] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0045] Existing health management systems for sub-health prevention programs often rely on active user input, timed check-ins, or questionnaire responses. Consequently, when user data is missing during a period, it is often considered normal inactivity, failing to identify hidden health risks.

[0046] The present invention generates user behavior files to dynamically monitor the health data records of users within a given management cycle, and uses preset behavior pattern models to determine abnormal phenomena such as record interruptions or insufficiency, thereby assisting in accurate assessment of health risks and outputting personalized intervention strategies that match the risk level; it integrates traditional physical examinations, health self-tests, and periodic assessment routine steps, and also quantifies the hidden phenomenon of "silence" in user behavior, effectively improving the defect of traditional systems that are insufficiently sensitive to abnormal silence phenomena, thereby achieving more refined and dynamic health management in prevention programs for sub-healthy people; the following embodiments are proposed.

[0047] Example 1, with reference to Figure 1 This embodiment provides an artificial intelligence-based health management system for preventing sub-healthy people, including:

[0048] The data collection unit obtains user behavior data and health data information from the health management process; the health management process includes smart wearable devices, mobile terminals and regular physical examinations;

[0049] Data integration unit, which integrates health data information to form a user health information database;

[0050] A user behavior profile generating unit, which generates a behavior profile within a predetermined period based on a user health information database;

[0051] The anomaly detection unit analyzes the behavior files based on the preset behavior pattern model, determines whether the user's records are missing or below the expected record level within a given period, and generates an anomaly indication signal;

[0052] The behavior pattern model includes a continuous expectation model built based on statistical and machine learning methods, which is used to dynamically predict and judge the missing user behavior data;

[0053] In the anomaly detection unit, a preset behavior pattern model is established to predict the records that a user should have at a specific moment, so as to determine whether there is a missing record phenomenon. The model expression is:

[0054]

[0055] in, represents the expected record value at time t, i represents the index of the historical record, d i represents the weight of the i-th historical record, x t-i represents the actual recorded value at time ti, e represents the bias parameter of the model, and n represents the number of historical records involved in the model calculation;

[0056] The least squares method is used to train the model parameters, and the objective function used is:

[0057]

[0058] Among them, x t Indicates the actual recorded value at time t, t represents the time index, n represents the number of historical records involved in the model calculation, d i represents the weight of the i-th historical record, x t-i represents the actual recorded value at time ti, e represents the bias parameter of the model, and T represents the end time of the record in the training period;

[0059] In the anomaly detection unit, during the anomaly detection phase, an anomaly indication signal is generated by comparing the actual record with the expected record, which is mathematically expressed as: Among them, I t Indicates the abnormal indication signal at time t. If the condition is met, the value is 1, otherwise it is 0. t represents the actual recorded value at time t, c represents the preset abnormality judgment threshold factor, represents the expected record value at time t, 1{·} represents an indicator function, which takes 1 if the conditions in the braces are met and 0 otherwise;

[0060] In the anomaly detection phase, the threshold factor is dynamically adapted, and its calculation formula is: c = 1-δ / σ x , where c represents the preset abnormality judgment threshold factor, δ represents the allowable deviation range, σ x Indicates the standard deviation of historical data;

[0061] Specifically, a continuous expectation model is established here, using historical behavior data to predict the expected record values of users at each moment. The least squares method is used to optimize the parameters to make the predicted values close to the actual record changes. A dynamic threshold factor is used to adaptively correct the detection standard so that when the actual record is significantly lower than the expected level, an abnormal indication signal can be generated in time. A combination of statistics and machine learning methods is used to dynamically judge the phenomenon of missing records.

[0062] The risk assessment unit combines the abnormal indication signal with the user's historical health data to determine the user's health risk level through comparison;

[0063] The personalized intervention decision-making unit outputs an intervention strategy that matches the user's actual situation based on the risk level and user personalized information;

[0064] The personalized intervention decision-making unit further integrates data from the health management process to provide proactive warnings based on user silence data and lack of behavioral continuity. Data from the health management process also includes health self-testing, risk prediction, and health assessment data.

[0065] In the personalized intervention decision unit, based on the abnormality indication results of the abnormality detection unit and the user's personality characteristics, an intervention strategy is generated by constructing a decision function, and the decision parameters are dynamically adjusted based on the feedback information;

[0066] In the personalized intervention decision-making unit, the process of determining the intervention strategy includes:

[0067] The user's personality characteristics are weighted and summarized to generate an overall characteristic score, the mathematical expression of which is:

[0068]

[0069] Among them, θ represents the user's overall personality feature score, j is the index of the user feature, and w j represents the weight of the jth user feature, u j represents the jth user feature value, and m represents the total number of user features;

[0070] In the personalized intervention decision-making unit, the process of determining the intervention strategy also includes:

[0071] The anomaly detection results are combined with the overall feature scores, and the intervention strategy is generated through decision mapping. The decision function adopts the form of a logical function, and its expression is:

[0072]

[0073] Among them, S t represents the personalized intervention strategy generated at time t, I t represents the abnormality indication signal at time t, refers to the detection result, α represents the weight of the abnormal signal in the decision, β represents the weight of the user's overall feature score in the decision, γ represents the decision bias, and exp(·) represents the natural exponential function, which is used here to implement the logistic mapping;

[0074] In the personalized intervention decision-making unit, a feedback update mechanism is also established, setting r t is the actual feedback value of the intervention effect, η is the learning rate, and the parameter update rule is:

[0075] α new =α old +η(r t -S t ),

[0076] β new =β old +η(r t -S t )θ,

[0077] γ new =γ old +η(r t -S t ),

[0078] w j,new =w j,old +η(r t -S t )u j ,

[0079] Among them, α old and α new Represent the abnormal signal weights before and after adjustment, β old and β new Represent the user feature weights before and after adjustment, γ old and γ new Respectively represent the decision bias before and after adjustment, w j,old and w j,new Respectively represent the values of the j-th user feature weight before and after adjustment, r t It represents the actual feedback of the intervention at time t, reflecting the effect of the intervention strategy; η represents the learning rate, which controls the step size of the feedback adjustment;

[0080] Specifically, when generating intervention strategies, the personalized intervention decision-making unit integrates test results and user personality characteristics to construct a decision mapping model with a logic function as the core. It first performs a weighted summation of the user's multi-dimensional features to obtain a comprehensive score, which is then combined with the abnormality indication signal and passed to the logic function through a linear combination to generate a normalized intervention strategy that not only reflects the current abnormality risk but also integrates individual differences.

[0081] The dynamic adjustment mechanism uses actual intervention feedback to continuously correct various decision parameters, so that the strategy output is consistent with the user's actual status, and further improves the system adaptability and intervention effect through feedback adjustment.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based health management system for sub-healthy population prevention programs, characterized by: include, Data collection unit, which obtains user behavior data and health data information from the health management link; The health management process includes smart wearable devices, mobile terminals and regular physical examinations; a data integration unit for fusing the health data information to form a user health information database; A user behavior profile generating unit, which generates a behavior profile within a predetermined period based on a user health information database; An anomaly detection unit analyzes the behavior profile based on a preset behavior pattern model, determines whether the user's records are missing or below an expected level within a predetermined period, and generates an anomaly indication signal; a risk assessment unit, combining the abnormal indication signal with the user's historical health data to determine the user's health risk level by comparison; The personalized intervention decision unit outputs an intervention strategy that matches the user's actual situation based on the risk level and the user's personalized information.

2. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 1, characterized in that: The behavior pattern model includes a continuity expectation model constructed based on statistics and machine learning methods, which is used to dynamically predict and judge the missing of user behavior data.

3. The artificial intelligence-based health management system for sub-healthy population prevention program according to claim 2, characterized in that: In the anomaly detection unit, a preset behavior pattern model is established to predict the records that the user should have at a specific time, so as to determine whether there is a record missing phenomenon. The model expression is: in, represents the expected record value at time t, i represents the index of the historical record, d i represents the weight of the i-th historical record, x t-i represents the actual recorded value at time ti, e represents the bias parameter of the model, and n represents the number of historical records involved in the model calculation; The least squares method is used to train the model parameters, and the objective function used is: Among them, x t Indicates the actual recorded value at time t, t represents the time index, n represents the number of historical records involved in the model calculation, d i represents the weight of the i-th historical record, x t-i Represents the actual recorded value at time ti, e represents the bias parameter of the model, and T represents the end time of the record during the training period.

4. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 3, characterized in that: In the anomaly detection unit, during the anomaly detection phase, an anomaly indication signal is generated by comparing the actual record with the expected record, which is mathematically expressed as: Among them, I t Indicates the abnormal indication signal at time t. If the condition is met, the value is 1, otherwise it is 0. t represents the actual recorded value at time t, c represents the preset abnormality judgment threshold factor, represents the expected record value at time t, 1{·} represents an indicator function, which takes 1 if the conditions in the braces are met and 0 otherwise.

5. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 4, characterized in that: In the anomaly detection stage, the threshold factor is dynamically adapted, and its calculation formula is: c = 1-δ / σ x , where c represents the preset abnormality judgment threshold factor, δ represents the allowable deviation range, σ x Indicates the standard deviation of historical data.

6. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 1, characterized in that: The personalized intervention decision-making unit further integrates data from the health management process and issues proactive warnings based on user silence data and lack of behavioral continuity; The data in the health management link also includes health self-test, risk prediction and health assessment data.

7. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 6, characterized in that: In the personalized intervention decision unit, an intervention strategy is generated by constructing a decision function according to the abnormal indication result of the abnormality detection unit and the user's personality characteristics, and the decision parameters are dynamically adjusted based on the feedback information.

8. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 7, characterized in that: In the personalized intervention decision-making unit, the process of determining the intervention strategy includes: The user's personality characteristics are weighted and summarized to generate an overall characteristic score, the mathematical expression of which is: Among them, θ represents the user's overall personality feature score, j is the index of the user feature, and w j represents the weight of the jth user feature, u j represents the jth user feature value, and m represents the total number of user features.

9. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 8, characterized in that: In the personalized intervention decision-making unit, the process of determining the intervention strategy also includes: The anomaly detection results are combined with the overall feature scores, and the intervention strategy is generated through decision mapping. The decision function adopts the form of a logical function, and its expression is: Among them, S t represents the personalized intervention strategy generated at time t, I t represents the abnormality indication signal at time t, refers to the detection result, α represents the weight of the abnormal signal in the decision, β represents the weight of the user's overall feature score in the decision, γ represents the decision bias, and exp(·) represents the natural exponential function, which is used here to implement the logistic mapping.

10. The artificial intelligence-based health management system for preventing sub-healthy people according to claim 9, characterized in that: In the personalized intervention decision-making unit, a feedback update mechanism is also established, assuming r t is the actual feedback value of the intervention effect, η is the learning rate, and the parameter update rule is: a new =a old +η(r t -S t ), b new =b old +η(r t -S t )θ, c new =c old +η(r t -S t ), w j,new =w j,old +η(r t -S t )u j , Among them, α old and α new Represent the abnormal signal weights before and after adjustment, β old and β new Represent the user feature weights before and after adjustment, γ old and γ new Respectively represent the decision bias before and after adjustment, w j,old and w j,new Respectively represent the values of the j-th user feature weight before and after adjustment, r t It represents the actual feedback of the intervention at time t, reflecting the effect of the intervention strategy. η represents the learning rate, which controls the step size of the feedback adjustment.

Images