Myopia prevention and control system and method based on multi-source data

Abstract

The invention discloses a myopia prevention and control system and method based on multi-source data, and relates to the technical field of myopia prevention and control, and the method comprises the steps: collecting multi-source data through an intelligent wearable device, a camera and an environment sensor; the method comprises the following steps: preprocessing collected multi-source data to form a structured time sequence data set; decomposing the structured time sequence data set into a plurality of time scales, and extracting time sequence feature values; according to a prediction result, defining a reinforcement learning environment, and using a deep reinforcement learning algorithm to generate a dynamic prevention and control strategy scheme; according to the generated dynamic prevention and control strategy scheme, eye using habit changes of the user are tracked for a long time, and a myopia prevention and control scheme is generated. According to the method, the structured time sequence data set is decomposed into a plurality of time scales by using wavelet transform, and the time sequence feature values are extracted, so that multi-level description of eye using behaviors of the user and environmental changes is realized, and the accuracy and reliability of myopia risk prediction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of myopia prevention and control, and in particular to a myopia prevention and control system and method based on multi-source data. Background Art

[0002] In recent years, with the rapid development of information technology, especially the advancement of hardware technologies such as smart wearable devices, cameras and environmental sensors, data collection and analysis have been widely used in the field of medical health. Myopia prevention and control, as an important part of youth health, has received more and more attention. Traditional myopia prevention and control methods mainly rely on regular eye examinations and professional advice from doctors. Although this method is effective, it has certain limitations. First, the time interval between regular eye examinations is long, and it is impossible to monitor the user's eye habits and environmental changes in real time; second, the doctor's professional advice is usually based on static data, which is difficult to dynamically adjust according to individual differences.

[0003] Although existing myopia prevention and control methods have made certain progress, there are still some shortcomings. First, existing technologies still have limitations in data collection. Traditional myopia prevention and control methods usually rely only on a single data source, such as ophthalmic examination results or users' self-reported eye habits, and lack comprehensive analysis of multiple data sources. This leads to the one-sidedness and incompleteness of the data, making it difficult to fully reflect the user's actual eye use. For example, relying solely on ophthalmic examinations cannot capture subtle changes in the user's eye use duration, frequency, rest intervals, and other subtle changes in daily life, which are crucial for assessing myopia risk. Summary of the invention

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

[0005] Therefore, the present invention provides a myopia prevention and control method based on multi-source data to solve the problems of single data source and inability to effectively extract multi-time scale features in the prior art.

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

[0007] In the first aspect, the present invention provides a myopia prevention and control method based on multi-source data, which includes collecting multi-source data through smart wearable devices, cameras and environmental sensors; preprocessing the collected multi-source data to form a structured time series data set; decomposing the structured time series data set into multiple time scales, and extracting time series eigenvalues; based on the time series eigenvalues, weightedly fusing the eigenvalues ​​of different time scales to predict the user's myopia risk and generate a prediction result; based on the prediction result, defining a reinforcement learning environment, using a deep reinforcement learning algorithm, and generating a dynamic prevention and control strategy plan; based on the generated dynamic prevention and control strategy plan, tracking the changes in the user's eye habits over a long period of time to generate a myopia prevention and control plan.

[0008] As a preferred solution of the myopia prevention and control method based on multi-source data described in the present invention, the multi-source data includes eye behavior data, environmental data, image data and physiological data.

[0009] As a preferred solution of the myopia prevention and control method based on multi-source data of the present invention, the collected multi-source data are pre-processed to form a structured time series data set. The specific steps are as follows:

[0010] Based on the collected multi-source data, Python is used to read the data and delete missing and duplicate values. The Scikit-learn library standardizes the numerical data.

[0011] The standardized multi-source data are organized into a structured format suitable for time series analysis and integrated to form a structured time series data set.

[0012] As a preferred solution of the myopia prevention and control method based on multi-source data of the present invention, the structured time series data set is decomposed into multiple time scales, and the time series feature values ​​are extracted. The specific steps are as follows:

[0013] The wavelet transform is used to perform multi-scale decomposition on the structured time series data set. The expression is:

[0014]

[0015] Where W(a,b) represents the decomposition of the structured time series data set under the time scale parameter a and the translation parameter b, a represents the scale parameter, b represents the translation parameter, D(t) represents the comprehensive eigenvalue of the preprocessed multi-source data at time t, ψ(t) represents the wavelet basis function at time t, dt represents the integration of the variable at time t, and t represents the time point of the time series data;

[0016] By decomposing the structured time series data set, we can obtain the time series features at different time scales;

[0017] Extract the time series feature values according to the time series characteristics at different time scales. The expression is as follows:

[0018]

[0019] Among them, F k (a) represents the time series feature value of the k-th type of feature at the time scale a, and δ i represents the weight coefficient of the i-th time series feature, n represents the number of time series features, and σ i represents the non-linear activation function of the i-th time series feature, and φ i (t) represents the kernel function of the i-th time series feature at time t, k represents the index of feature classification, and i represents the index variable of the time series feature.

[0020] As a preferred solution of the myopia prevention and control method based on multi-source data according to the present invention, wherein: based on the time series feature values, the feature values of different time scales are weighted and fused to predict the myopia risk of the user and generate a prediction result. The specific steps are as follows:

[0021] Based on the time series feature values, the feature values of different time scales are weighted and fused to generate a comprehensive myopia risk prediction value. The expression is as follows:

[0022]

[0023] Among them, P(t) represents the comprehensive myopia risk prediction value at time t, m represents the number of time scales, and w ν represents the weight coefficient of the ν-th time scale, and ν represents the index of the time scale;

[0024] Set the low-risk threshold M and the high-risk threshold N according to the historical risk division data;

[0025] Based on the myopia risk prediction value, divide the risk level through the low-risk threshold M and the high-risk threshold N to predict the myopia risk of the user and generate a prediction result;

[0026] If P(t) < M, it is determined as low myopia risk;

[0027] If M ≤ P(t) < N, it is determined as medium myopia risk;

[0028] If P(t) ≥ N, it is determined as high myopia risk.

[0029] As a preferred solution of the myopia prevention and control method based on multi-source data according to the present invention, wherein: according to the prediction result, define the reinforcement learning environment and use the deep reinforcement learning algorithm to generate a dynamic prevention and control strategy plan. The specific steps are as follows:

[0030] According to the prediction results, the key features that affect myopia risk are determined, including the user's current myopia risk prediction value, eye behavior characteristics, environmental characteristics, and user physiological characteristics, which are combined into a state vector to define the state space;

[0031] Determine the actions to be taken to improve the user's eye habits and environmental conditions, including adjusting the interval of eye time reminders, optimizing the indoor light intensity and color temperature, adding outdoor activity suggestions, and adjusting the screen blue light ratio. The actions are formalized into a set and the action space is defined;

[0032] Taking the change in the user's myopia risk as the main optimization target, the reward value is defined as follows:

[0033]

[0034] Where R represents the reward value, ΔP(t) represents the change in the user's comprehensive myopia risk prediction value within the time interval Δt, ξ(A) represents the execution cost function of the action set A, η represents the user's acceptance feedback function of the strategy, and λ 1 represents the risk change weight, λ 2 represents the execution cost weight, λ 3 represents the weight of user feedback on the reward value, and A represents the action set;

[0035] Based on the defined state space, action space and reward value, a deep reinforcement learning algorithm is used to collect experience through environmental interaction, and the experience replay mechanism is used to obtain the best action to generate a dynamic prevention and control strategy plan.

[0036] As a preferred solution of the myopia prevention and control method based on multi-source data of the present invention, wherein: according to the generated dynamic prevention and control strategy solution, the user's eye habits changes are tracked for a long time to generate a myopia prevention and control solution, the specific steps are as follows:

[0037] According to the generated dynamic prevention and control strategy, the action value of each eye behavior feature is calculated, and the expression is:

[0038]

[0039] Among them, V(U,t) represents the action value of eye behavior feature U at time t, U represents the eye behavior feature, α represents the total number of actions in the action space, and β j represents the weight coefficient of the action in the jth spatial action, and j represents the index variable of the action in the spatial action;

[0040] Based on the action value of eye behavior characteristics, a myopia prevention and control plan is generated through weighted fusion, which is expressed as:

[0041]

[0042] Among them, O(t) represents the myopia prevention and control plan at time t, represents the number of eye behavior characteristics, ω f represents the weight coefficient of the f-th eye behavior feature, and f represents the index variable of the eye behavior feature.

[0043] In the second aspect, the present invention provides a myopia prevention and control system based on multi-source data, including a multi-source data acquisition module, a time series data set formation module, a time series feature value extraction module, a prediction result generation module, a dynamic prevention and control strategy scheme generation module and a myopia prevention and control scheme generation module; the multi-source data acquisition module is used to collect multi-source data through smart wearable devices, cameras and environmental sensors; the time series data set formation module is used to pre-process the collected multi-source data to form a structured time series data set; the time series feature value extraction module is used to decompose the structured time series data set into multiple time scales and extract time series feature values; the prediction result generation module is used to weightedly fuse the feature values ​​of different time scales based on the time series feature values, predict the user's myopia risk, and generate a prediction result; the dynamic prevention and control strategy scheme generation module is used to define a reinforcement learning environment according to the prediction results, and use a deep reinforcement learning algorithm to generate a dynamic prevention and control strategy scheme; the myopia prevention and control scheme generation module is used to track the changes in the user's eye habits over a long period of time according to the generated dynamic prevention and control strategy scheme, and generate a myopia prevention and control plan.

[0044] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the myopia prevention and control method based on multi-source data as described in the first aspect of the present invention is implemented.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the myopia prevention and control method based on multi-source data as described in the first aspect of the present invention.

[0046] The beneficial effects of the present invention are as follows: by using wavelet transform to decompose the structured time series data set into multiple time scales and extracting the time series eigenvalues, a multi-level description of the user's eye behavior and environmental changes is achieved, thereby improving the accuracy and reliability of myopia risk prediction. Secondly, a reinforcement learning environment is defined based on the prediction results, and a deep reinforcement learning algorithm is used to generate a dynamic prevention and control strategy plan, so that the system can dynamically adjust the prevention and control measures according to the user's real-time data and provide personalized myopia prevention and control recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0048] Figure 1 This is a flow chart of the myopia prevention and control method based on multi-source data in Example 1.

[0049] Figure 2 This is a schematic diagram of the myopia prevention and control system based on multi-source data in Example 1. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and 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.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0053] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a myopia prevention and control method based on multi-source data, comprising the following steps:

[0054] S1, collect multi-source data through smart wearable devices, cameras and environmental sensors;

[0055] The multi-source data includes eye behavior data, environmental data, image data and physiological data.

[0056] S2, preprocessing the collected multi-source data to form a structured time series data set;

[0057] Based on the collected multi-source data, Python is used to read the data and delete missing and duplicate values. The Scikit-learn library standardizes the numerical data.

[0058] It should be noted that missing values ​​and duplicate values ​​were deleted to ensure the integrity and uniqueness of the data. Numerical data were standardized through the Scikit-learn library so that data from different sources had the same dimension, which facilitated the subsequent time series analysis.

[0059] Organize the standardized multi-source data into a structured format suitable for time series analysis and merge them to form a structured time series data set;

[0060] It should be noted that the data from different sources are aligned according to the timestamp to ensure that all data are consistent at the same time point. The multi-source data are merged into a unified DataFrame to form a structured time series data set, which is convenient for subsequent time series analysis.

[0061] S3, decompose the structured time series data set into multiple time scales and extract time series feature values;

[0062] The wavelet transform is used to perform multi-scale decomposition on the structured time series data set. The expression is:

[0063]

[0064] Where W(a,b) represents the decomposition of the structured time series data set under the time scale parameter a and the translation parameter b, a represents the scale parameter, b represents the translation parameter, D(t) represents the comprehensive eigenvalue of the preprocessed multi-source data at time t, ψ(t) represents the wavelet basis function at time t, dt represents the integration of the variable at time t, and t represents the time point of the time series data;

[0065] It should be noted that a controls the scaling degree of the wavelet basis function, a larger a value corresponds to a lower frequency, and a smaller a value corresponds to a higher frequency; b controls the position of the wavelet basis function on the time axis; ψ(t) is usually an oscillating function with finite energy; D(t) is usually a standardized numerical data, representing the user's eye behavior, environmental conditions or physiological state at a certain moment; The integration operation multiplies the wavelet basis function with the original signal and sums them over the entire time range.

[0066] By decomposing the structured time series data set, we can obtain the time series features at different time scales;

[0067] It should be noted that the number of decomposition layers indicates how many components of different time scales the signal is decomposed into. A higher number of decomposition layers means more scale components, but it also increases the computational complexity. By capturing feature changes at different time scales, the user's myopia risk can be predicted more accurately.

[0068] According to the time series characteristics at different time scales, the time series characteristic values ​​are extracted, and the expression is:

[0069]

[0070] Among them, F k (a) represents the time series eigenvalue of the kth feature at time scale a, δ i represents the weight coefficient of the i-th time series feature, n represents the number of time series features, σ i Represents the nonlinear activation function of the i-th time series feature, φ i (t) represents the kernel function of the i-th time series feature at time t, k represents the index of the feature classification, and i represents the index variable of the time series feature;

[0071] It should be noted that i (t) is usually a set of orthogonal basis functions used to extract useful features from the decomposed coefficients; σ i Used to perform nonlinear transformation on the inner product result.

[0072] S4. Based on the time series eigenvalues, the eigenvalues ​​of different time scales are weighted and fused to predict the user's myopia risk and generate a prediction result;

[0073] Based on the time series eigenvalues, the eigenvalues ​​of different time scales are weighted and fused to generate a comprehensive myopia risk prediction value, which is expressed as:

[0074]

[0075] Where P(t) represents the comprehensive myopia risk prediction value at time t, m represents the number of time scales, and w ν represents the weight coefficient of the νth time scale, ν represents the index of the time scale;

[0076] It should be noted that F k (a) The eigenvalues ​​reflect the user’s eye behavior, environmental conditions, and physiological state at different time scales; w ν Indicates the importance of each time scale. Different time scales may have different impacts on myopia risk, so it is necessary to assign appropriate weights to each time scale. Weighted fusion is to obtain the comprehensive myopia risk prediction value by multiplying the feature values ​​of different time scales by the corresponding weight coefficients and normalizing them.

[0077] According to the historical risk classification data, set the low risk threshold M and the high risk threshold N;

[0078] It should be noted that historical risk classification data refers to the results of users' myopia risk assessments collected over a past period of time. The data usually includes users' eye-using behaviors, environmental conditions, physiological states, and the corresponding myopia risk levels.

[0079] Based on the myopia risk prediction value, the risk level is divided by the low-risk threshold M and the high-risk threshold N to predict the myopia risk of the user and generate a prediction result.

[0080] If P(t) < M, it is determined as a low myopia risk.

[0081] If M ≤ P(t) < N, it is determined as a medium myopia risk.

[0082] If P(t) ≥ N, it is determined as a high myopia risk.

[0083] It should be noted that when the myopia risk is low, the user's myopia risk is relatively low, indicating that their current eye-using behaviors, environmental conditions, and physiological states are relatively healthy; when the myopia risk is medium, the user's myopia risk is at a medium level, indicating that some of their eye-using behaviors or environmental conditions may need to be adjusted; when the myopia risk is high, the user's myopia risk is relatively high, indicating that there may be relatively large problems with their current eye-using behaviors, environmental conditions, or physiological states.

[0084] S5. According to the prediction result, define the reinforcement learning environment and use the deep reinforcement learning algorithm to generate a dynamic prevention and control strategy plan.

[0085] According to the prediction result, determine the key features affecting the myopia risk, including the user's current myopia risk prediction value, eye-using behavior features, environmental features, and user physiological features, combine them into a state vector, and define the state space.

[0086] It should be noted that the eye-using behavior features describe the user's eye-using habits over a period of time, such as: the time of using electronic devices every day, the time of reading or doing close work, the frequency and duration of eye rest; the environmental features describe the impact of the user's environment on the eyes, such as: indoor light intensity, screen brightness and contrast, environmental noise level; the user physiological features describe the user's physical condition and physiological parameters, such as: pupil size change, vision change trend, sleep quality and duration. Combining the above key features into a state vector can comprehensively describe the user's current state.

[0087] Determine the actions to be taken to improve the user's eye-using habits and environmental conditions, including adjusting the reminder interval of eye-using time, optimizing the indoor light intensity and color temperature, increasing the suggestion of outdoor activities, and adjusting the screen blue light ratio. The actions are formalized into a set to define the action space.

[0088] It should be noted that the interval of reminders for eye use time should be adjusted. According to the characteristics of the user's eye behavior, the time interval for reminding users to rest their eyes should be dynamically adjusted. Optimize the intensity and color temperature of indoor light. Adjust the intensity and color temperature of indoor light to provide a more suitable eye environment. Increase outdoor activity suggestions to encourage users to increase outdoor activity time and reduce close-up eye use time, which will help reduce the risk of myopia. Adjust the proportion of blue light on the screen. Reduce the proportion of blue light on the screen to reduce stimulation to the eyes, especially when using electronic devices at night. Formalize the above actions into a set and define the action space.

[0089] Taking the change in the user's myopia risk as the main optimization target, the reward value is defined as follows:

[0090]

[0091] Where R represents the reward value, ΔP(t) represents the change in the user's comprehensive myopia risk prediction value within the time interval Δt, ξ(A) represents the execution cost function of the action set A, η represents the user's acceptance feedback function of the strategy, and λ 1 represents the risk change weight, λ 2 represents the execution cost weight, λ 3 represents the weight of user feedback on the reward value, and A represents the action set;

[0092] It should be noted that a negative value of ΔP(t) indicates that the user's myopia risk is reduced, and a positive value of ΔP(t) indicates that the user's myopia risk is increased; λ 1 Used to adjust the impact of risk changes on reward values. A larger λ 1 It means that the system pays more attention to the changes in myopia risk; ξ(A) cost function includes factors such as time, energy, and economy; a higher η means that the user is more satisfied with the current strategy, and a lower η means that the user is not very satisfied.

[0093] Based on the defined state space, action space and reward value, a deep reinforcement learning algorithm is used to collect experience through environmental interaction, and the experience replay mechanism is used to obtain the best action to generate a dynamic prevention and control strategy plan;

[0094] It should be noted that the experience replay mechanism is used to improve learning efficiency and stability. The agent stores the experience (state, action, reward, next state) of each interaction with the environment in an experience pool, and randomly extracts a batch of experiences for learning during training. This helps to break the correlation between data and improve the generalization ability of the model.

[0095] S6. Based on the generated dynamic prevention and control strategy, track the user's eye habits over a long period of time and generate a myopia prevention and control plan;

[0096] According to the generated dynamic prevention and control strategy, the action value of each eye behavior feature is calculated, and the expression is:

[0097]

[0098] Among them, V(U,t) represents the action value of eye behavior feature U at time t, U represents the eye behavior feature, α represents the total number of actions in the action space, and β j represents the weight coefficient of the action in the jth spatial action, and j represents the index variable of the action in the spatial action;

[0099] It should be noted that the action value V(U,t) is a comprehensive indicator used to measure the overall effect of a series of actions taken for a certain eye behavior feature U at a specific time point t. The higher the value, the better the improvement effect of these actions on the eye behavior feature. j Indicates the importance or priority of each action. Different actions may have different effects on the user's eye behavior, so it is necessary to assign a suitable weight coefficient to each action.

[0100] Based on the action value of eye behavior characteristics, a myopia prevention and control plan is generated through weighted fusion, which is expressed as:

[0101]

[0102] Among them, O(t) represents the myopia prevention and control plan at time t, represents the number of eye behavior characteristics, ω f represents the weight coefficient of the f-th eye behavior feature, and f represents the index variable of the eye behavior feature;

[0103] It should be noted that O(t) is used to measure the overall effect of a series of actions taken for all eye behavior characteristics at a specific time point t. The higher this value is, the better the effect of these actions on improving the user's eye habits and reducing the risk of myopia. Eye behavior characteristics include the time spent using electronic devices, reading or working at close range, and the frequency and duration of eye rests per day; f It represents the importance or priority of each eye behavior feature. Different eye behavior features may have different effects on the user's myopia risk, so it is necessary to assign a suitable weight coefficient to each feature.

[0104] The present embodiment also provides a myopia prevention and control system based on multi-source data, including: a multi-source data acquisition module, a time series data set formation module, a time series feature value extraction module, a prediction result generation module, a dynamic prevention and control strategy scheme generation module and a myopia prevention and control scheme generation module; the multi-source data acquisition module is used to collect multi-source data through smart wearable devices, cameras and environmental sensors; the time series data set formation module is used to pre-process the collected multi-source data to form a structured time series data set; the time series feature value extraction module is used to decompose the structured time series data set into multiple time scales and extract time series feature values; the prediction result generation module is used to weightedly fuse the feature values ​​of different time scales based on the time series feature values, predict the user's myopia risk, and generate a prediction result; the dynamic prevention and control strategy scheme generation module is used to define a reinforcement learning environment according to the prediction results, and use a deep reinforcement learning algorithm to generate a dynamic prevention and control strategy scheme; the myopia prevention and control scheme generation module is used to track the changes in the user's eye habits over a long period of time according to the generated dynamic prevention and control strategy scheme, and generate a myopia prevention and control scheme.

[0105] This embodiment also provides a computer device, which is suitable for the case of a myopia prevention and control method based on multi-source data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the myopia prevention and control method based on multi-source data proposed in the above embodiment.

[0106] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0107] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the myopia prevention and control method based on multi-source data proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0108] In summary, the present invention achieves a multi-level description of user eye behavior and environmental changes by using wavelet transform to decompose structured time series data sets into multiple time scales and extract time series eigenvalues, thereby improving the accuracy and reliability of myopia risk prediction. Secondly, a reinforcement learning environment is defined based on the prediction results, and a deep reinforcement learning algorithm is used to generate a dynamic prevention and control strategy plan, so that the system can dynamically adjust prevention and control measures according to the user's real-time data and provide personalized myopia prevention and control recommendations.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. 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. A myopia prevention and control method based on multi-source data, characterized by: including, collecting multi-source data through smart wearable devices, cameras, and environmental sensors; preprocessing the collected multi-source data to form a structured time series dataset; decomposing the structured time series dataset into multiple time scales and extracting time series feature values; weightedly fusing the feature values of different time scales based on the time series feature values to predict the myopia risk of the user and generate a prediction result; defining a reinforcement learning environment according to the prediction result and using a deep reinforcement learning algorithm to generate a dynamic prevention and control strategy plan; generating a myopia prevention and control plan by long-term tracking of the changes in the user's eye use habits according to the generated dynamic prevention and control strategy plan.

2. The myopia prevention and control method based on multi-source data according to claim 1, characterized in that: The multi-source data includes eye use behavior data, environmental data, image data, and physiological data.

3. The myopia prevention and control method based on multi-source data according to claim 2, characterized in that: The preprocessing of the collected multi-source data to form a structured time series dataset is specifically as follows. Based on the collected multi-source data, use Python to read the data and delete missing values and duplicate values, and use the Scikit-learn library to standardize the numerical data; Organize the standardized multi-source data into a structured format suitable for time series analysis for fusion to form a structured time series dataset.

4. The myopia prevention and control method based on multi-source data according to claim 3, characterized in that: The decomposition of the structured time series dataset into multiple time scales and the extraction of time series feature values are specifically as follows. Use wavelet transform to perform multi-scale decomposition on the structured time series dataset, and the expression is: where W(a, b) represents the decomposition of the structured time series dataset under the time scale parameter a and the translation parameter b, a represents the scale parameter, b represents the translation parameter, D(t) represents the comprehensive feature value of the multi-source data after preprocessing at time t, ψ(t) represents the wavelet basis function at time t, dt represents the integration of the variable of time t, and t represents the time point of the time series data; Obtain time series features at different time scales through the decomposition of the structured time series dataset; Extract time series feature values according to the time series features at different time scales, and the expression is: Among them, F k (a) represents the time series eigenvalue of the kth feature at time scale a, δ i represents the weight coefficient of the i-th time series feature, n represents the number of time series features, σ i Represents the nonlinear activation function of the i-th time series feature, φ i (t) represents the kernel function of the i-th time series feature at time t, k represents the index of the feature classification, and i represents the index variable of the time series feature.

5. The myopia prevention and control method based on multi-source data according to claim 4, characterized in that: The weighted fusion of the feature values of different time scales based on the time series feature values to predict the myopia risk of the user and generate a prediction result is specifically as follows. Based on the time series feature values, weightedly fuse the feature values of different time scales to generate a comprehensive myopia risk prediction value, and the expression is: Where P(t) represents the comprehensive myopia risk prediction value at time t, m represents the number of time scales, and w ν represents the weight coefficient of the νth time scale, ν represents the index of the time scale; Set a low risk threshold M and a high risk threshold N according to the historical risk division data; Based on the myopia risk prediction value, divide the risk level through the low risk threshold M and the high risk threshold N to predict the myopia risk of the user and generate a prediction result; If P(t) < M, it is determined as low myopia risk; If M ≤ P(t) < N, it is determined as medium myopia risk; If P(t) ≥ N, it is determined as high myopia risk.

6. The myopia prevention and control method based on multi-source data according to claim 5, characterized in that: The definition of a reinforcement learning environment according to the prediction result and the use of a deep reinforcement learning algorithm to generate a dynamic prevention and control strategy plan are specifically as follows. According to the prediction results, the key features that affect myopia risk are determined, including the user's current myopia risk prediction value, eye behavior characteristics, environmental characteristics, and user physiological characteristics, which are combined into a state vector to define the state space; Determine the actions to be taken to improve the user's eye habits and environmental conditions, including adjusting the interval of eye time reminders, optimizing the indoor light intensity and color temperature, adding outdoor activity suggestions, and adjusting the screen blue light ratio. The actions are formalized into a set and the action space is defined; Taking the change in the user's myopia risk as the main optimization target, the reward value is defined as follows: Where R represents the reward value, ΔP(t) represents the change in the user's comprehensive myopia risk prediction value within the time interval Δt, ξ(A) represents the execution cost function of the action set A, η represents the user's acceptance feedback function of the strategy, λ1 represents the risk change weight, λ2 represents the execution cost weight, λ3 represents the weight of the user feedback on the reward value, and A represents the action set; Based on the defined state space, action space and reward value, a deep reinforcement learning algorithm is used to collect experience through environmental interaction, and the experience replay mechanism is used to obtain the best action to generate a dynamic prevention and control strategy plan.

7. The myopia prevention and control method based on multi-source data according to claim 6, characterized in that: According to the generated dynamic prevention and control strategy, the user's eye habits are tracked over a long period of time to generate a myopia prevention and control plan. The specific steps are as follows: According to the generated dynamic prevention and control strategy, the action value of each eye behavior feature is calculated, and the expression is: Among them, V(U,t) represents the action value of eye behavior feature U at time t, U represents the eye behavior feature, α represents the total number of actions in the action space, and β j represents the weight coefficient of the action in the jth spatial action, and j represents the index variable of the action in the spatial action; Based on the action value of eye behavior characteristics, a myopia prevention and control plan is generated through weighted fusion, which is expressed as: Among them, O(t) represents the myopia prevention and control plan at time t, represents the number of eye behavior characteristics, ω f represents the weight coefficient of the f-th eye behavior feature, and f represents the index variable of the eye behavior feature.

8. A myopia prevention and control system based on multi-source data, based on the myopia prevention and control method based on multi-source data according to any one of claims 1 to 7, characterized in that: It includes a multi-source data acquisition module, a time series data set formation module, a time series feature value extraction module, a prediction result generation module, a dynamic prevention and control strategy program generation module, and a myopia prevention and control program generation module; Multi-source data acquisition module, used to collect multi-source data through smart wearable devices, cameras and environmental sensors; The time series data set formation module is used to pre-process the collected multi-source data to form a structured time series data set; The time series feature value extraction module is used to decompose the structured time series data set into multiple time scales and extract the time series feature values; A prediction result generation module is used to weight and fuse the feature values ​​of different time scales based on the time series feature values, predict the user's myopia risk, and generate a prediction result; The dynamic prevention and control strategy generation module is used to define the reinforcement learning environment based on the prediction results and use the deep reinforcement learning algorithm to generate dynamic prevention and control strategy solutions; The myopia prevention and control plan generation module is used to generate a myopia prevention and control plan based on the generated dynamic prevention and control strategy plan and long-term tracking of changes in users' eye habits.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the myopia prevention and control method based on multi-source data described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the myopia prevention and control method based on multi-source data described in any one of claims 1 to 7 are implemented.

Images