Children's Myopia Prevention and Control Prediction System and Method Based on Data Mining

Abstract

The present invention discloses a prediction system and method for preventing and controlling children's myopia based on data mining, which relates to the technical fields of health management and data mining. It includes collecting the dynamic modal data and static modal data of children, and preprocessing the dynamic modal data and static modal data; extracting the dynamic modal features and static modal features of children based on the two kinds of preprocessed modal data; generating personalized myopia risk features by using a multimodal fusion algorithm based on the dynamic modal features and static modal features of children; constructing a myopia risk prediction model, and predicting the myopia probability of children based on the personalized myopia risk features; generating a personalized prevention and control strategy according to the myopia probability of children. The present invention constructs a prediction model by using a deep neural network, converts these features into specific myopia probabilities, provides high-precision personalized prediction, and supports early warning and customized prevention and control strategies.

Description

Technical Field

[0001] The present invention relates to the technical fields of health management and data mining, and particularly to a child myopia prevention and control prediction system and method based on data mining. Background Art

[0002] With the rapid development of information technology and the wide application of data mining technology, data-driven health management methods have gradually become a research hotspot. Child myopia, as a globally prevalent public health problem, has been showing a continuous upward trend in its incidence rate in recent years. Especially under the influence of the popularization of electronic devices and the increasing academic burden, the trend of myopia at a younger age has become even more obvious. Currently, predicting the risk of myopia occurrence and formulating prevention and control strategies by combining data mining technology with children's biometric and behavioral data has become an important research direction for solving this problem. Existing research mainly focuses on the analysis and modeling of single-modal data (such as physiological indicators like axial length of the eye, refractive power, etc.) or behavioral statistical data (such as reading duration, screen usage time, etc.). However, since the occurrence and development of child myopia are the result of the combined action of multiple factors, the single-modal data processing method often fails to accurately capture the complex interaction relationships between multi-dimensional features, thus limiting the accuracy of the myopia risk prediction model and the personalization level of the prevention and control plan.

[0003] Although traditional myopia prediction methods based on statistical analysis or machine learning have achieved the assessment of myopia risk to a certain extent, the existing technologies still have significant deficiencies. First, the existing methods have weak capabilities in fusing and processing multi-modal data and cannot fully utilize the complementary information between dynamic modal data (such as blink frequency, head posture changes) and static modal data (such as axial length of the eye, refractive power, etc.). Dynamic modal data can reflect the real-time changes in children's daily eye use behaviors, while static modal data reflects the long-term trends of children's physiological characteristics. When the two types of data are not effectively fused, the prediction accuracy of the model is relatively low. Second, most current myopia risk prediction models adopt simple linear regression or traditional machine learning methods, which are difficult to capture the non-linear relationships between multi-modal features, resulting in inaccurate assessment results of myopia risk, which further affects the pertinence and effectiveness of the prevention and control strategies. Summary of the Invention

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

[0005] Therefore, the present invention provides a child myopia prevention and control prediction system and method based on data mining, which solves the problems of insufficient multi-modal data fusion and limited non-linear feature capture ability in the prior art.

[0006] The solutions are as follows:

[0007] In a first aspect, the present invention provides a method for predicting and preventing myopia in children based on data mining, which includes: collecting dynamic modal data and static modal data of children, and preprocessing the two types of modal data. The dynamic modal data includes the blinking frequency, pupil diameter change, gaze time distribution, and head pose change of children, and the static modal data includes the axial length of the eye, refractive power, and reading distance of children; extracting dynamic modal features and static modal features of children based on the preprocessed two types of modal data, calculating the dot product of each dynamic modal feature and static modal feature with a weight vector, assigning weights to each dynamic modal feature and static modal feature, and using a multi-modal fusion algorithm to perform feature interaction calculation to generate personalized myopia risk features; constructing a myopia risk prediction model based on a deep neural network, and predicting the myopia probability of children based on the personalized myopia risk features to generate a personalized prevention and control strategy.

[0008] As a preferred embodiment of the method for predicting and preventing myopia in children based on data mining according to the present invention, wherein: the preprocessing of the two types of modal data is specifically as follows: removing outliers in the dynamic modal data by the box plot method; smoothing the dynamic modal data using Gaussian filtering; converting the static modal data by standardization processing to eliminate different dimensions; performing data dimensionality reduction processing on the static modal data by the principal component analysis method.

[0009] As a preferred embodiment of the method for predicting and preventing myopia in children based on data mining according to the present invention, wherein: the extraction of dynamic modal features and static modal features of children based on the preprocessed two types of modal data is specifically as follows:

[0010] Performing time series analysis on the blinking frequency data to extract the average blinking frequency feature of children within a unit time;

[0011] For the pupil diameter change data, using Fourier transform to extract the frequency components and fluctuation amplitude features of the pupil diameter change data of children;

[0012] For the head pose change data, extracting the stability feature of the head pose of children within a unit time;

[0013] Performing time series analysis on the axial length data of the eye to extract the growth feature of the axial length of the eye of children;

[0014] Statistically analyzing the change amplitude and direction of the refractive power to extract the refractive power change feature of children;

[0015] According to the preprocessed reading distance, using statistical analysis methods to extract the reading distance variability feature of children.

[0016] As a preferred solution of the method for predicting the prevention and control of myopia in children based on data mining according to the present invention, wherein: weight assignment is performed on the dynamic modal features, and the expression is:

[0017] ;

[0018] wherein, is the weight coefficient of the th dynamic modal feature, is the value of the th dynamic modal feature, is the total number of dynamic modal features, is the transpose of the weight vector of the dynamic modal features, is the index variable of the dynamic modal features in the normalization process, is the value of the th dynamic modal feature in the normalization process;

[0019] Weight assignment is performed on the static modal features, and the expression is:

[0020] ;

[0021] Based on the dynamic modal features, wherein, is the weight coefficient of the th static modal feature, is the value of the th static modal feature, is the total number of static modal features, is the transpose of the weight vector of the static modal features, is the index variable of the static modal features in the normalization process, is the value of the th static modal feature in the normalization process;

[0022] The weight coefficient and the weight coefficient of the static modal features are used to perform feature interaction calculation by using a multi-modal fusion algorithm to capture the non-linear relationship between each dynamic modal feature and the static modal feature, and generate personalized myopia risk features.

[0023] As a preferred solution of the method for predicting the prevention and control of myopia in children based on data mining according to the present invention, wherein: the expression of the personalized myopia risk feature is:

[0024] ;

[0025] wherein, is the personalized myopia risk feature, is the value of the th dynamic modal feature The eigenvalue obtained after non - linear transformation, is the value of the th static modal feature The eigenvalue obtained after non - linear transformation, is a positive constant that regulates the differential influence of dynamic and static modal features, is the original value of the th modal feature, is the mean value of the th modal feature,

[0026] As a preferred solution of the method for predicting and preventing myopia in children based on data mining according to the present invention, wherein: the myopia risk prediction model is constructed based on a deep neural network, and the myopia probability of children is predicted based on personalized myopia risk characteristics. The specific steps are as follows:

[0027] Use DNN as the framework of the myopia risk prediction model;

[0028] Based on the framework of the myopia risk prediction model, construct an input layer, a first hidden layer, a second hidden layer, and an output layer;

[0029] The first hidden layer receives the personalized myopia risk characteristics of the input layer , performs a linear transformation on the personalized myopia risk characteristics according to the weight matrix and the bias vector , and performs an activation process through the ReLU activation function to generate a preliminary feature representation;

[0030] The second hidden layer receives the preliminary feature representation, repeats the linear transformation and the activation process of the ReLU activation function, and generates a deep - level feature representation;

[0031] The output layer receives the deep - level feature representation and performs an activation process through the Sigmoid activation function;

[0032] Combine the input layer, the first hidden layer, the second hidden layer, the output layer, and the framework of the myopia risk prediction model to construct a myopia risk prediction model;

[0033] Based on the personalized myopia risk characteristics, predict the myopia probability of children. The expression is:

[0034] ;

[0035] Wherein, is the myopia probability of children, is the weight matrix of the first hidden layer, is the weight matrix of the second hidden layer, the weight matrix of the output layer, is the bias vector of the first hidden layer, is the bias vector of the second hidden layer, is the bias vector of the output layer.

[0036] As a preferred solution of the method for predicting and preventing myopia in children based on data mining according to the present invention, wherein: the steps of generating a personalized prevention and control strategy are as follows:

[0037] Based on the statistical analysis of the probability distributions of low risk and high risk in the historical myopia population data, define a low risk threshold P1 and a high risk threshold P2;

[0038] When ≤P1, it is considered that the current myopia risk of the child is at low risk, and the current healthy eye - using habits are maintained;

[0039] When P1 < ≤P2, it is considered that the current myopia risk of the child is at medium risk, increase the outdoor activity time, and use a standardized reading light environment;

[0040] When ≥P1, it is considered that the current myopia risk of the child is at high risk, limit the use time of electronic screens, carry out myopia intervention, and conduct regular vision checks.

[0041] In a second aspect, the present invention provides a system for predicting and preventing myopia in children based on data mining, including a data acquisition module, a feature extraction module, a feature fusion module, a myopia probability prediction module, and a prevention and control strategy generation module; the data acquisition module is used to collect the dynamic modal data and static modal data of children, and pre - process the dynamic modal data and static modal data; the feature extraction module is used to extract the dynamic modal features and static modal features of children based on the pre - processed two - modal data; the feature fusion module is used to generate personalized myopia risk features by using a multi - modal fusion algorithm based on the dynamic modal features and static modal features of children; the myopia probability prediction module is used to construct a myopia risk prediction model and predict the myopia probability of children based on the personalized myopia risk features; the prevention and control strategy generation module is used to generate a personalized prevention and control strategy according to the myopia probability of children.

[0042] In a third aspect, the present invention provides a computer device, including a memory and a processor, where: when the computer program is executed by the processor, it implements any step of the method for predicting and preventing myopia in children based on data mining as described in the first aspect of the present invention.

[0043] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by a processor, any step of the method for predicting the prevention and control of myopia in children based on data mining as described in the first aspect of the present invention is implemented.

[0044] The beneficial effects of the present invention are as follows: By integrating dynamic and static modal features and using the attention mechanism to generate personalized myopia risk features, accurate risk assessment is achieved. Then, a deep neural network is used to construct a myopia risk prediction model, which converts these features into specific myopia probabilities, provides high-precision personalized predictions, and supports early warning and customized prevention and control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is a flowchart of the method for predicting the prevention and control of myopia in children based on data mining in Embodiment 1.

[0047] Figure 2 It is a schematic diagram of the system for predicting the prevention and control of myopia in children based on data mining in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0049] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0050] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0051] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a method for predicting the prevention and control of myopia in children based on data mining, including the following steps:

[0052] S1: Collect the dynamic modality data and static modality data of children, and preprocess the dynamic modality data and static modality data.

[0053] The dynamic modality data includes the blinking frequency, pupil diameter change, gaze time distribution and head pose change of children.

[0054] It should be noted that the collection of dynamic modality data mainly relies on intelligent wearable devices and screen monitoring software. Through sensors worn on children, such as eye trackers, head pose sensors, etc., the blinking frequency, pupil diameter change and head pose change can be monitored in real time. These devices can capture data points at a high frequency to ensure the accuracy and continuity of the data. At the same time, through the dedicated software installed on the electronic device, the gaze time distribution and screen usage time of children can be recorded, so as to comprehensively understand the eye use behavior pattern of children.

[0055] The static modality data includes the axial length of the eye, refractive power and reading distance of children.

[0056] It should be noted that the static modality data is obtained through regular medical examinations and questionnaires. The data of the axial length of the eye and refractive power are usually measured by professional ophthalmologists using precision instruments and recorded in the health files of children. The reading distance can be collected through questionnaires filled out by parents or teachers, or can be assisted by intelligent devices (such as smart watches or mobile phone applications) for recording to ensure the consistency and objectivity of the data. This method combines professional measurement with daily observation, providing a solid foundation for evaluating the visual health of children.

[0057] Remove the outliers in the dynamic modality data by the box plot method.

[0058] Furthermore, when processing the dynamic modality data, first apply the box plot method to identify and remove outliers. This method is based on statistical principles and uses the interquartile range (IQR) between the first quartile (Q1) and the third quartile (Q3) to define the data points within the normal range. Any data points outside this range are regarded as outliers and excluded, ensuring that the subsequent analyzed data set is more pure and accurate.

[0059] Smooth the dynamic modality data by using Gaussian filtering.

[0060] Furthermore, use Gaussian filtering technology to smooth the dynamic modality data. Gaussian filtering is a common image processing and data analysis technology. By applying a window with a Gaussian distribution weight to the data points, the new value of each data point becomes the weighted average result of its neighboring points. This method can effectively reduce noise interference while retaining the main features and trends of the original data.

[0061] Convert the static modal data through standardization to eliminate different dimensions.

[0062] Furthermore, apply standardization to eliminate the influence of different dimensions. This process transforms all features to the same scale, usually by subtracting the mean and dividing by the standard deviation. The standardized static modal data ensures the comparability of each feature in subsequent analysis, avoids biases caused by unit or magnitude differences, and thus improves the accuracy of data analysis.

[0063] Perform data dimensionality reduction on the static modal data through principal component analysis.

[0064] Furthermore, use principal component analysis (PCA) to perform dimensionality reduction on the static modal data. Principal component analysis is a statistical method used to identify the linear relationships between variables in a dataset and transform these variables into a new set of orthogonal variables - namely, principal components. By selecting the first few principal components that explain most of the variance, the data dimension can be significantly reduced while retaining important information, simplifying the complexity in the subsequent modeling process.

[0065] S2: Extract the dynamic modal features and static modal features of children based on the two preprocessed modal data.

[0066] Conduct time series analysis on the blink frequency data to extract the average blink frequency feature of children per unit time;

[0067] Furthermore, when processing the preprocessed blink frequency data, apply time series analysis methods. By analyzing the blink frequency data at a series of time points and calculating the average value per unit time, the average blink frequency feature of children during this period can be extracted. This method can accurately capture the regular changes in children's blink behavior and provide a stable basis for subsequent analysis.

[0068] For the pupil diameter change data, use Fourier transform to extract the frequency components and fluctuation amplitude features of the pupil diameter change data of children;

[0069] Furthermore, for the pupil diameter change data, use Fourier transform technology for analysis. This process converts the pupil diameter change in the time domain to a frequency domain representation to identify the frequency components and fluctuation amplitude features. In this way, the periodic and aperiodic patterns of the pupil diameter change over time can be revealed, providing important information about children's visual responses.

[0070] For the head pose change data, extract the stability feature of the head pose of children per unit time;

[0071] Further, perform stability analysis on the preprocessed head pose change data. By evaluating the change of the head pose within a unit time, determine the stability characteristics of the head position. This step helps to understand whether the head remains relatively stationary during the child's eye use and whether there are frequent or large-amplitude pose changes, which is crucial for understanding the child's eye use habits.

[0072] Perform time series analysis on the axial length data to extract the growth characteristics of the child's axial length;

[0073] Further, when processing the preprocessed axial length data, also use time series analysis. By analyzing the measurement results of the axial length at different time points, extract the growth trend characteristics of the axial length. This process aims to track the change of the axial length over time and provide a key basis for evaluating the development of myopia in children.

[0074] Statistically analyze the change amplitude and direction of the diopter to extract the diopter change characteristics of the child;

[0075] Further, for the diopter data, perform statistical analysis to determine the change amplitude and direction. By quantifying the degree of change of the diopter and its increasing or decreasing trend, extract the diopter change characteristics. This method can effectively reflect the dynamic changes of the child's visual state and help identify possible vision problems.

[0076] According to the preprocessed reading distance, use statistical analysis methods to extract the variability characteristics of the child's reading distance.

[0077] Further, based on the preprocessed reading distance data, apply statistical analysis methods. By calculating the distribution of the reading distance, including statistical quantities such as the minimum value, maximum value, and standard deviation, extract the variability characteristics of the reading distance. This step aims to evaluate the change of the distance between the child's eyes and the book or other reading materials during reading and provide a reference for optimizing the eye use environment.

[0078] It should be noted that the dynamic modal characteristics include: average blink frequency characteristics, frequency components and fluctuation amplitude characteristics of the pupil diameter, and stability characteristics of the head pose; the static modal characteristics include axial length growth characteristics, diopter change characteristics, and variability characteristics of the reading distance;

[0079] S3: Based on the child's dynamic modal characteristics and static modal characteristics, use a multi-modal fusion algorithm to generate personalized myopia risk characteristics.

[0080] Based on the child's dynamic modal characteristics and static modal characteristics, through the attention mechanism, calculate the dot product of each dynamic modal characteristic and static modal characteristic with the weight vector to assign weights to each dynamic modal characteristic and static modal characteristic;

[0081] Furthermore, when processing the dynamic modal features and static modal features of children, the attention mechanism is adopted to calculate the dot product of each feature and the weight vector, so as to assign corresponding weights to each feature. In this process, the dynamic modal features and static modal features are first input into the attention model. By calculating the dot product between the feature vector and the weight vector, the importance degree of each feature is determined. Subsequently, according to these dot product results, the dynamic modal features and static modal features respectively obtain corresponding weight values, so that in subsequent analysis, more critical features can receive higher attention.

[0082] The weight assignment for dynamic modal features is expressed as:

[0083] ;

[0084] where, is the weight coefficient of the th dynamic modal feature, is the value of the th dynamic modal feature, is the total number of dynamic modal features, is the transpose of the weight vector of dynamic modal features, is the index variable of the dynamic modal feature in the normalization process, is the th value of the dynamic modal feature in the normalization process;

[0085] It should be noted that when assigning weights to dynamic modal features, first calculate the dot product between each dynamic modal feature and the weight vector. Then, apply the softmax function to convert these dot products into weight coefficients in the form of probability distribution. Specifically, for the th dynamic modal feature, by calculating the dot product of its value and the weight vector, and after transforming the result through the exponential function, dividing it by the sum of the exponential transformation values of the dot products corresponding to all dynamic modal features, the weight coefficient of this feature is obtained. This process ensures that the sum of the weight coefficients of all dynamic modal features is 1, achieving the normalization process.

[0086] For example, when considering multiple dynamic modal features such as blink frequency and pupil diameter change, each feature will obtain a weight coefficient reflecting its relative importance according to the above formula. In this way, those features that can better reflect children's eye use habits and visual behavior patterns will be given higher weights, so as to occupy a more important position in subsequent analysis. This method not only ensures the comparability between different features, but also enhances the accuracy and rationality of personalized myopia risk assessment.

[0087] The weight assignment for static modal features is expressed as:

[0088] ;

[0089] Among them, is the weight coefficient of the th static modal feature, is the value of the th static modal feature, is the total number of static modal features, is the transpose of the weight vector of static modal features, is the index variable of the static modal feature during the normalization process, is the value of the th static modal feature during the normalization process;

[0090] It should be noted that when assigning weights to static modal features, first calculate the dot product between each static modal feature and the weight vector. Then, apply the softmax function to convert these dot products into weight coefficients in the form of a probability distribution . Specifically, for the th static modal feature, by calculating the dot product of its value and the weight vector, and after transforming the result through the exponential function, dividing it by the sum of the exponential transformation values of the dot products corresponding to all static modal features, the weight coefficient of this feature is obtained. This process ensures that the sum of the weight coefficients of all static modal features is 1, achieving the normalization process.

[0091] For example, when considering static modal features such as axial length and refractive power, each feature obtains a weight coefficient reflecting its relative importance according to the above formula. This means that those features that can better reflect the long-term vision change trend of children will be assigned higher weights. This method not only ensures the comparability between different static modal features but also enhances the accuracy and rationality of personalized myopia risk assessment. In this way, the importance of static modal features is quantified, thus better supporting subsequent analysis and prediction work.

[0092] Based on the weight coefficient of the dynamic modal feature and the weight coefficient of the static modal feature, a multi-modal fusion algorithm is used to perform feature interaction calculations, capture the non-linear relationship between each dynamic modal feature and the static modal feature, and generate personalized myopia risk features. The expression is:

[0093] ;

[0094] Among them, is the personalized myopia risk feature, is the value of the th dynamic modal feature The feature value obtained after non-linear transformation, is the value of the th static modal feature The eigenvalue obtained after non - linear transformation, is a positive constant that adjusts the differential influence between dynamic and static modal features, is the normalization factor, is the original value of the th modal feature, is the mean value of the th modal feature,

[0095] It should be noted that the "modal feature" mentioned in the original value of the th modal feature includes both dynamic and static modal features.

[0096] It should be noted that in the process of generating personalized myopia risk features, a multi - modal fusion algorithm is adopted. The weighted eigenvalue is processed by non - linear transformation to capture the non - linear relationship between dynamic and static modal features. Next, all weighted and transformed eigenvalues are integrated, and normalization is performed to ensure the balance of the importance of different features. Finally, an exponential decay function is applied to adjust the differential influence between features to generate the final personalized myopia risk feature R. This process integrates various feature information, improving the accuracy and personalization level of risk assessment.

[0097] S4: Build a myopia risk prediction model to predict the myopia probability of children based on personalized myopia risk features.

[0098] Use DNN (Deep Neural Networks) as the framework of the myopia risk prediction model;

[0099] It should be noted that using DNN as the framework of the myopia risk prediction model is because DNN can automatically learn and extract the non - linear relationship between complex features, thus providing a more accurate and personalized myopia risk prediction.

[0100] Based on the framework of the myopia risk prediction model, build an input layer, a first hidden layer, a second hidden layer, and an output layer;

[0101] The first hidden layer receives the personalized myopia risk features from the input layer and performs a linear transformation on the personalized myopia risk features according to the weight matrix and bias vector and then activates the result through the ReLU activation function to generate a preliminary feature representation;

[0102] Furthermore, the first hidden layer receives the personalized myopia risk feature R from the input layer. The specific process is as follows: The personalized myopia risk feature R is passed to the first hidden layer, where each feature value is multiplied by a weight matrix and added with a bias vector. This linear transformation operation aims to capture the linear relationship between the input features and the output. Subsequently, the transformed result is processed by the ReLU activation function, which sets all negative values to zero while positive values remain unchanged. This step not only enhances the model's ability to learn non-linear relationships but also ensures that only meaningful feature representations are retained and enhanced. For example, if the personalized myopia risk feature of a certain child indicates that their axial length is growing rapidly, after being activated by ReLU, the importance of this feature will be more prominent.

[0103] The second hidden layer receives the preliminary feature representation and repeats the activation process of the linear transformation and the ReLU activation function to generate a deep feature representation;

[0104] Furthermore, the second hidden layer receives the preliminary feature representations generated from the first hidden layer, which have undergone linear transformation and ReLU activation processing. In the second hidden layer, the same linear transformation is applied to the preliminary feature representations again, that is, the weight matrix and the bias vector are applied again for transformation. Then, the result of the linear transformation is processed by the ReLU activation function again. This repeated process further abstracts the features and generates deeper feature representations. For example, assuming that the first hidden layer identifies that an abnormally high blink frequency may be one of the risk factors for myopia, then the second hidden layer may further analyze the association between this high-frequency blink and other eye-using behaviors (such as long-time screen viewing), thereby more accurately reflecting the potential myopia risk.

[0105] The output layer receives the deep feature representation and is activated by the Sigmoid activation function;

[0106] Furthermore, the output layer receives the deep feature representation generated by the second hidden layer and performs the final processing through the Sigmoid activation function. The Sigmoid function maps any real value to between 0 and 1, enabling the output to be directly interpreted as a probability value. In this process, the information in the deep feature representation is integrated to produce a value between 0 and 1, representing the myopia probability of the child. For example, if the deep feature representation of a child shows that their eye-using habits and physiological characteristics all point to a high myopia risk, the Sigmoid activation function will output a value close to 1, indicating a high myopia probability.

[0107] Combine the input layer, the first hidden layer, the second hidden layer, the output layer, and the framework of the myopia risk prediction model to construct a myopia risk prediction model;

[0108] Furthermore, the construction of the myopia risk prediction model combines an input layer, a first hidden layer, a second hidden layer, and an output layer. Each layer is connected in sequence to form a complete network structure for processing personalized myopia risk features and predicting the myopia probability of children. The input layer receives the original personalized myopia risk features, the first hidden layer and the second hidden layer are responsible for gradually refining and abstracting these features, and the output layer generates the final prediction result.

[0109] Based on the personalized myopia risk features, the myopia probability of children is predicted, and the expression is:

[0110] ;

[0111] where is the myopia probability of the child, is the weight matrix of the first hidden layer, is the weight matrix of the second hidden layer, the weight matrix of the output layer, is the bias vector of the first hidden layer, is the bias vector of the second hidden layer, is the bias vector of the output layer.

[0112] It should be noted that after the model construction is completed, the personalized myopia risk features are input into the myopia risk prediction model. Inside the model, the features first undergo a linear transformation and ReLU activation process in the first hidden layer, then a deeper abstraction process in the second hidden layer, and finally are transformed into a specific myopia probability P through the Sigmoid activation function in the output layer. The whole process integrates multi-level information processing, ensuring the accuracy and reliability of the prediction results. For example, for a child with specific eye-using habits and physiological characteristics, the model can output a specific myopia probability value based on their personalized myopia risk features, helping parents or doctors take corresponding preventive measures.

[0113] S5: Generate a personalized prevention and control strategy according to the myopia probability of the child.

[0114] Based on the statistical analysis of the low-risk and high-risk probability distributions in the historical myopia population data, a low-risk threshold P1 and a high-risk threshold P2 are defined;

[0115] When ≤P1, it is considered that the current child's myopia risk is at low risk, and the current healthy eye-using habits are maintained;

[0116] It should be noted that when the myopia probability P is less than or equal to the low-risk threshold P1, for example, the calculated myopia probability of a certain child is 0.2, which is much lower than the set low-risk threshold. This means that the current child's myopia risk is at a low level. In this case, it is recommended to continue to maintain the existing healthy eye-using habits, such as keeping an appropriate reading distance, taking regular breaks for the eyes, and maintaining good lighting conditions. Parents do not need to take additional measures, but only need to continuously pay attention to the child's vision condition.

[0117] When P1 < ≤ P2, it is considered that the current child's myopia risk is at a medium risk. Increase the outdoor activity time and use a standard reading lighting environment;

[0118] It should be noted that when the myopia probability P is greater than the low-risk threshold P1 but less than or equal to the medium-risk threshold P2, for example, the myopia probability of a certain child is 0.6, which is between low risk and high risk. This indicates that the current child's myopia risk is at a medium level. In this case, it is recommended to increase the outdoor activity time, ensure at least a certain amount of natural light exposure every day, and at the same time ensure the use of a standard reading lighting environment to reduce the burden on the eyes. Through these preventive measures, the possibility of myopia development can be effectively reduced.

[0119] When ≥ P1, it is considered that the current child's myopia risk is at a high risk. Limit the use time of electronic screens, carry out myopia intervention, and conduct regular vision checks.

[0120] It should be noted that when the myopia probability P is greater than or equal to the high-risk threshold P1, for example, the myopia probability of a certain child is as high as 0.85, exceeding the set high-risk threshold. This means that the current child's myopia risk is relatively high. In this case, immediate action should be taken to limit the use time of electronic screens, avoid long-term close-range eye use, and consider professional myopia intervention measures, such as wearing corrective glasses or receiving vision training. In addition, regular vision checks should also be arranged to monitor vision changes in a timely manner and adjust the treatment plan.

[0121] This embodiment also provides a children's myopia prevention and control prediction system based on data mining, including: a data collection module, a feature extraction module, a feature fusion module, a myopia probability prediction module, and a prevention and control strategy generation module; the data collection module is used to collect the dynamic modal data and static modal data of children, and preprocess the dynamic modal data and static modal data; the feature extraction module is used to extract the dynamic modal features and static modal features of children based on the preprocessed two types of modal data; the feature fusion module is used to generate personalized myopia risk features based on the dynamic modal features and static modal features of children by using a multimodal fusion algorithm; the myopia probability prediction module is used to construct a myopia risk prediction model and predict the myopia probability of children based on the personalized myopia risk features; the prevention and control strategy generation module is used to generate personalized prevention and control strategies according to the myopia probability of children.

[0122] This embodiment also provides a computer device, which is applicable to the situation of the children's myopia prevention and control prediction method based on data mining, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the children's myopia prevention and control prediction method based on data mining proposed in the above embodiment.

[0123] This computer device can be a terminal, and this computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this 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 this 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, a carrier network, NFC (Near Field Communication) or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device can be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the computer device housing, or an external keyboard, a touchpad or a mouse, etc.

[0124] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for predicting the prevention and control of myopia in children based on data mining as proposed in the above embodiment; 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, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0125] In summary, the present invention: by fusing dynamic and static modal features and using the attention mechanism to generate personalized myopia risk features, realizes accurate risk assessment. Then, a deep neural network (DNN) is used to construct a myopia risk prediction model, which converts these features into specific myopia probabilities, provides high-precision personalized prediction, and supports early warning and customized prevention and control strategies.

[0126] Example 2, referring to Table 1, is the second embodiment of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the method for predicting the prevention and control of myopia in children based on data mining is given.

[0127] To verify the effectiveness of the method for predicting the prevention and control of myopia in children based on data mining, an experimental study was designed. This experiment covered 120 children aged between 6 and 12 years old. The myopia assessment was carried out on each child using the method of the present invention and the traditional method respectively: the experimental group (using the method of the present invention) and the control group (using the traditional assessment method of the existing technology). All children wore intelligent wearable devices, such as eye trackers and head pose sensors, and used electronic devices installed with screen monitoring software. At the same time, medical examinations were regularly carried out to collect static modal data, including axial length, refractive power, etc., and the reading distance of children was determined through questionnaires.

[0128] First, in this experiment, the dynamic modal data (such as blink frequency, pupil diameter change) and static modal data (such as axial length, refractive power) of children were collected through intelligent wearable devices and screen monitoring software. The box plot method was used to remove outliers, and Gaussian filtering was used to smooth the dynamic data; the static data was standardized and reduced in dimension by principal component analysis.

[0129] Secondly, based on the preprocessed data, dynamic modal features and static modal features were extracted. For example, time series analysis was performed on the blink frequency to extract the average blink frequency; Fourier transform was applied to the pupil diameter change data to extract the frequency components and fluctuation amplitudes; the stability of the head pose was evaluated; the growth trend of the eye axis length was analyzed; the changes in refractive power were statistically analyzed; and the variability of the reading distance was calculated.

[0130] Then, personalized myopia risk features were generated through a multimodal fusion algorithm. The attention mechanism was used to assign weights to each feature and perform a non-linear transformation to capture the non-linear relationship between dynamic and static features, and finally personalized myopia risk feature R was generated.

[0131] Finally, a deep neural network (DNN) model was constructed to predict the myopia probability P. This model received the personalized myopia risk feature R, underwent linear transformation and ReLU activation processing through two hidden layers, and output the myopia probability through the Sigmoid function in the output layer. At the same time, the low-risk threshold P1 was defined as 0.30, and the high-risk threshold P2 was defined as 0.70.

[0132] The control group adopted the traditional myopia risk assessment method, which was limited to simple feature superposition and did not apply advanced data mining techniques and multimodal fusion algorithms for personalized risk assessment.

[0133] Specifically, it is shown in Table 1 below:

[0134] Table 1 Comparison Table of Experimental Data for Children's Myopia Risk Assessment

[0135] Test subjects Blinking frequency (times / minute) Change in pupil diameter (millimeters) Axial length growth (millimeters / year) Change in refractive power (Diopters) Predicted myopia probability P of the experimental group Predicted myopia probability P of the control group Actual myopia status (1 = myopia, 0 = non-myopia) Child A 26.5 3.80 0.27 -1.00 0.78 0.25 1 Child B 15.2 3.20 0.18 -0.50 0.22 0.75 0 Child C 24.5 3.70 0.26 -0.92 0.83 0.78 1

[0136] By analyzing the experimental data, it can be seen that the prediction results of the experimental group of the present invention are significantly better than those of the control group. In Child A, the experimental group predicted the myopia probability to be 0.78 (high risk, P > P2), identifying it as high-risk myopia, which was consistent with the actual situation (myopia, actual value is 1), while the control group predicted it to be 0.25 (low risk, P < P1), misassessing it as low risk and underestimating the risk. In Child B, the experimental group predicted the myopia probability to be 0.22 (low risk, P < P1), judging it as low risk, which was consistent with the actual situation (not myopic, actual value is 0), while the control group predicted it to be 0.75 (high risk, P > P2), overestimating the risk incorrectly and falsely reporting it as high risk. In Child C, both the experimental group and the control group predicted it to be high risk (P > P2), but the predicted value of the experimental group was 0.83, which was closer to the actual situation (myopia, actual value is 1), better than the predicted value of 0.78 of the control group. The present invention accurately captures the myopia risk features of children through multimodal data fusion and deep learning models, showing higher prediction accuracy and reliability.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the prevention and control of children's myopia based on data mining, characterized in that: Including: Collecting the dynamic modal data and static modal data of children, and preprocessing the two types of modal data. The dynamic modal data includes the blinking frequency, pupil diameter change, gaze time distribution and head pose change of children, and the static modal data includes the eye axis length, refractive power and reading distance of children; Extracting the dynamic modal features and static modal features of children based on the preprocessed two types of modal data, calculating the dot product of each dynamic modal feature and static modal feature with the weight vector, assigning weights to each dynamic modal feature and static modal feature, and using a multimodal fusion algorithm to perform feature interaction calculation to generate personalized myopia risk features; Constructing a myopia risk prediction model based on a deep neural network, predicting the myopia probability of children based on the personalized myopia risk features, and generating personalized prevention and control strategies; Assigning weights to the dynamic modal features, and the expression is: ； Among them, is the weight coefficient of the th dynamic modal feature, is the value of the th dynamic modal feature, is the total number of dynamic modal features, is the transpose of the weight vector of the dynamic modal feature, is the index variable of the dynamic modal feature in the normalization process, is the value of the th dynamic modal feature in the normalization process; Assigning weights to the static modal features, and the expression is: ； Among them, is the weight coefficient of the th static modal feature, is the value of the th static modal feature, is the total number of static modal features, is the transpose of the weight vector of the static modal feature, is the index variable of the static modal feature in the normalization process, is the value of the th static modal feature in the normalization process; Weight coefficients based on dynamic modal features and weight coefficients of static modal features , use a multimodal fusion algorithm to perform feature interaction calculations, capture the non-linear relationship between each dynamic modal feature and static modal feature, and generate personalized myopia risk features; The expression of the personalized myopia risk features is: ； Among them, is the personalized myopia risk feature, is the value of the th dynamic modal feature after non - linear transformation, is the value of the th static modal feature after non - linear transformation, is a positive constant that adjusts the differential influence of dynamic and static modal features, is the regularization factor, is the original value of the th modal feature, is the mean value of the th modal feature, is the index variable of all dynamic and static modal features.

2. The method for predicting and preventing myopia in children based on data mining according to claim 1, wherein: The preprocessing of the two types of modal data is specifically as follows: Removing outliers in the dynamic modal data through the box plot method; Smoothing the dynamic modal data using Gaussian filtering; Converting the static modal data through standardization processing to eliminate different dimensions; Performing data dimensionality reduction processing on the static modal data through the principal component analysis method.

3. The method for predicting and preventing myopia in children based on data mining according to claim 2, wherein: The extraction of the dynamic modal features and static modal features of children based on the preprocessed two types of modal data is specifically as follows: Performing time series analysis on the blinking frequency data to extract the average blinking frequency feature of children within a unit time; For the pupil diameter change data, using Fourier transform to extract the frequency components and fluctuation amplitude features of the pupil diameter change data of children; For the head pose change data, extracting the stability feature of the head pose of children within a unit time; Performing time series analysis on the eye axis length data to extract the growth feature of the eye axis length of children; Statistically analyzing the change amplitude and direction of the refractive power to extract the refractive power change feature of children; According to the preprocessed reading distance, using statistical analysis methods to extract the reading distance variability feature of children.

4. The method for predicting and preventing myopia in children based on data mining according to claim 3, wherein: The construction of a myopia risk prediction model based on a deep neural network, predicting the myopia probability of children based on the personalized myopia risk features, is specifically as follows: Using DNN as the framework of the myopia risk prediction model; Based on the framework of the myopia risk prediction model, constructing an input layer, a first hidden layer, a second hidden layer and an output layer; The first hidden layer receives the personalized myopia risk features of the input layer , performs a linear transformation on the personalized myopia risk features according to the weight matrix and the bias vector , and performs an activation process through the ReLU activation function to generate a preliminary feature representation; The second hidden layer receives the preliminary feature representation, undergoes repeated linear transformation and activation processing by the ReLU activation function to generate a deep feature representation; The output layer receives the deep feature representation and undergoes activation processing through the Sigmoid activation function; Combining the input layer, the first hidden layer, the second hidden layer, the output layer and the framework of the myopia risk prediction model to construct a myopia risk prediction model; Predicting the myopia probability of children based on the personalized myopia risk features, and the expression is: ； Among them, is the myopia probability of children, is the weight matrix of the first hidden layer, is the weight matrix of the second hidden layer, is the weight matrix of the output layer, is the bias vector of the first hidden layer, is the bias vector of the second hidden layer, is the bias vector of the output layer.

5. The method for predicting the prevention and control of children's myopia based on data mining according to claim 4, wherein: The generation of personalized prevention and control strategies is specifically as follows: Based on the statistical analysis of the probability distributions of low risk and high risk in the historical myopia population data, define the low-risk threshold P1 and the high-risk threshold P2; When ≤ P1, it is considered that the current myopia risk of the child is at a low risk, and the current healthy eye-using habits are maintained; When P1 < ≤ P2, it is considered that the current myopia risk of the child is at a medium risk. Increase the outdoor activity time and use a standardized reading light environment; When ≥P1, it is considered that the current myopia risk of children is at a high level. Restrict the usage time of electronic screens, conduct myopia intervention, and conduct regular vision checks.

6. A child myopia prevention and control prediction system based on data mining, which is used to implement the child myopia prevention and control prediction method based on data mining according to any one of claims 1 to 5, and is characterized in that: It includes a data acquisition module, a feature extraction module, a feature fusion module, a myopia probability prediction module, and a prevention and control strategy generation module; The data acquisition module is used to collect the dynamic modality data and static modality data of children, and preprocess the dynamic modality data and static modality data; The feature extraction module is used to extract the dynamic modality features and static modality features of children based on the two preprocessed modality data; The feature fusion module is used to generate personalized myopia risk features based on the dynamic modality features and static modality features of children by using a multimodal fusion algorithm; The myopia probability prediction module is used to construct a myopia risk prediction model and predict the myopia probability of children based on the personalized myopia risk features; The prevention and control strategy generation module is used to generate personalized prevention and control strategies according to the myopia probability of children.

7. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the data mining-based children's myopia prevention and control prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the data mining-based children's myopia prevention and control prediction method according to any one of claims 1 to 5.

Images