Digital children growth and development predictive guidance method

Abstract

The invention discloses a digital children growth and development predictive guidance method, particularly relates to the field of children growth and development evaluation, solves the problems of individual growth trajectory matching and specificity recognition, and comprises the following steps: segmenting a normal growth curve into local reference windows and constructing a long-term standard curve; generating a target growth curve in combination with a dynamic time warping method, and capturing the growth characteristic difference between short-term fluctuation and global trend of a target individual; by extracting a relative curvature unbalance factor and a cross-section extension entropy difference value, dynamic deviation of a target growth curve in local and global ranges is quantitatively described; performing time domain division dynamic evaluation by using a short-term mutation deviation index and a long-term gradual deviation index, generating an explanation credibility coefficient of a normal target growth trajectory, dividing individual growth characteristic grades, and determining an explanation capability range of a normal growth curve to the target trajectory, thereby being beneficial to distinguishing the matching degree of a target and a normal growth mode, and improving the target growth accuracy. And the growth specificity is recognized.

Description

Technical Field

[0001] The present invention relates to the field of children's growth and development evaluation, and more specifically, to a digital children's growth and development predictive guidance method. Background Art

[0002] Children's growth and development is a complex and highly personalized process, which is jointly influenced by multiple factors such as genetics, environment, diet and lifestyle. Current digital growth monitoring technology can construct individual growth curves by regularly collecting key indicators such as children's height, weight, bone age, etc., and compare them with standard normal growth curves to assess growth status. However, in actual scenarios, individual growth behaviors often show significant differences in short-term fluctuations and long-term trends. This complexity makes the method of relying solely on normal growth curves as an evaluation standard, which may have the hidden danger of insufficient applicability when facing real growing individuals. For example, the normal growth curve is generated based on group statistical data, and its ability to explain local details and specific behaviors of individual growth may not fully meet actual needs, especially for children with atypical growth rhythms.

[0003] When existing technologies rely too much on normal growth curves for evaluation, they ignore the uniqueness and differences in the growth characteristics of the target individuals, especially the correlation analysis between local fluctuations and long-term trends is relatively weak. This technical limitation may cause the evaluation results to deviate from the actual situation of the individual, fail to accurately identify the dynamic changes in the individual growth trajectory, and lack guidance for individualized intervention plans. In addition, when dealing with the differences between the target growth curve and the normal growth curve, the existing methods lack an evaluation mechanism for the interpretation ability of the normal growth curve, making it difficult to determine whether the individual growth data can be effectively covered by the normal standard. This deficiency further limits the accurate capture of abnormal or specific growth behaviors, and there is an urgent need for a differentiated evaluation method that can combine individual growth characteristics with the standard normal growth curve to provide more scientific and personalized growth and development evaluation and guidance support.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a digital child growth and development predictive guidance method, which divides the normal growth curve into local reference windows and constructs a long-term standard curve, and generates a target growth curve in combination with a dynamic time warping method, so as to accurately capture the growth characteristic differences between the short-term fluctuations and the global trend of the target individual; through the relative curvature imbalance factor and the cross-segment extension entropy difference extraction, the dynamic deviation characteristics of the target growth curve in the local and global ranges are quantitatively described; by using the time-domain dynamic evaluation of the short-term mutation deviation index and the long-term progressive deviation index, combined with a multi-level feature interaction network and an adaptive weight adjustment mechanism, a normal interpretation credibility coefficient of the target growth trajectory is generated, and the individual growth characteristic levels are divided accordingly, and the range of the normal growth curve's explanatory ability for the target trajectory is clarified, which helps to distinguish the degree of matching between the target trajectory and the normal growth pattern, identify local or global growth specificity, and provide a scientific basis for personalized developmental intervention to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A digital child growth and development predictive guidance method, comprising the steps of:

[0008] S1: The normal growth curve is divided into multiple local reference windows to capture short-term fluctuation characteristics, and a long-term standard curve is constructed on a global scale. The local observation data of the target individual is mapped to the corresponding reference window through the dynamic time warping method and then secondary fused at the global level to generate the target growth curve;

[0009] S2: Based on the difference between the target growth curve and the normal growth curve, the relative curvature imbalance factor is extracted using the local change of the curve to measure the local deviation, and the cross-segment extension entropy difference is calculated based on the cross-window characteristics to measure the overall extension and contraction characteristics;

[0010] S3: Generate a short-term mutation deviation index by aggregating the deviation values ​​of the target growth curve within a short-term observation window, and obtain a long-term gradual deviation index by combining the complete historical data and accumulating them in segments, so as to evaluate the dynamic difference characteristics of the target growth curve in different time domains;

[0011] S4: The short-term mutation deviation index and the long-term asymptotic deviation index are input into the machine learning model. By integrating the features of the relative curvature imbalance factor and the cross-segment extension entropy difference and adjusting the weights, the credibility coefficient of the normal explanation of the target is calculated and the growth-specific assessment results of the target individual are obtained.

[0012] In a preferred embodiment, step S1 includes the following contents:

[0013] By calculating the local change rate of the normal growth curve through the second-order derivative, the change trend of the curve is identified and the characteristic points are extracted, where the extreme point of the absolute value of the change rate is used to delineate the boundary of the segmentation window; according to the position of the characteristic point, the normal growth curve is divided into several local reference windows, and the curve segment in each window independently describes the growth characteristics in the corresponding time period to form a set of short-term characteristic curves; the characteristic curves of all local windows are spliced ​​according to the time series, and optimized at the segment boundary through the curve smoothness constraint to construct a global long-term standard curve to express the complete normal growth law;

[0014] Subsequently, the growth observation data of the target individual are taken as discrete time series and aligned one by one with the local reference window of the normal growth curve using the dynamic time warping method; after completing the local mapping, all local fitting results are recombined according to the global time scale of the normal growth curve to construct the preliminary growth curve of the target individual;

[0015] Finally, the target growth curve and the global long-term standard curve are fused secondary in the overall range; the fusion adopts the method of local feature enhancement, which retains the specific characteristics of the target growth curve in local fitting while ensuring that its trend at the global level is consistent with the normal growth curve, and finally generates a target growth curve that can both describe the growth characteristics of the target individual and compare with the normal standard.

[0016] In a preferred embodiment, the target growth curve is secondarily fused with the global long-term standard curve within the overall range to generate a target growth curve that can describe the growth characteristics of the target individual and compare with the normal standard; the specific process is as follows:

[0017] a. After the target individual growth observation data is aligned with the local reference window of the normal growth curve, the key characteristic values ​​of the target growth curve in each local window are extracted;

[0018] b. The global standard trend of the normal growth curve is generated by splicing all local window curves, while ensuring the continuity and smoothness of the curve at the splicing points;

[0019] c. The following mathematical model is used for fusion: ;in, is the final target growth curve; is the fitting result of the target growth curve in the local window; is the normal global standard curve; It is the strengthening factor of dynamic calculation.

[0020] In a preferred embodiment, step S2 includes the following contents:

[0021] a. Extract relative curvature imbalance factor: define the target growth curve as The normal growth curve is ; Both in time The curvatures are: , ; In each local reference window The cumulative value of the curvature deviation of the target growth curve relative to the normal growth curve is calculated, which is called the relative curvature imbalance factor. : ; The relative curvature imbalance factor is used to quantify the degree of deviation between the fluctuation characteristics of the target growth curve in the local reference window and the normal growth curve.

[0022] In a preferred embodiment, step S2 further includes the following contents:

[0023] b. Calculate the difference in cross-segment extension entropy: For the target growth curve and the normal growth curve, calculate the total extension in each reference window: , ; Based on the window extension, calculate the global extension distribution probability of the target growth curve and the extended distribution probability of the normal growth curve : , ; Calculate the cross-segment extension entropy difference , the formula is: The cross-segment extension entropy difference was used to quantitatively evaluate the global difference between the overall extension characteristics of the target growth curve and the normal growth curve.

[0024] In a preferred embodiment, step S3 includes the following contents:

[0025] Select the recent observation window of the target growth curve and define the time range as The window length is determined by the target data acquisition frequency; within the selected window, the relative curvature imbalance factor and the cross-segment extension entropy difference are accumulated to calculate the overall mutation deviation within the window. : ;in, and To dynamically adjust the parameters.

[0026] In a preferred embodiment, step S3 further includes the following contents:

[0027] The complete historical time range of the target growth curve Divide into several equally spaced time periods ; The relative curvature imbalance factor and the cross-segment extension entropy difference in each time period are accumulated and summed globally to calculate the long-term asymptotic deviation index : .

[0028] In a preferred embodiment, step S4 includes the following contents:

[0029] Firstly, the short-term mutation deviation index and the long-term asymptotic deviation index are fused with the relative curvature imbalance factor and the cross-segment extension entropy difference to construct a multidimensional feature vector. Then a multi-level feature interaction network structure is adopted, in which each layer captures the complex correlation between different features through nonlinear transformation, ensuring that the relative curvature imbalance factor and the cross-segment extension entropy difference can be complementarily fused in different time windows.

[0030] In a preferred embodiment, step S4 further includes the following contents:

[0031] Then, the adaptive weight adjustment mechanism is used to dynamically allocate the contribution of the short-term mutation deviation index and the long-term gradual deviation index in the overall model according to the importance of the features, so as to optimize the sensitivity of the model to different deviation features.

[0032] In a preferred embodiment, step S4 further includes the following contents:

[0033] Subsequently, the weighted feature vector is input into a pre-trained deep neural network model, which outputs the credibility coefficient of the normal to target growth curve by learning the nonlinear relationship between the features;

[0034] Finally, according to the numerical range of the interpretation credibility coefficient, the growth-specific assessment results of the target individual are divided into different levels, including high credibility assessment, medium credibility assessment and low credibility assessment.

[0035] The technical effects and advantages of the digital child growth and development predictive guidance method of the present invention are as follows:

[0036] The present invention divides the normal growth curve into multiple local reference windows and constructs a long-term standard curve, combines the dynamic time warping method to generate the target growth curve, accurately captures the growth characteristic differences of the target individual in short-term fluctuations and global trends; through the extraction of the relative curvature imbalance factor and the cross-segment extension entropy difference, quantitatively describes the dynamic deviation characteristics of the target growth curve in the local and global range, and solves the problem of difficulty in identifying individual growth specificity in the prior art; further utilizes the time-domain dynamic evaluation of the short-term mutation deviation index and the long-term progressive deviation index, combines the feature fusion and adaptive weight adjustment mechanism of the multi-level feature interaction network structure, generates the normal interpretation credibility coefficient of the target growth trajectory, accurately divides the individual growth characteristic level, clarifies the scope of the normal growth curve's interpretation ability for the target growth trajectory, and helps to accurately identify the growth specificity of the target individual, thereby providing a scientific basis for the formulation of targeted personalized development intervention programs. This program not only improves the accuracy and scientificity of growth and development evaluation, but also provides reliable data support for personalized growth and development intervention, and effectively enhances the applicability and practicality of the evaluation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flowchart of a digital child growth and development predictive guidance method. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Embodiment 1: Figure 1 The present invention provides a digital child growth and development predictive guidance method, comprising:

[0040] S1: The normal growth curve is divided into multiple local reference windows to capture short-term fluctuation characteristics, and a long-term standard curve is constructed on a global scale. The local observation data of the target individual is mapped to the corresponding reference window through the dynamic time warping method and then secondary fused at the global level to generate the target growth curve;

[0041] S2: Based on the difference between the target growth curve and the normal growth curve, the relative curvature imbalance factor is extracted using the local change of the curve to measure the local deviation, and the cross-segment extension entropy difference is calculated based on the cross-window characteristics to measure the overall extension and contraction characteristics;

[0042] S3: Generate a short-term mutation deviation index by aggregating the deviation values ​​of the target growth curve within a short-term observation window, and obtain a long-term gradual deviation index by combining the complete historical data and accumulating them in segments, so as to evaluate the dynamic difference characteristics of the target growth curve in different time domains;

[0043] S4: The short-term mutation deviation index and the long-term asymptotic deviation index are input into the machine learning model. By integrating the features of the relative curvature imbalance factor and the cross-segment extension entropy difference and adjusting the weights, the credibility coefficient of the normal explanation of the target is calculated and the growth-specific assessment results of the target individual are obtained.

[0044] In the evaluation of children's growth and development, the difference between the growth status of the target individual and the normal growth curve is an important basis for evaluation specificity. However, since individual growth characteristics have both local short-term fluctuations and long-term trends, it is difficult to fully characterize the growth characteristics of the target individual based solely on overall or local analysis. To this end, it is necessary to construct a growth curve for the target individual and accurately compare it with the normal growth curve at the local and global levels, so as to capture the subtle features of the target growth curve and ensure global consistency with the normal standard, laying a data foundation for subsequent feature extraction and credibility assessment.

[0045] Step S1 includes the following contents:

[0046] First, the normal growth curve is defined as a function that changes with time, which is used to describe the growth trend characteristics of normal individuals within a specific time range. The local change rate of the normal growth curve is calculated by the second-order derivative, the change trend of the curve is identified and the characteristic points are extracted, where the extreme points of the absolute value of the change rate are used to delineate the boundaries of the segmentation window. According to the position of the characteristic points, the normal growth curve is divided into several local reference windows. The curve segments in each window independently describe the growth characteristics in the corresponding time period, forming a set of short-term characteristic curves. To ensure the consistency of the overall trend, the characteristic curves of all local windows are spliced ​​in time series, and optimized at the segment boundaries through curve smoothness constraints to construct a global long-term standard curve to express the complete normal growth law.

[0047] Subsequently, the growth observation data of the target individual is taken as a discrete time series and aligned with the local reference window of the normal growth curve one by one using the dynamic time warping method. Dynamic time warping minimizes the difference in values ​​between the target data and the reference window by nonlinearly adjusting the time series of the target data and the reference window, thereby finding the best matching path of the target individual observation point within the normal reference window. In this way, each target observation data point is mapped to the corresponding local reference window to obtain the local fitting result of the target data. After completing the local mapping, all local fitting results are recombined according to the global time scale of the normal growth curve to construct the preliminary growth curve of the target individual.

[0048] Finally, in order to further ensure the integrity of the target individual growth curve and the consistency of the normal growth law, the target growth curve and the global long-term standard curve are fused twice within the overall range. The fusion adopts the method of local feature enhancement, which retains the specific characteristics of the target growth curve in the local fitting while ensuring that its trend at the global level is consistent with the normal growth curve, and finally generates a target growth curve that can describe the growth characteristics of the target individual and compare with the normal standard. The specific process is as follows:

[0049] a. After the target individual growth observation data is aligned with the local reference window of the normal growth curve, the key characteristic values ​​of the target growth curve in each local window are extracted, including the curvature change, fluctuation amplitude and trend direction within the window. These local characteristic data are stored in the form of an array, for example, the average growth rate within the window, the position and value of the curve convexity and concavity change point.

[0050] b. The global standard trend of the normal growth curve is generated by splicing all local window curves, while ensuring the continuity and smoothness of the curve at the splicing point. In order to further enhance the trend characteristics of the global standard curve, the splicing segments of the curves between windows are optimized by applying a smoothing function to ensure that the global trend conforms to the integrity of the normal growth law.

[0051] c. The following mathematical model is used for fusion:

[0052] ;

[0053] in, is the final target growth curve; is the fitting result of the target growth curve in the local window; is the normal global standard curve; is a dynamically calculated reinforcement factor. The value of the reinforcement factor is determined by the fit between the local characteristics of the target growth curve and the global trend. The specific calculation formula is:

[0054] ;

[0055] in and They are the derivatives of the target local curve and the global standard curve, respectively, and are used to measure their changing trends.

[0056] The target growth curve generated by the above model has both local detail characteristics and global trend consistency on the time axis.

[0057] The normal growth curve is a standard curve generated through statistical analysis and model fitting based on the growth and development data of a large number of healthy individuals. It usually takes time as the horizontal axis and the key growth indicators of individuals (such as height, weight, etc.) as the vertical axis, describing the upper and lower limits and average trends of normal growth in a specific age group. The upper and lower limits are defined as the physiological range differences within a certain interval, forming a reference area for evaluating growth patterns. The growth observation data of the target individual refers to the key growth-related indicators obtained through continuous monitoring, including height measurement values, weight measurement values, bone age assessment results, and the time points of observation data collection. These data are usually recorded in the form of time series to characterize the growth status and change trends of the target individual at different time nodes.

[0058] In step S1, the target growth curve has been generated through multi-scale alignment and dynamic time warping methods, and it is ensured that the target growth curve captures the local characteristics of the target individual and maintains consistency with the normal growth curve in terms of global trends. Step S2 aims to extract key characteristics describing the relationship between the target growth curve and the normal growth curve based on the difference between the target growth curve and the normal growth curve, including the relative curvature imbalance factor of local deviation and the cross-segment extension entropy difference of the overall extension characteristics.

[0059] Step S2 includes the following contents:

[0060] a. Extract relative curvature imbalance factor:

[0061] The core of the relative curvature imbalance factor is to measure the degree of deviation of the curvature of the target growth curve from the normal growth curve in the local reference window. Curvature is a quantitative description of the degree of curvature of the growth curve in space. Its first-order derivative represents the growth rate, and the second-order derivative represents the speed of change of the rate. By calculating the curvature values ​​of the target growth curve and the normal growth curve in each reference window respectively, and integrating the difference between the two on the time axis, the cumulative curvature difference of the two curves within the time range can be obtained. This cumulative value is the relative curvature imbalance factor. This factor can reflect the morphological difference of the target growth curve relative to the normal growth curve in the local range, and is particularly sensitive to capturing the local characteristics of abnormal mutations in the target growth curve. The specific calculation logic is as follows:

[0062] Define the target growth curve as The normal growth curve is . Both in time The curvatures are:

[0063] , ;

[0064] In each local reference window The cumulative value of the curvature deviation of the target growth curve relative to the normal growth curve is calculated, which is called the relative curvature imbalance factor. :

[0065] .

[0066] The relative curvature imbalance factor is used to quantify the degree of deviation between the fluctuation characteristics of the target growth curve in the local reference window and the normal growth curve, which can help identify whether the target individual has a local change pattern that is different from the normal growth trajectory, such as abnormal growth jumps or stagnation. The introduction of this factor provides a reliable quantitative basis for accurately capturing local specificity.

[0067] b. Calculate the cross-segment extension entropy difference:

[0068] The cross-segment extension entropy difference is used to measure the difference between the growth distribution of the target growth curve in the global range and the overall extension characteristics of the normal growth curve. Extensibility describes the total change of the curve in a certain period of time, and is usually expressed by integrating the absolute value of the first-order derivative of the curve to represent the total growth amplitude of each time period. By calculating the extension of the target growth curve and the normal growth curve in each local reference window and normalizing them to the global distribution probability, the growth distribution characteristics of each curve in different parts of the time axis can be obtained. Subsequently, by calculating the entropy difference between the distribution probability of the target growth curve and the distribution probability of the normal growth curve, the difference in growth rhythm and distribution characteristics between the two is quantified. When the growth distribution of the target growth curve is concentrated in a few windows and deviates significantly from the distribution pattern of the normal growth curve, the entropy difference will increase significantly, reflecting the significant deviation between the overall growth rhythm of the target individual and the normal pattern. The specific calculation logic is as follows:

[0069] For both the target growth curve and the normal growth curve, the total extension in each reference window was calculated:

[0070] , ;

[0071] Based on the window extension, calculate the global extension distribution probability of the target growth curve and the extended distribution probability of the normal growth curve :

[0072] , ;

[0073] Calculate the cross-segment extension entropy difference , the formula is:

[0074] .

[0075] The cross-segment extension entropy difference is used to quantitatively evaluate the global difference between the overall extension characteristics of the target growth curve and the normal growth curve, and can reveal whether there is a global growth rhythm deviation or trend change in the target individual. This characteristic is of great value in judging the abnormal long-term growth trend of the target individual, especially in analyzing the rationality of the growth of the target individual within the upper and lower limits of the normal growth trajectory.

[0076] Through the implementation of step S2, two key characteristics are extracted from the comparison between the target growth curve and the normal growth curve - the relative curvature imbalance factor and the cross-segment extension entropy difference, which comprehensively characterize the growth characteristics of the target individual from the two perspectives of local fluctuation characteristics and global growth distribution. The relative curvature imbalance factor reveals the abnormal changes that may exist in the target individual in certain time periods, while the cross-segment extension entropy difference focuses on the difference between the overall growth rhythm of the target individual and the normal growth pattern, which is particularly suitable for identifying anomalies in long-term trends. These characteristics can not only reflect the dynamic characteristics of the target growth curve, but also lay a data foundation for subsequent short-term and long-term deviation assessments, truly realizing a deep analysis of the target growth behavior, and providing strong support for the application of the present invention in growth and development guidance.

[0077] In step S2, the relative curvature imbalance factor and the cross-segment extension entropy difference have been extracted from the comparison between the target growth curve and the normal growth curve, respectively quantifying the deviation characteristics of the target individual in the local fluctuation and global growth pattern. Step S3 aims to further utilize these characteristics to calculate the short-term and long-term deviation indexes from the perspective of time scale, clarify the dynamic difference between the growth trajectory of the target individual and the normal pattern through multi-time domain analysis, and provide a basis for subsequent credibility assessment and specificity analysis.

[0078] Step S3 includes the following contents:

[0079] Select the recent observation window of the target growth curve and define the time range as ,The window length is determined by the target data acquisition frequency.

[0080] In the selected window, the relative curvature imbalance factor and the cross-segment extension entropy difference are accumulated to calculate the overall mutation deviation degree in the window. :

[0081] ;

[0082] in, and It is a dynamically adjusted parameter used to balance the weights of local fluctuations and global growth characteristics in short-term analysis.

[0083] Short-term mutation deviation index It quantitatively describes the intensity of abnormal changes in the target growth curve within a recent window and can identify sudden deviations from the recent growth trajectory, such as accelerated growth or abnormal fluctuations.

[0084] The complete historical time range of the target growth curve Divide into several equally spaced time periods .

[0085] The relative curvature imbalance factor and the cross-segment extension entropy difference in each time period are accumulated and summed globally to calculate the long-term asymptotic deviation index. :

[0086] ;

[0087] The long-term asymptotic deviation index reflects the cumulative difference between the target growth curve and the normal growth curve over a long period of time, and is particularly suitable for capturing subtle changing trends in the growth trajectory.

[0088] Through the implementation of step S3, the deviation characteristics of the target growth curve are refined into two levels, short-term and long-term, in terms of time scale, and a short-term mutation deviation index and a long-term progressive deviation index are generated. The short-term index emphasizes the capture of abnormal changes in the recent growth trajectory, while the long-term index focuses on the evaluation of the cumulative overall deviation of the growth trajectory of the target individual. These indices can not only independently describe the growth characteristics in different time domains, but also provide multi-dimensional input for subsequent credibility analysis, further enhancing the comprehensive characterization capability of the present invention for the dynamic differences of the target growth trajectory. Combined with the background of the present invention, this deviation analysis method combining short-term and long-term effectively makes up for the shortcomings of the prior art in the analysis of the dynamics of the growth trajectory, and has important practical application value.

[0089] In step S3, the dynamic difference characteristics of the target growth curve and the normal growth curve at different time scales have been comprehensively evaluated by calculating the short-term mutation deviation index and the long-term asymptotic deviation index. The goal of step S4 is to deeply integrate these deviation indices with the relative curvature imbalance factor and the cross-segment extension entropy difference extracted in step S2, calculate the interpretation credibility coefficient of the normal growth curve for the target growth curve through an advanced machine learning model, and derive the growth-specific evaluation results of the target individual based on this, thereby providing a scientific basis for personalized growth and development guidance.

[0090] Step S4 includes the following contents:

[0091] First, the short-term mutation deviation index and long-term gradual deviation index generated in step S3 are fused with the relative curvature imbalance factor and cross-segment extension entropy difference extracted in step S2 to construct a multi-dimensional feature vector. A multi-level feature interaction network structure is adopted, in which each layer captures the complex correlation between different features through nonlinear transformation, ensuring that the relative curvature imbalance factor and cross-segment extension entropy difference can be complementary and fused in different time windows.

[0092] The multi-level feature interaction network structure is designed as follows:

[0093] Input layer: receives multi-dimensional feature vectors, namely, short-term mutation deviation index, long-term asymptotic deviation index, relative curvature imbalance factor and cross-segment extension entropy difference;

[0094] Hidden layer 1: Apply nonlinear activation functions (such as ReLU) to perform preliminary transformations on input features to generate intermediate representations;

[0095] Hidden layer 2: Captures high-order correlations between features through feature interaction mechanisms (such as feature multiplication or attention mechanisms) and applies nonlinear activation functions;

[0096] Hidden layer three: further enhances the complementary fusion of features and integrates features through convolution operations or fully connected layers;

[0097] Output layer: Generates fused feature representation for subsequent credibility coefficient calculation.

[0098] Then, the adaptive weight adjustment mechanism is used to dynamically allocate the contribution of the short-term mutation deviation index and the long-term gradual deviation index in the overall model according to the importance of the features, so as to optimize the sensitivity of the model to different deviation features. Specifically, the dynamic weight allocation formula based on the entropy weight method is introduced to calculate the information entropy of each feature. To determine its weight , the formula is as follows:

[0099] ;

[0100] in, For the The weight of the feature, is the information entropy of the feature, is the total number of features. Information entropy Calculated by the following formula:

[0101] ;

[0102] in Representation characteristics In category The conditional probability distribution under is the total number of categories. Weight The introduction of ensures that features with more information have a higher influence in model training.

[0103] Subsequently, the weighted feature vectors are input into a pre-trained deep neural network model, which outputs the interpretation credibility coefficient of the normal to target growth curve by learning the nonlinear relationship between the features. The calculation of the interpretation credibility coefficient is based on the model's comprehensive understanding of the data after feature fusion, reflecting the ability of the normal growth curve to interpret the target growth curve in the multidimensional feature space.

[0104] Finally, according to the numerical range of the interpretation credibility coefficient, the growth-specific assessment results of the target individual are divided into different levels, including high credibility assessment (normal growth curve can fully explain the characteristics of the target growth curve), medium credibility assessment (normal growth curve has partial explanatory power for the target growth curve) and low credibility assessment (normal growth curve is difficult to explain the characteristics of the target growth curve). This grading method not only provides a quantitative assessment of the growth characteristics of the target individual, but also provides a clear guiding direction for the formulation of subsequent personalized intervention plans.

[0105] According to the numerical range of the interpretation credibility coefficient, the growth-specific assessment results of the target individual are divided into three levels: high, medium, and low. For example, the specific division is as follows:

[0106] High confidence assessment: When the interpretation confidence coefficient is in the upper quartile of the overall distribution (i.e., above 75%), it indicates that the normal growth curve can fully explain the characteristics of the target growth curve.

[0107] Moderate confidence assessment: When the interpretation confidence coefficient is in the middle quartile (i.e., between 25% and 75%), it indicates that the normal growth curve has partial explanatory power for the target growth curve.

[0108] Low confidence assessment: When the interpretation confidence coefficient is in the lower quartile (i.e., below 25%), it means that the normal growth curve is difficult to explain the characteristics of the target growth curve.

[0109] Step S4 deeply integrates the short-term mutation deviation index, long-term asymptotic deviation index, relative curvature imbalance factor and cross-segment extension entropy difference of the target individual, uses a multi-level feature interaction network to capture the complex relationship between local fluctuations and global trends, and dynamically optimizes the feature contribution through an adaptive weight adjustment mechanism to generate the interpretation credibility coefficient of the normal growth curve for the target growth curve. Based on this credibility coefficient, the growth-specific assessment of the target individual is divided into three levels: high, medium and low, reflecting the situations where the normal growth curve can fully explain, partially explain and is difficult to explain the target growth curve. This process realizes the accurate quantification and graded assessment of the growth trajectory of the target individual, effectively reveals the dynamic differences in individual growth characteristics, provides a scientific basis for guiding personalized growth and development interventions, and enhances the accuracy and practicality of growth and development evaluation.

[0110] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0111] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0112] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0113] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A digital child growth and development predictive guidance method, characterized in that: Includes steps: S1: The normal growth curve is divided into multiple local reference windows to capture short-term fluctuation characteristics, and a long-term standard curve is constructed on a global scale. The local observation data of the target individual is mapped to the corresponding reference window through the dynamic time warping method and then secondary fused at the global level to generate the target growth curve; S2: Based on the difference between the target growth curve and the normal growth curve, the relative curvature imbalance factor is extracted using the local change of the curve to measure the local deviation, and the cross-segment extension entropy difference is calculated based on the cross-window characteristics to measure the overall extension and contraction characteristics; S3: Generate a short-term mutation deviation index by aggregating the deviation values ​​of the target growth curve within a short-term observation window, and obtain a long-term gradual deviation index by combining the complete historical data and accumulating them in segments, so as to evaluate the dynamic difference characteristics of the target growth curve in different time domains; S4: The short-term mutation deviation index and the long-term asymptotic deviation index are input into the machine learning model. By integrating the features of the relative curvature imbalance factor and the cross-segment extension entropy difference and adjusting the weights, the credibility coefficient of the normal explanation of the target is calculated and the growth-specific assessment results of the target individual are obtained.

2. A digital child growth and development predictive guidance method according to claim 1, characterized in that: Step S1 includes the following contents: By calculating the local change rate of the normal growth curve through the second-order derivative, the change trend of the curve is identified and the characteristic points are extracted, where the extreme point of the absolute value of the change rate is used to delineate the boundary of the segmentation window; according to the position of the characteristic point, the normal growth curve is divided into several local reference windows, and the curve segment in each window independently describes the growth characteristics in the corresponding time period to form a set of short-term characteristic curves; the characteristic curves of all local windows are spliced ​​according to the time series, and optimized at the segment boundary through the curve smoothness constraint to construct a global long-term standard curve to express the complete normal growth law; Subsequently, the growth observation data of the target individual are taken as discrete time series and aligned one by one with the local reference window of the normal growth curve using the dynamic time warping method; after completing the local mapping, all local fitting results are recombined according to the global time scale of the normal growth curve to construct the preliminary growth curve of the target individual; Finally, the target growth curve and the global long-term standard curve are fused secondary in the overall range; the fusion adopts the method of local feature enhancement, which retains the specific characteristics of the target growth curve in local fitting while ensuring that its trend at the global level is consistent with the normal growth curve, and finally generates a target growth curve that can both describe the growth characteristics of the target individual and compare with the normal standard.

3. A digital child growth and development predictive guidance method according to claim 2, characterized in that: The target growth curve and the global long-term standard curve are secondary fused within the overall range to generate a target growth curve that can describe the growth characteristics of the target individual and compare with the normal standard; the specific process is as follows: a. After the target individual growth observation data is aligned with the local reference window of the normal growth curve, the key characteristic values ​​of the target growth curve in each local window are extracted; b. The global standard trend of the normal growth curve is generated by splicing all local window curves, while ensuring the continuity and smoothness of the curve at the splicing points; c. The following mathematical model is used for fusion: ;in, is the final target growth curve; is the fitting result of the target growth curve in the local window; is the normal global standard curve; It is the strengthening factor of dynamic calculation.

4. A digital child growth and development predictive guidance method according to claim 3, characterized in that: Step S2 includes the following contents: a. Extract relative curvature imbalance factor: define the target growth curve as The normal growth curve is ; Both in time The curvatures are: , ; In each local reference window The cumulative value of the curvature deviation of the target growth curve relative to the normal growth curve is calculated, which is called the relative curvature imbalance factor. : ; The relative curvature imbalance factor is used to quantify the degree of deviation of the fluctuation characteristics of the target growth curve in the local reference window from the normal growth curve.

5. A digital child growth and development predictive guidance method according to claim 4, characterized in that: Step S2 also includes the following contents: b. Calculate the difference in cross-segment extension entropy: For the target growth curve and the normal growth curve, calculate the total extension in each reference window: , ; Based on the window extension, calculate the global extension distribution probability of the target growth curve and the extended distribution probability of the normal growth curve : , ; Calculate the cross-segment extension entropy difference , the formula is: The cross-segment extension entropy difference was used to quantitatively evaluate the global difference between the overall extension characteristics of the target growth curve and the normal growth curve.

6. A digital child growth and development predictive guidance method according to claim 5, characterized in that: Step S3 includes the following contents: Select the recent observation window of the target growth curve and define the time range as The window length is determined by the target data acquisition frequency; within the selected window, the relative curvature imbalance factor and the cross-segment extension entropy difference are accumulated to calculate the overall mutation deviation within the window. : ;in, and To dynamically adjust the parameters.

7. A digital child growth and development predictive guidance method according to claim 6, characterized in that: Step S3 also includes the following: The complete historical time range of the target growth curve Divide into several equally spaced time periods ; The relative curvature imbalance factor and the cross-segment extension entropy difference in each time period are accumulated and summed globally to calculate the long-term asymptotic deviation index : .

8. A digital child growth and development predictive guidance method according to claim 7, characterized in that: Step S4 includes the following contents: Firstly, the short-term mutation deviation index and the long-term gradual deviation index are fused with the relative curvature imbalance factor and the cross-segment extension entropy difference to construct a multi-dimensional feature vector. Then a multi-level feature interaction network structure is adopted, in which each layer captures the complex correlation between different features through nonlinear transformation, ensuring that the relative curvature imbalance factor and the cross-segment extension entropy difference can be complementarily fused in different time windows.

9. A digital child growth and development predictive guidance method according to claim 8, characterized in that: Step S4 also includes the following contents: Then, the adaptive weight adjustment mechanism is used to dynamically allocate the contribution of the short-term mutation deviation index and the long-term gradual deviation index in the overall model according to the importance of the features, so as to optimize the sensitivity of the model to different deviation features.

10. A digital child growth and development predictive guidance method according to claim 9, characterized in that: Step S4 also includes the following contents: Subsequently, the weighted feature vector is input into a pre-trained deep neural network model, which outputs the credibility coefficient of the normal to target growth curve by learning the nonlinear relationship between the features; Finally, according to the numerical range of the interpretation credibility coefficient, the growth-specific assessment results of the target individual are divided into different levels, including high credibility assessment, medium credibility assessment and low credibility assessment.

Images