Intervention effect evaluation system for overweight and obese children based on data intelligence

Abstract

The application relates to the field of data intelligence, and particularly discloses an intervention effect evaluation system and method for overweight and obese children based on data intelligence. By using artificial intelligence technology based on deep learning and neural networks, the system and method can deeply mine the correlation features of body index data of the overweight and obese children before and after intervention, and high-dimensional hidden features of exercise scheme data and diet scheme data in the intervention scheme, so that the effect of the intervention scheme for the overweight and obese children can be accurately and publicly evaluated on the basis of getting rid of the limitations and finiteness of human cognition. In this way, the problem of overweight and obesity of children and adolescents can be solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent intervention for overweight and obese children, and more specifically, to a data-intelligent-based system and method for evaluating the intervention effect of overweight and obese children. Background Technology

[0002] Over the past 30 years, the global rate of childhood overweight and obesity has been on the rise, becoming an increasingly serious public health problem. Between 1985 and 2014, the prevalence of overweight and obesity among school-aged children in China increased from 5.3% to 20.5%. In 2020, my country's "Implementation Plan for the Prevention and Control of Childhood and Adolescent Obesity" set forth the overall goal of reducing overweight and obesity rates. The "Healthy China 2030" Plan outlines the goal of cultivating self-disciplined healthy behaviors and significantly slowing the growth rate of the overweight and obese population by 2030. Overweight and obesity are risk factors for diabetes and hypertension in children and adolescents, and also impose psychological and cognitive burdens on this group. Childhood obesity often leads to adult obesity, accompanied by associated risks of chronic diseases, including cardiometabolic diseases, non-alcoholic fatty liver disease, type 2 diabetes, and kidney disease.

[0003] Studies have found that unhealthy lifestyle behaviors such as low dietary diversity, low physical activity, and sedentary behavior related to screen time are risk factors for obesity; a well-balanced diet with controlled calorie intake, regular physical exercise, and limiting screen time are beneficial for reducing body fat. Therefore, numerous lifestyle intervention studies have been conducted both domestically and internationally to explore effective intervention methods for preventing overweight and obesity. However, finding low-cost, sustainable, and easily scalable intervention methods is a pressing issue. Understandably, before promoting intervention methods, accurate and impartial evaluation of their effectiveness is crucial.

[0004] Therefore, an evaluation scheme is needed to assess the effectiveness of interventions in overweight and obese children. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a data-intelligence-based intervention effect evaluation system and method for overweight and obese children. By using artificial intelligence technologies based on deep learning and neural networks, it deeply mines the correlation characteristics of the physical indicator data of the overweight and obese children before and after intervention, as well as the high-dimensional implicit features of the exercise and dietary data in the intervention program. This allows for an accurate and impartial evaluation of the effectiveness of the intervention program for overweight and obese children, overcoming the limitations and constraints of human cognition. This can contribute to solving the problem of overweight and obesity in children and adolescents.

[0006] According to one aspect of this application, a data-driven intelligent intervention effect evaluation system for overweight and obese children is provided, comprising:

[0007] The pre- and post-intervention body indicator data acquisition unit is used to acquire the body indicator data of the obese children to be assessed before the intervention and the body indicator data of the obese children to be assessed after the intervention, including: body mass score, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content and protein content.

[0008] An intervention plan acquisition unit is used to acquire intervention plans, which include exercise plan data and diet plan data.

[0009] The pre-intervention indicator data encoding unit is used to obtain multiple feature vectors by passing the pre-intervention body indicator data through a context encoder containing an embedding layer, and to concatenate the multiple feature vectors to obtain a first feature vector.

[0010] The post-intervention indicator data encoding unit is used to pass the post-intervention body indicator data through the context encoder containing the embedding layer to obtain multiple feature vectors, and to concatenate the multiple feature vectors to obtain a second feature vector.

[0011] The body change feature extraction unit is used to calculate the transition matrix of the second feature vector relative to the first feature vector as the first feature matrix. The transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention.

[0012] The first intervention scheme encoding unit is used to generate a third feature vector by passing the motion scheme data of the intervention scheme through the temporal encoder of the joint encoder. The temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers.

[0013] The second intervention program encoding unit is used to generate a fourth feature vector by passing the dietary program data of the intervention program through the semantic encoding model of the joint encoder. The semantic encoding model includes an embedding layer and a bidirectional long short-term memory model.

[0014] The intervention scheme joint coding unit is used to use the cross-modal feature fusion module of the joint encoder to calculate the correlation matrix of the third feature vector and the fourth feature vector as the second feature matrix;

[0015] A feature fusion unit is configured to fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix; and

[0016] The evaluation result generation unit is used to pass the classification feature matrix through a classifier to obtain classification results, which are the evaluation labels for the effectiveness of the intervention program.

[0017] Compared with existing technologies, the data-intelligence-based intervention effect evaluation system and method for overweight and obese children provided in this application utilize artificial intelligence technologies based on deep learning and neural networks to deeply mine the correlation characteristics of the physical indicator data of overweight and obese children before and after intervention, as well as the high-dimensional implicit features of the exercise and diet data in the intervention program. This allows for an accurate and impartial evaluation of the effectiveness of the intervention program for overweight and obese children, overcoming the limitations and constraints of human cognition. This can contribute to solving the problem of overweight and obesity in children and adolescents. Attached Figure Description

[0018] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 This is a block diagram of a data-driven intelligent intervention effect evaluation system for overweight and obese children according to an embodiment of this application.

[0020] Figure 2 This is a flowchart of the evaluation method for the data intelligence-based intervention effect evaluation system for overweight and obese children according to an embodiment of this application.

[0021] Figure 3 This is a schematic diagram of the evaluation method of the data intelligence-based intervention effect evaluation system for overweight and obese children according to an embodiment of this application. Detailed Implementation

[0022] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0023] Scene Overview

[0024] As mentioned earlier, the global rate of childhood overweight and obesity has been increasing over the past 30 years, becoming a growing public health problem. Between 1985 and 2014, the prevalence of overweight and obesity among school-aged children in China rose from 5.3% to 20.5%. In 2020, my country's "Implementation Plan for the Prevention and Control of Childhood and Adolescent Obesity" set forth the overall goal of reducing overweight and obesity rates. The "Healthy China 2030" plan outlines the goal of cultivating self-disciplined healthy behaviors and significantly slowing the growth rate of the overweight and obese population by 2030. Overweight and obesity are risk factors for diabetes and hypertension in children and adolescents, and also impose psychological and cognitive burdens on this group. Childhood obesity often leads to adult obesity, accompanied by associated risks of chronic diseases, including cardiometabolic diseases, non-alcoholic fatty liver disease, type 2 diabetes, and kidney disease.

[0025] Studies have found that unhealthy lifestyle behaviors such as low dietary diversity, low physical activity, and sedentary behavior related to screen time are risk factors for obesity; a well-balanced diet with controlled calorie intake, regular physical exercise, and limiting screen time are beneficial for reducing body fat. Therefore, numerous lifestyle intervention studies have been conducted both domestically and internationally to explore effective intervention methods for preventing overweight and obesity. However, finding low-cost, sustainable, and easily scalable intervention methods is a pressing issue. Understandably, before promoting intervention methods, accurate and impartial evaluation of their effectiveness is crucial.

[0026] Therefore, an evaluation scheme is needed to assess the effectiveness of interventions in overweight and obese children.

[0027] Numerous randomized controlled trials abroad have demonstrated the effectiveness of lifestyle interventions for childhood obesity. In China, existing interventions for childhood overweight and obesity are beginning to incorporate a balanced diet combined with exercise. Previous obesity intervention studies have primarily been school-based, with some based in families, communities, or hospitals. Currently, this field lacks relatively flexible, sustainable, widely applicable, and low-cost intensive lifestyle intervention studies. Short-video health education, through vivid and engaging content, allows participants to gain a deeper understanding of relevant knowledge, representing a highly efficient, rapidly disseminated, and attractive potential method of information dissemination.

[0028] Based on this, the technical solution of this application is based on the research on the impact of enhanced lifestyle intervention using popular science short videos as an intervention method on overweight and obesity in children and adolescents, in order to promote the solution of the problem of overweight and obesity in children and adolescents.

[0029] In recent years, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and text signal processing. Furthermore, deep learning and neural networks have demonstrated near-human or even surpassed human-level performance in areas such as image classification, object detection, semantic segmentation, and text translation. The development of deep learning and neural networks has provided solutions and approaches for evaluating the effectiveness of interventions in overweight and obese children.

[0030] Compared to traditional evaluation methods, such as those based on statistical models or expert rules, evaluation schemes based on deep learning and neural networks can extract high-dimensional hidden features from the collected big data, thus overcoming the limitations and constraints of human cognition.

[0031] Specifically, in the technical solution of this application, the body indicator data of overweight and obese children before and after intervention are first obtained. The body indicator data includes: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content. Of course, in other embodiments of this application, more quantity and more dimensions of body indicator data can be collected to more comprehensively evaluate the body indicator data of overweight and obese children before and after intervention. In this embodiment, a context encoder including an embedding layer is used to encode the body indicator data before and after intervention to fully extract the implicit correlation features of each body indicator data in a high-dimensional feature space, thereby characterizing the body indicator features before and after intervention. Then, a transition matrix between the body indicator features before and after intervention is calculated to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after intervention.

[0032] Then, an intervention plan is obtained. In the technical solution of this application, the intervention plan includes exercise plan data and diet plan data. The exercise plan data includes exercise type and calorie consumption, and the diet plan data is the menu data for each meal on each day. Considering that the exercise plan data is sequence data and the diet plan is text data, a cross-modal joint encoder is used to encode the intervention plan in the technical solution of this application. Specifically, the temporal encoder of the joint encoder is used to encode the exercise plan data in the intervention plan, and the semantic encoding model of the joint encoder is used to encode the diet plan data in the intervention plan. The cross-modal feature fusion module of the joint encoder is used to calculate the correlation matrix of the third feature vector and the fourth feature vector to represent the high-dimensional hidden feature information of the intervention plan.

[0033] Then, by fusing the high-dimensional implicit feature information of the intervention plan and the implicit transformation feature information of the body indicators of the obese children to be evaluated before and after the intervention, the intervention effect of the intervention plan can be classified and judged. However, since the transition matrix is ​​used to represent the implicit transformation feature of the body indicators of the obese children to be evaluated before and after the intervention, while the second feature matrix is ​​used to represent the implicit feature information of the intervention plan, and the transition matrix has a certain response relationship with the implicit feature information, the response information needs to be considered when fusing the transition matrix and the second feature matrix.

[0034] Specifically, the fusion matrix is: in This represents exponentiation using a vector as a power. It means taking the value at each position of the vector as the exponent, and then filling the result into the corresponding positions of the vector to obtain the vector operation result. and Let represent positional subtraction and addition of two matrices, respectively. The representation of numbers and matrices is calculated element-wise. Hyperparameters used to control response weights.

[0035] This matrix fusion uses the second feature matrix as the response source of the transition matrix. Through a smooth response model, the latent feature representation of the transition matrix is ​​expressed as the posterior feature distribution of the second feature matrix, which is the prior feature distribution. Thus, when optimizing the model based on the loss function of a multi-label classifier using this fusion matrix, a smoother and more consistent transformation between the matrices to be fused can be achieved through optimization using this fusion matrix as a special form of the model's objective function. This improves the accuracy of evaluating the intervention effect of the intervention plan.

[0036] Based on this, this application proposes a data-driven intelligent intervention effect evaluation system for overweight and obese children, comprising: a pre- and post-intervention body indicator data acquisition unit, used to acquire body indicator data of the obese child to be evaluated before intervention and body indicator data of the obese child to be evaluated after intervention, including: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content; an intervention plan acquisition unit, used to acquire an intervention plan, the intervention plan including exercise plan data and diet plan data; a pre-intervention indicator data encoding unit, used to pass the pre-intervention body indicator data through a context encoder containing an embedding layer to obtain multiple feature vectors, and concatenate the multiple feature vectors to obtain a first feature vector; a post-intervention indicator data encoding unit, used to pass the post-intervention body indicator data through the context encoder containing an embedding layer to obtain multiple feature vectors, and concatenate the multiple feature vectors to obtain a second feature vector; and a body change feature extraction unit, used to calculate the transition matrix of the second feature vector relative to the first feature vector as... A first feature matrix, wherein the transition matrix represents the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention; a first intervention program encoding unit, used to generate a third feature vector by passing the exercise program data of the intervention program through the temporal encoder of the joint encoder, wherein the temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers; a second intervention program encoding unit, used to generate a fourth feature vector by passing the diet program data of the intervention program through the semantic encoding model of the joint encoder, wherein the semantic encoding model includes an embedding layer and a bidirectional long short-term memory model; an intervention program joint encoding unit, used to calculate the correlation matrix between the third feature vector and the fourth feature vector as a second feature matrix using the cross-modal feature fusion module of the joint encoder; a feature fusion unit, used to fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix; and an evaluation result generation unit, used to pass the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is an evaluation label for the effectiveness of the intervention program.

[0037] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0038] Exemplary System

[0039] Figure 1 The diagram illustrates a block diagram of a data-driven intelligent intervention effect evaluation system for overweight and obese children according to an embodiment of this application. Figure 1As shown, the data-driven intelligent intervention effect evaluation system 200 for overweight and obese children according to an embodiment of this application includes: a pre- and post-intervention body indicator data acquisition unit 210, used to acquire body indicator data of the obese child to be evaluated before intervention and body indicator data of the obese child to be evaluated after intervention, including: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content; an intervention plan acquisition unit 220, used to acquire an intervention plan, the intervention plan including exercise plan data and diet plan data; a pre-intervention indicator data encoding unit 230, used to pass the pre-intervention body indicator data through a context encoder containing an embedding layer to obtain multiple feature vectors, and concatenate the multiple feature vectors to obtain a first feature vector; a post-intervention indicator data encoding unit 240, used to pass the post-intervention body indicator data through the context encoder containing an embedding layer to obtain multiple feature vectors, and concatenate the multiple feature vectors to obtain a second feature vector; and a body change feature extraction unit 250, used to calculate the transition matrix of the second feature vector relative to the first feature vector. The system comprises: a first feature matrix, wherein the transition matrix represents the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention; a first intervention program encoding unit 260, which generates a third feature vector by passing the exercise program data of the intervention program through the temporal encoder of the joint encoder, wherein the temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers; a second intervention program encoding unit 270, which generates a fourth feature vector by passing the diet program data of the intervention program through the semantic encoding model of the joint encoder, wherein the semantic encoding model includes an embedding layer and a bidirectional long short-term memory model; an intervention program joint encoding unit 280, which calculates the correlation matrix between the third feature vector and the fourth feature vector as a second feature matrix using the cross-modal feature fusion module of the joint encoder; a feature fusion unit 290, which fuses the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix; and an evaluation result generation unit 300, which passes the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is an evaluation label for the effectiveness of the intervention program.

[0040] Specifically, in this embodiment, the pre- and post-intervention body indicator data acquisition unit 210 and the intervention plan acquisition unit 220 are used to acquire body indicator data of the obese child to be assessed before intervention and body indicator data of the obese child to be assessed after intervention, including: body mass score, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content, and to acquire the intervention plan, which includes exercise plan data and dietary plan data. As mentioned above, studies have found that poor lifestyle behaviors such as low dietary diversity, low physical activity, and screen-related sedentary behavior are risk factors for obesity; a well-structured diet with controlled calorie intake, regular physical exercise, and limiting screen time are beneficial for reducing body fat. Therefore, numerous lifestyle intervention studies have been conducted both domestically and internationally to explore effective intervention methods for preventing overweight and obesity. However, finding low-cost, sustainable, and easily promoted intervention methods is an urgent problem to be solved. It should be understood that before promoting intervention methods, how to accurately and fairly evaluate the intervention effect of intervention methods is crucial.

[0041] It is understandable that, compared to traditional evaluation methods, such as those based on statistical models or expert rules, evaluation schemes based on deep learning and neural networks can extract high-dimensional hidden features from the collected big data to overcome the limitations and constraints of human cognition.

[0042] Therefore, specifically, in the technical solution of this application, the first step is to obtain the body indicator data of overweight and obese children before and after the intervention. Here, the body indicator data includes: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content. Of course, in other embodiments of this application, more quantity and more dimensions of body indicator data can be collected to more comprehensively evaluate the body indicator data of the overweight and obese children before and after the intervention. Then, it is also necessary to obtain the intervention plan. In the technical solution of this application, the intervention plan includes exercise plan data and dietary plan data. Here, the exercise plan data includes the type of exercise and calorie consumption, and the dietary plan data is the menu data for each meal on each day.

[0043] Specifically, in this embodiment, the pre-intervention indicator data encoding unit 230 and the post-intervention indicator data encoding unit 240 are used to obtain multiple feature vectors from the pre-intervention body indicator data through a context encoder containing an embedding layer, and concatenate the multiple feature vectors to obtain a first feature vector; and to obtain multiple feature vectors from the post-intervention body indicator data through the context encoder containing an embedding layer, and concatenate the multiple feature vectors to obtain a second feature vector. That is, in the technical solution of this application, after obtaining the pre-intervention and post-intervention body indicator data, a context encoder containing an embedding layer is further used to encode the pre-intervention and post-intervention body indicator data to fully extract the implicit correlation features of each body indicator data in a high-dimensional feature space, thereby representing the pre-intervention and post-intervention body indicator features.

[0044] More specifically, in this embodiment, the pre-intervention indicator data encoding unit includes: first, using the embedding layer of the encoder model containing the embedding layer to convert the pre-intervention body indicator data into input vectors to obtain a sequence of input vectors; then, using the converter of the encoder model containing the embedding layer to perform global contextual semantic encoding on the sequence of input vectors to obtain multiple feature vectors; finally, concatenating the multiple feature vectors to obtain the first feature vector.

[0045] Specifically, in this embodiment, the body change feature extraction unit 250 is used to calculate the transition matrix of the second feature vector relative to the first feature vector as a first feature matrix. The transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention. It should be understood that in the technical solution of this application, after obtaining the second feature vector after intervention and the first feature vector before intervention, a transition matrix between the body indicator features before and after intervention is further calculated to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after intervention.

[0046] More specifically, in this embodiment of the application, the body change feature extraction unit is further configured to: calculate the transition matrix of the second feature vector relative to the first feature vector as the first feature matrix using the following formula;

[0047] The formula is as follows: in F Let T represent the first eigenvector, T represent the transition matrix, and S represent the second eigenvector.

[0048] Specifically, in this embodiment, the first intervention scheme encoding unit 260 and the second intervention scheme encoding unit 270 are used to generate a third feature vector by passing the motion scheme data of the intervention scheme through the temporal encoder of the joint encoder. The temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers. The second intervention scheme's dietary scheme data is then passed through the semantic encoding model of the joint encoder to generate a fourth feature vector. The semantic encoding model includes an embedding layer and a bidirectional long short-term memory model. It should be understood that, considering the motion scheme data is sequence data and the dietary scheme is text data, a cross-modal joint encoder is used to encode the intervention scheme in this technical solution. Specifically, the temporal encoder of the joint encoder is used to encode the motion scheme data in the intervention scheme, and the semantic encoding model of the joint encoder is used to encode the dietary scheme data in the intervention scheme, thereby obtaining the third feature vector containing high-dimensional implicit correlation feature information of the motion scheme and the fourth feature vector containing global dietary scheme data correlation feature information. Accordingly, in a specific example, the dietary plan data of the intervention program can first be segmented and then input into the embedding layer of the semantic coding model of the joint encoder to obtain a sequence of word embedding vectors corresponding to the daily dietary plan data; then, the sequence of word embedding feature vectors is passed through the bidirectional long short-term memory model of the semantic coding model of the joint encoder to obtain the fourth feature vector.

[0049] More specifically, in this embodiment, the first intervention scheme encoding unit is further configured to: arrange the motion scheme data of the intervention scheme into a one-dimensional input vector; and use the one-dimensional convolutional layer of the temporal encoder of the joint encoder to perform one-dimensional convolutional encoding on the input vector using the following formula to extract high-dimensional implicit correlation features between feature values ​​at each position in the input vector, wherein the formula is: in, a For the convolution kernel in x Width in direction F For the convolution kernel parameter vector, G For the local vector matrix that operates with the convolution kernel function, w Where is the size of the convolutional kernel; the fully connected layer of the temporal encoder of the joint encoder is used to fully encode the input vector using the following formula to extract the high-dimensional latent features of the feature values ​​at each position in the input vector, wherein the formula is: ,in It is the input vector, It is the output vector. It is a weight matrix. It is a bias vector. This represents matrix multiplication.

[0050] Specifically, in this embodiment, the intervention scheme joint encoding unit 280 is used to calculate the correlation matrix of the third feature vector and the fourth feature vector as a second feature matrix using the cross-modal feature fusion module of the joint encoder. That is, in the technical solution of this application, after obtaining the third feature vector with high-dimensional implicit correlation feature information of the exercise scheme and the fourth feature vector with global dietary scheme data correlation feature information, the cross-modal feature fusion module of the joint encoder is further used to calculate the correlation matrix of the third feature vector and the fourth feature vector to represent the high-dimensional implicit feature information of the intervention scheme, thereby obtaining the second feature matrix.

[0051] More specifically, in the embodiments of this application, the intervention scheme joint coding unit is further configured to: use the cross-modal feature fusion module of the joint encoder to calculate the correlation matrix of the third feature vector and the fourth feature vector as the second feature matrix using the following formula;

[0052] The formula is as follows: in The third feature vector, For the fourth feature vector, Let be the correlation matrix.

[0053] Specifically, in this embodiment, the feature fusion unit 290 is used to fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix. It should be understood that in the technical solution of this application, after obtaining the first feature matrix and the second feature matrix, further fusing the high-dimensional implicit feature information of the intervention scheme and the implicit transformation feature information of the body indicators of the obese child to be evaluated before and after the intervention can classify and judge the intervention effect of the intervention scheme. However, since the transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention, and the second feature matrix is ​​used to represent the implicit feature information of the intervention scheme, the transition matrix has a certain response relationship with the implicit feature information. Therefore, when fusing the transition matrix and the second feature matrix, the response information needs to be considered. That is, specifically, the first feature matrix and the second feature matrix are further fused based on the response relationship between the first feature matrix and the second feature matrix to obtain the classification feature matrix.

[0054] More specifically, in embodiments of this application, the feature fusion unit is further configured to: fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix using the following formula to obtain the classification feature matrix;

[0055] The formula is as follows: in This represents exponentiation using a matrix as a power, where the value at each position in the matrix is ​​used as the exponent, and the result is then populated into the corresponding positions in the matrix to obtain the matrix operation result. and Let represent positional subtraction and addition of two matrices, respectively. The representation of numbers and matrices is calculated element-wise. To control the hyperparameters of the response weights, it should be understood that the matrix fusion, by using the second feature matrix as the response source of the transition matrix, uses a smooth response model to represent the latent feature expression of the transition matrix as the posterior feature distribution of the second feature matrix, which is the prior feature distribution. Thus, when optimizing the model based on the loss function of a multi-label classifier using the fusion matrix, a smoother and more consistent transformation between the matrices to be fused can be achieved through optimization using the fusion matrix as a special form of the model objective function. This improves the accuracy of evaluating the intervention effect of the intervention plan.

[0056] Specifically, in this embodiment, the evaluation result generation unit 300 is used to pass the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is an effect evaluation label for the intervention program. That is, in this technical solution, the classification feature matrix is ​​further passed through a classifier to obtain a classification result representing an effect evaluation label for the intervention program. In a specific example, the classifier processes the classification feature matrix using the following formula to generate the classification result, wherein the formula is: ,in This indicates that the classification feature matrix is ​​projected into a vector. to Here are the weight matrices for each fully connected layer. to This represents the bias matrix of each fully connected layer.

[0057] In summary, the data-driven intelligent intervention effect evaluation system 200 for overweight and obese children, as described in this application embodiment, is explained. It utilizes artificial intelligence technology based on deep learning and neural networks to deeply mine the correlation characteristics of the physical indicator data of overweight and obese children before and after intervention, as well as the high-dimensional implicit features of the exercise and diet data within the intervention plan. This allows for an accurate and impartial evaluation of the effectiveness of the intervention plan for overweight and obese children, overcoming the limitations and constraints of human cognition. This can contribute to solving the problem of overweight and obesity in children and adolescents.

[0058] As described above, the data-intelligence-based intervention effect evaluation system 200 for overweight and obese children according to embodiments of this application can be implemented in various terminal devices, such as servers for data-intelligence-based intervention effect evaluation algorithms for overweight and obese children. In one example, the data-intelligence-based intervention effect evaluation system 200 for overweight and obese children according to embodiments of this application can be integrated into a terminal device as a software module and / or a hardware module. For example, the data-intelligence-based intervention effect evaluation system 200 for overweight and obese children can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the data-intelligence-based intervention effect evaluation system 200 for overweight and obese children can also be one of many hardware modules of the terminal device.

[0059] Alternatively, in another example, the data-driven intelligent intervention effect evaluation system 200 for overweight and obese children and the terminal device can also be separate devices, and the data-driven intelligent intervention effect evaluation system 200 for overweight and obese children can be connected to the terminal device via wired and / or wireless networks, and transmit interactive information in accordance with the agreed data format.

[0060] Exemplary methods

[0061] Figure 2 The diagram illustrates the evaluation method of a data-driven intelligence-based intervention effectiveness assessment system for overweight and obese children. Figure 2As shown, the evaluation method of the data-intelligence-based intervention effect evaluation system for overweight and obese children according to an embodiment of this application includes the following steps: S110, acquiring the body indicator data of the obese child to be evaluated before intervention and the body indicator data of the obese child to be evaluated after intervention, including: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content; S120, acquiring the intervention plan, the intervention plan including exercise plan data and diet plan data; S130, passing the body indicator data before intervention through a context encoder containing an embedding layer to obtain multiple feature vectors, and concatenating the multiple feature vectors to obtain a first feature vector; S140, passing the body indicator data after intervention through the context encoder containing an embedding layer to obtain multiple feature vectors, and concatenating the multiple feature vectors to obtain a second feature vector; S150, calculating the transition matrix of the second feature vector relative to the first feature vector as the first feature vector. The feature matrix, wherein the transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention; S160, the exercise program data of the intervention program is passed through the temporal encoder of the joint encoder to generate a third feature vector, the temporal encoder being composed of alternating one-dimensional convolutional layers and fully connected layers; S170, the diet program data of the intervention program is passed through the semantic encoding model of the joint encoder to generate a fourth feature vector, the semantic encoding model including an embedding layer and a bidirectional long short-term memory model; S180, the cross-modal feature fusion module of the joint encoder is used to calculate the correlation matrix between the third feature vector and the fourth feature vector as a second feature matrix; S190, the first feature matrix and the second feature matrix are fused based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix; and S200, the classification feature matrix is ​​passed through a classifier to obtain a classification result, the classification result being the effect evaluation label of the intervention program.

[0062] Figure 3 The illustration shows a schematic diagram of the evaluation method for a data-driven intelligence-based intervention effect evaluation system for overweight and obese children, according to an embodiment of this application. Figure 3 As shown, in the network architecture of the evaluation method of the data-driven intelligence-based intervention effect evaluation system for overweight and obese children, firstly, the obtained pre-intervention body indicator data (e.g., such as...) are... Figure 3 The P1 shown is transmitted through a context encoder containing an embedding layer (e.g., as shown in the image). Figure 3 E1 as shown in the figure) to obtain multiple feature vectors (e.g., such as Figure 3 The VF1 shown in the diagram), and concatenate the multiple feature vectors to obtain a first feature vector (e.g., as shown in the diagram). Figure 3The VF2 shown in the diagram); then, the obtained post-intervention body index data (e.g., such as...) Figure 3 P2 as shown) is transmitted through the context encoder containing the embedding layer (e.g., as shown in the figure). Figure 3 E1 as shown in the figure) to obtain multiple feature vectors (e.g., such as Figure 3 The VF3 shown in the diagram), and concatenate the multiple feature vectors to obtain a second feature vector (e.g., as shown in the diagram). Figure 3 (as shown in VF4); then, the transition matrix of the second eigenvector relative to the first eigenvector is calculated as the first eigenvector matrix (e.g., as shown in VF4). Figure 3 The intended MF1); then, the movement plan of the intervention program (e.g., as shown) Figure 3 The Q1 data shown is transmitted through a time encoder (e.g., as indicated by the joint encoder) by the time encoder. Figure 3 E2 as shown in the figure is used to generate a third feature vector (e.g., as shown in the figure). Figure 3 The VF5 shown); then, the dietary regimen of the intervention (e.g., as shown) Figure 3 The Q2 data shown is encoded by the semantic encoding model of the joint encoder (e.g., as shown in the image). Figure 3 The SUM shown is used to generate the fourth feature vector (e.g., as shown in the figure). Figure 3 The VF6 shown in the diagram); then, the cross-modal feature fusion module of the joint encoder (e.g., as shown in the diagram) is used. Figure 3 The correlation matrix between the third eigenvector and the fourth eigenvector (as shown in E3) is used to calculate the second eigenma matrix (e.g., as shown in E3). Figure 3 (as shown in MF2); then, based on the response relationship between the first feature matrix and the second feature matrix, the first feature matrix and the second feature matrix are fused to obtain a classification feature matrix (e.g., as shown in MF2). Figure 3 The MF shown in the diagram); and finally, the classification feature matrix is ​​passed through a classifier (e.g., such as...). Figure 3 (The classifier shown in the figure) is used to obtain the classification result, which is the label for evaluating the effectiveness of the intervention program.

[0063] More specifically, in steps S110 and S120, the body indicator data of the obese child to be assessed before the intervention and the body indicator data of the obese child to be assessed after the intervention are obtained, including: body mass score, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content, and an intervention plan is obtained, which includes exercise plan data and dietary plan data. It should be understood that compared to traditional assessment methods, such as assessment methods based on statistical models or expert rules, assessment methods based on deep learning and neural networks can extract high-dimensional implicit features from the collected big data to overcome the limitations and constraints of human cognition.

[0064] Therefore, specifically, in the technical solution of this application, the first step is to obtain the body indicator data of overweight and obese children before and after the intervention. Here, the body indicator data includes: body mass fraction, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content, and protein content. Of course, in other embodiments of this application, more quantity and more dimensions of body indicator data can be collected to more comprehensively evaluate the body indicator data of the overweight and obese children before and after the intervention. Then, it is also necessary to obtain the intervention plan. In the technical solution of this application, the intervention plan includes exercise plan data and dietary plan data. Here, the exercise plan data includes the type of exercise and calorie consumption, and the dietary plan data is the menu data for each meal on each day.

[0065] More specifically, in steps S130 and S140, the pre-intervention body indicator data is processed through a context encoder containing an embedding layer to obtain multiple feature vectors, and the multiple feature vectors are concatenated to obtain a first feature vector. Similarly, the post-intervention body indicator data is processed through the context encoder containing an embedding layer to obtain multiple feature vectors, and the multiple feature vectors are concatenated to obtain a second feature vector. That is, in the technical solution of this application, after obtaining the pre- and post-intervention body indicator data, a context encoder containing an embedding layer is further used to encode the pre- and post-intervention body indicator data to fully extract the implicit correlation features of each body indicator data in a high-dimensional feature space, thereby characterizing the pre- and post-intervention body indicator features.

[0066] More specifically, in this embodiment, the pre-intervention indicator data encoding unit includes: first, using the embedding layer of the encoder model containing the embedding layer to convert the pre-intervention body indicator data into input vectors to obtain a sequence of input vectors; then, using the converter of the encoder model containing the embedding layer to perform global contextual semantic encoding on the sequence of input vectors to obtain multiple feature vectors; finally, concatenating the multiple feature vectors to obtain the first feature vector.

[0067] More specifically, in step S150, the transition matrix of the second feature vector relative to the first feature vector is calculated as the first feature matrix. This transition matrix represents the implicit transformation characteristics of the body indicators of the obese child to be evaluated before and after the intervention. It should be understood that in the technical solution of this application, after obtaining the second feature vector after intervention and the first feature vector before intervention, a transition matrix between the body indicator features before and after intervention is further calculated to represent the implicit transformation characteristics of the body indicators of the obese child to be evaluated before and after intervention.

[0068] More specifically, in steps S160 and S170, the exercise plan data of the intervention plan is processed by the temporal encoder of the joint encoder to generate a third feature vector. The temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers. The diet plan data of the intervention plan is processed by the semantic encoding model of the joint encoder to generate a fourth feature vector. The semantic encoding model includes an embedding layer and a bidirectional long short-term memory model. It should be understood that, considering the exercise plan data is sequence data and the diet plan is text data, a cross-modal joint encoder is used to encode the intervention plan in this application. Specifically, the temporal encoder of the joint encoder is used to encode the exercise plan data in the intervention plan, and the semantic encoding model of the joint encoder is used to encode the diet plan data in the intervention plan, thereby obtaining the third feature vector containing high-dimensional implicit correlation feature information of the exercise plan and the fourth feature vector containing global correlation feature information of the diet plan data. Accordingly, in a specific example, the dietary plan data of the intervention program can first be segmented and then input into the embedding layer of the semantic coding model of the joint encoder to obtain a sequence of word embedding vectors corresponding to the daily dietary plan data; then, the sequence of word embedding feature vectors is passed through the bidirectional long short-term memory model of the semantic coding model of the joint encoder to obtain the fourth feature vector.

[0069] More specifically, in step S180, the cross-modal feature fusion module of the co-encoder is used to calculate the correlation matrix between the third feature vector and the fourth feature vector as the second feature matrix. That is, in the technical solution of this application, after obtaining the third feature vector containing high-dimensional implicit correlation feature information of the exercise plan and the fourth feature vector containing global dietary plan data correlation feature information, the cross-modal feature fusion module of the co-encoder is further used to calculate the correlation matrix between the third feature vector and the fourth feature vector to represent the high-dimensional implicit feature information of the intervention plan, thereby obtaining the second feature matrix.

[0070] More specifically, in step S190, the first feature matrix and the second feature matrix are fused based on the response relationship between them to obtain a classification feature matrix. It should be understood that in the technical solution of this application, after obtaining the first feature matrix and the second feature matrix, further fusing the high-dimensional implicit feature information of the intervention plan and the implicit transformation feature information of the obese child's physical indicators before and after the intervention can classify and judge the intervention effect of the intervention plan. However, since the transition matrix is ​​used to represent the implicit transformation features of the obese child's physical indicators before and after the intervention, and the second feature matrix is ​​used to represent the implicit feature information of the intervention plan, the transition matrix has a certain response relationship with the implicit feature information. Therefore, when fusing the transition matrix and the second feature matrix, the response information needs to be considered. That is, specifically, the first feature matrix and the second feature matrix are further fused based on the response relationship between them to obtain the classification feature matrix.

[0071] More specifically, in step S200, the classification feature matrix is ​​processed by a classifier to obtain a classification result, which is a label for evaluating the effectiveness of the intervention program. That is, in this application's technical solution, the classification feature matrix is ​​further processed by a classifier to obtain a classification result representing a label for evaluating the effectiveness of the intervention program. In a specific example, the classifier processes the classification feature matrix using the following formula to generate the classification result, wherein the formula is: ,in This indicates that the classification feature matrix is ​​projected into a vector. to Here are the weight matrices for each fully connected layer. to This represents the bias matrix of each fully connected layer.

[0072] In summary, the evaluation method of the data-intelligence-based intervention effect evaluation system for overweight and obese children, as described in the embodiments of this application, is elucidated. It utilizes artificial intelligence technologies based on deep learning and neural networks to deeply mine the correlation characteristics of the physical indicator data of overweight and obese children before and after intervention, as well as the high-dimensional implicit features of the exercise and dietary data within the intervention plan. This allows for an accurate and impartial evaluation of the effectiveness of the intervention plan for overweight and obese children, overcoming the limitations and constraints of human cognition. This can contribute to addressing the problem of overweight and obesity in children and adolescents.

[0073] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0074] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0075] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0076] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0077] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A data-driven intelligence-based intervention effect evaluation system for overweight and obese children, characterized in that, include: The pre- and post-intervention body indicator data acquisition unit is used to acquire the body indicator data of the obese children to be assessed before the intervention and the body indicator data of the obese children to be assessed after the intervention, including: body mass score, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content and protein content. An intervention plan acquisition unit is used to acquire intervention plans, which include exercise plan data and diet plan data. The pre-intervention indicator data encoding unit is used to obtain multiple feature vectors by passing the pre-intervention body indicator data through a context encoder containing an embedding layer, and to concatenate the multiple feature vectors to obtain a first feature vector. The post-intervention indicator data encoding unit is used to pass the post-intervention body indicator data through the context encoder containing the embedding layer to obtain multiple feature vectors, and to concatenate the multiple feature vectors to obtain a second feature vector. The body change feature extraction unit is used to calculate the transition matrix of the second feature vector relative to the first feature vector as the first feature matrix. The transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention. The first intervention scheme encoding unit is used to generate a third feature vector by passing the motion scheme data of the intervention scheme through the temporal encoder of the joint encoder. The temporal encoder consists of alternating one-dimensional convolutional layers and fully connected layers. The second intervention program encoding unit is used to generate a fourth feature vector by passing the dietary program data of the intervention program through the semantic encoding model of the joint encoder. The semantic encoding model includes an embedding layer and a bidirectional long short-term memory model. The intervention scheme joint coding unit is used to use the cross-modal feature fusion module of the joint encoder to calculate the correlation matrix of the third feature vector and the fourth feature vector as the second feature matrix; A feature fusion unit is configured to fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix to obtain a classification feature matrix; and The evaluation result generation unit is used to pass the classification feature matrix through a classifier to obtain classification results, which are the evaluation labels for the effectiveness of the intervention program.

2. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 1, wherein, The pre-intervention indicator data encoding unit is further configured to: use the embedding layer of the encoder model containing the embedding layer to convert the pre-intervention body indicator data into input vectors to obtain a sequence of input vectors; The converter of the encoder model containing the embedding layer performs global contextual semantic encoding on the sequence of input vectors to obtain multiple feature vectors; and concatenates the multiple feature vectors to obtain the first feature vector.

3. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 2, wherein, The body change feature extraction unit is further configured to: calculate the transition matrix of the second feature vector relative to the first feature vector as the first feature matrix using the following formula; The formula is as follows: in F Let T represent the first eigenvector, T represent the transition matrix, and S represent the second eigenvector.

4. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 3, wherein, The first intervention scheme encoding unit is further configured to: arrange the motion scheme data of the intervention scheme into a one-dimensional input vector; and use the one-dimensional convolutional layer of the temporal encoder of the joint encoder to perform one-dimensional convolutional encoding on the input vector using the following formula to extract high-dimensional implicit correlation features between feature values ​​at each position in the input vector, wherein the formula is: in, a For the convolution kernel in x Width in direction F For the convolution kernel parameter vector, G For the local vector matrix that operates with the convolution kernel function, w Where is the size of the convolutional kernel; the fully connected layer of the temporal encoder of the joint encoder is used to fully encode the input vector using the following formula to extract the high-dimensional latent features of the feature values ​​at each position in the input vector, wherein the formula is: ,in It is the input vector, It is the output vector. It is a weight matrix. It is a bias vector. This represents matrix multiplication.

5. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 4, wherein, The second intervention scheme encoding unit is further configured to: segment the dietary plan data of the intervention scheme into words and input them into the embedding layer of the semantic encoding model of the joint encoder to obtain a sequence of word embedding vectors corresponding to the daily dietary plan data; and pass the sequence of word embedding feature vectors through the bidirectional long short-term memory model of the semantic encoding model of the joint encoder to obtain the fourth feature vector.

6. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 5, wherein, The intervention scheme joint coding unit is further configured to: use the cross-modal feature fusion module of the joint encoder to calculate the correlation matrix of the third feature vector and the fourth feature vector as the second feature matrix using the following formula; The formula is as follows: in The third feature vector, For the fourth feature vector, Let be the correlation matrix.

7. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 6, wherein, The feature fusion unit is further configured to: fuse the first feature matrix and the second feature matrix based on the response relationship between the first feature matrix and the second feature matrix using the following formula to obtain the classification feature matrix; The formula is as follows: in This represents exponentiation using a matrix as a power, where the value at each position in the matrix is ​​used as the exponent, and the result is then populated into the corresponding positions in the matrix to obtain the matrix operation result. and Let represent positional subtraction and addition of two matrices, respectively. The representation of numbers and matrices is calculated element-wise. This is a hyperparameter.

8. The data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 7, wherein, The evaluation result generation unit is further configured to: process the classification feature matrix using the following formula to generate a classification result, wherein the formula is: ,in This indicates that the classification feature matrix is ​​projected into a vector. to Here are the weight matrices for each fully connected layer. to This represents the bias matrix of each fully connected layer.

9. An evaluation method for a data-driven intelligent intervention effect evaluation system for overweight and obese children, characterized in that, include: Obtain physical index data of the obese children to be assessed before intervention and physical index data of the obese children to be assessed after intervention, including: body mass score, weight, fat percentage, water percentage, basal metabolic rate, visceral fat level, muscle mass, bone mineral content and protein content. Obtain intervention plans, which include exercise plan data and dietary plan data; The pre-intervention body index data is passed through a context encoder containing an embedding layer to obtain multiple feature vectors, and the multiple feature vectors are concatenated to obtain a first feature vector. The body index data after the intervention is passed through the context encoder containing the embedding layer to obtain multiple feature vectors, and the multiple feature vectors are concatenated to obtain a second feature vector; The transition matrix of the second feature vector relative to the first feature vector is calculated as the first feature matrix, and the transition matrix is ​​used to represent the implicit transformation features of the body indicators of the obese child to be evaluated before and after the intervention. The motion scheme data of the intervention scheme is passed through the temporal encoder of the joint encoder to generate a third feature vector, the temporal encoder being composed of alternating one-dimensional convolutional layers and fully connected layers; The dietary plan data of the intervention program is processed through the semantic encoding model of the joint encoder to generate a fourth feature vector, the semantic encoding model including an embedding layer and a bidirectional long short-term memory model; The cross-modal feature fusion module of the joint encoder is used to calculate the correlation matrix between the third feature vector and the fourth feature vector as the second feature matrix; The first feature matrix and the second feature matrix are fused based on the response relationship between them to obtain a classification feature matrix; and The classification feature matrix is ​​passed through a classifier to obtain classification results, which are used as labels for evaluating the effectiveness of the intervention program.

10. The evaluation method for the data-driven intelligent intervention effect evaluation system for overweight and obese children according to claim 9, wherein, The pre-intervention body index data is processed through a context encoder containing an embedding layer to obtain multiple feature vectors, and the multiple feature vectors are concatenated to obtain a first feature vector, including: The embedding layer of the encoder model containing the embedding layer transforms the pre-intervention body index data into input vectors to obtain a sequence of input vectors; the converter of the encoder model containing the embedding layer performs global contextual semantic encoding on the sequence of input vectors to obtain multiple feature vectors; and the multiple feature vectors are concatenated to obtain the first feature vector.

Images