Method and device for predicting adult height in juvenile stage based on machine learning
By combining particle swarm algorithm and support vector machine regression, a PSO-SVR model with adaptive weight coefficients and learning factor updates is proposed, which solves the shortcomings of traditional methods in predicting the adult height of children and adolescents, and improves the prediction accuracy and application value.
Patent Information
- Application Number
- CN202411004595.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-12-06
AI Technical Summary
Traditional methods have problems such as outdated methods and inapplicable to Chinese children when predicting adult heights of children and adolescents, and lack effective prediction methods.
Combining particle swarm algorithm (PSO) and support vector machine regression (SVR), a PSO-SVR model with adaptive weight coefficients and learning factor updates is proposed. Through the improvement of the algorithm, the global search and local optimization capabilities are improved to predict adult height.
The performance and accuracy of the prediction model are improved and have high application value in adult height prediction.
Smart Images

Figure CN119092139A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method and device for predicting adult height in the juvenile stage based on machine learning. Background Art
[0002] Predicting adult height during childhood and adolescence has important clinical, sports, psychological and scientific significance, which helps promote healthy development, rational intervention and scientific research. The growth of height in children and adolescents is a continuous and nonlinear process, and is affected by many factors, such as bone age, age, gender, etc.
[0003] When predicting adult height, traditional methods have problems such as being outdated and not applicable to Chinese children and adolescents. The development of artificial intelligence technology has provided a new approach for research on predicting adult height of children and adolescents. Summary of the invention
[0004] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provide a method and device for predicting adult height in the juvenile stage.
[0005] The present invention solves the technical problem by adopting the following technical solution:
[0006] A first aspect of the present invention relates to a method for predicting adult height from the juvenile stage based on machine learning, comprising the following steps:
[0007] Step S1: Model integration, combining particle swarm optimization (PSO) with support vector machine regression (SVR) to propose a PSO-SVR model;
[0008] Step S2: Algorithm optimization, propose adaptive weight coefficient and learning factor update method, improve the PSO algorithm in step S1, and improve its global search and local optimization capabilities;
[0009] Step S3: Use the improved algorithm in step S2 to predict height.
[0010] Wherein, step S1 specifically includes:
[0011] Step S1-1: Data preparation. First, prepare the dataset for training and testing, and divide the dataset into feature matrix (input) and target vector (output). Make sure the data has been preprocessed and standardized as necessary.
[0012] Step S1-2: Parameter setting. Set the relevant parameters of PSO and SVR, including the number of particles, number of iterations, inertia weight, acceleration factor, kernel function type, and regularization parameter, etc. The selection of these parameters can be determined by cross-validation or experience.
[0013] Step S1-3: Initialize the particle swarm. Randomly initialize the position and velocity of each particle in the particle swarm. The position of the particle represents the hyperparameter combination of the SVR model, and the velocity represents the moving direction and amplitude of the particle in the search space.
[0014] Step S1-4: Calculate fitness. For each particle, construct an SVR model based on its position and calculate the corresponding fitness using the training data. The fitness can be measured using the prediction error (such as mean square error) of the SVR model on the training set.
[0015] Step S1-5: Update particle speed and position. Using the idea of particle swarm optimization, update the speed and position of particles according to fitness and global optimal solution. Through iterative updates, particles will gradually gather to a better position.
[0016] Step S1-6: Repeat steps S1-4 and S1-5 until the preset number of iterations or convergence condition is reached. Each iteration updates the velocity and position of the particle and calculates a new fitness.
[0017] Step S1-7: Select the best model. After the iteration is completed, the SVR model with the best performance is selected as the final prediction model according to the final fitness value and particle position.
[0018] Wherein, step S2 specifically includes:
[0019] Step S2-1: The present invention dynamically updates the weight coefficient by calculating the Euclidean distance between the particle position and the optimal particle in each stage. The formula of the adaptive weight coefficient is as follows:
[0020]
[0021] Among them, w 0 is the initialization weight coefficient, generally between [0.5, 1], w i The weight coefficient after each stage update, is the position of the ith particle at time t.
[0022] Step S2-2: The sin and cos functions have the characteristics of periodic changes. By applying them to the update of the learning factor, the learning factor can show periodic changes in the iteration process. This periodic change can help the particle swarm to conduct more comprehensive and diverse exploration in the search space, thereby increasing the chance of finding the global optimal solution. The formula for updating the learning factor is as follows:
[0023]
[0024] Where i is the current iteration number, is the cognitive learning factor of the i-th iteration, is the social learning factor of the i-th iteration, and max_iter is the maximum number of iterations. When the learning factor is large, the particle swarm focuses more on development, that is, moving towards the current best solution; when the learning factor is small, the particle swarm focuses more on exploration and searches the solution space in a wider way. This balance helps avoid falling into the local optimal solution while effectively converging to a better solution.
[0025] Wherein, step S3 specifically includes:
[0026] Step S3-1: In order to obtain the optimal solution of the penalty parameter C, kernel parameter σ and boundary error ε in the SVR model, the parameters of the SVR are optimized by the improved ACPSO algorithm in step S2, and the optimal parameters are selected to improve the height prediction performance of the SVR model. The inverse mean square error of the adult height prediction result is used as the fitness function of the ACPSO algorithm, and the calculation formula is as follows:
[0027]
[0028] Where N is the total number of height samples, represents the predicted value of the i-th sample, y i represents the true value of the i-th sample, f MSE Represents the mean square error of the prediction results.
[0029] Step S3-2: Use the ACPSO algorithm to select the optimal parameters of the hyperparameters of the SVR model, repeat step S1 to form an ACPSO-SVR prediction model, and perform height prediction.
[0030] The second aspect of the present invention relates to a device for predicting adult height from the juvenile stage based on machine learning, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the method of predicting adult height from the juvenile stage based on machine learning of the present invention.
[0031] The third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the method of the present invention for predicting adult height at the juvenile stage based on machine learning.
[0032] The present invention has the following beneficial effects:
[0033] 1) This paper proposes a method for predicting the adult height of children and adolescents based on the ACPSO-SVR model, which innovatively adopts a particle swarm optimization algorithm with adaptive variable parameters to improve the performance and accuracy of the prediction model.
[0034] 2) The present invention has high application value in adult height prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 It is the overall flow chart of the present invention.
[0036] Figure 2 It is a flow chart of the ACPSO-SVR model used in the present invention. DETAILED DESCRIPTION
[0037] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0038] Example 1
[0039] Refer to the attached Figure 1 and Figure 2 , a method for predicting adult height from the juvenile stage based on machine learning, comprising the following steps:
[0040] Step S1: Model integration, combining particle swarm optimization (PSO) with support vector machine regression (SVR) to propose a PSO-SVR model;
[0041] Step S2: Algorithm optimization, propose adaptive weight coefficient and learning factor update method, improve the PSO algorithm in step S1, and improve its global search and local optimization capabilities;
[0042] Step S3: Use the improved algorithm in step S2 to predict height.
[0043] Wherein, step S1 specifically includes:
[0044] Step S1-1: Data preparation. First, prepare the dataset for training and testing, and divide the dataset into feature matrix (input) and target vector (output). Make sure the data has been preprocessed and standardized as necessary.
[0045] Step S1-2: Parameter setting. Set the relevant parameters of PSO and SVR, including the number of particles, number of iterations, inertia weight, acceleration factor, kernel function type, and regularization parameter, etc. The selection of these parameters can be determined by cross-validation or experience.
[0046] Step S1-3: Initialize the particle swarm. Randomly initialize the position and velocity of each particle in the particle swarm. The position of the particle represents the hyperparameter combination of the SVR model, and the velocity represents the moving direction and amplitude of the particle in the search space.
[0047] Step S1-4: Calculate fitness. For each particle, construct an SVR model based on its position and calculate the corresponding fitness using the training data. The fitness can be measured using the prediction error (such as mean square error) of the SVR model on the training set.
[0048] Step S1-5: Update particle speed and position. Using the idea of particle swarm optimization, update the speed and position of particles according to fitness and global optimal solution. Through iterative updates, particles will gradually gather to a better position.
[0049] Step S1-6: Repeat steps S1-4 and S1-5 until the preset number of iterations or convergence condition is reached. Each iteration updates the velocity and position of the particle and calculates a new fitness.
[0050] Step S1-7: Select the best model. After the iteration is completed, the SVR model with the best performance is selected as the final prediction model according to the final fitness value and particle position.
[0051] Wherein, step S2 specifically includes:
[0052] Step S2-1: The present invention dynamically updates the weight coefficient by calculating the Euclidean distance between the particle position and the optimal particle in each stage. The formula of the adaptive weight coefficient is as follows:
[0053]
[0054] Among them, w 0 is the initialization weight coefficient, generally between [0.5, 1], w i The weight coefficient after each stage update, is the position of the ith particle at time t.
[0055] Step S2-2: The sin and cos functions have the characteristics of periodic changes. By applying them to the update of the learning factor, the learning factor can show periodic changes in the iteration process. This periodic change can help the particle swarm to conduct more comprehensive and diverse exploration in the search space, thereby increasing the chance of finding the global optimal solution. The formula for updating the learning factor is as follows:
[0056]
[0057] Where i is the current iteration number, is the cognitive learning factor of the i-th iteration, is the social learning factor of the i-th iteration, and max_iter is the maximum number of iterations. When the learning factor is large, the particle swarm focuses more on development, that is, moving towards the current best solution; when the learning factor is small, the particle swarm focuses more on exploration and searches the solution space in a wider way. This balance helps avoid falling into the local optimal solution while effectively converging to a better solution.
[0058] Wherein, step S3 specifically includes:
[0059] Step S3-1: In order to obtain the optimal solution of the penalty parameter C, kernel parameter σ and boundary error ε in the SVR model, the parameters of the SVR are optimized by the improved ACPSO algorithm in step S2, and the optimal parameters are selected to improve the height prediction performance of the SVR model. The inverse mean square error of the adult height prediction result is used as the fitness function of the ACPSO algorithm, and the calculation formula is as follows:
[0060]
[0061] Where N is the total number of height samples, represents the predicted value of the i-th sample, y i represents the true value of the i-th sample, f MSE Represents the mean square error of the prediction results.
[0062] Step S3-2: Use the ACPSO algorithm to select the optimal parameters of the hyperparameters of the SVR model, repeat step S1 to form an ACPSO-SVR prediction model, and perform height prediction.
[0063] Example 2
[0064] The present embodiment relates to a device for predicting adult height from the juvenile stage based on machine learning, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the method for predicting adult height from the juvenile stage based on machine learning of Example 1.
[0065] Example 3
[0066] The present embodiment relates to a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the method of predicting adult height in the juvenile stage based on machine learning of embodiment 1 is implemented.
[0067] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned implementation modes, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned implementation modes, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the implementation modes of the present invention.
Claims
1. A method for predicting adult height from juvenile stage based on machine learning, Features: The following steps are involved: Step S1: Model integration, combining particle swarm optimization (PSO) with support vector machine regression (SVR) to propose a PSO-SVR model; Step S2: Algorithm optimization, propose adaptive weight coefficient and learning factor update method, improve the PSO algorithm in step S1, and improve its global search and local optimization capabilities; Step S3: Use the improved algorithm in step S2 to predict height.
2. A method for predicting adult height from the juvenile stage based on machine learning according to claim 1, It is characterized in that The specific steps of step S1 are as follows: Step S1-1: Data preparation. First, prepare the dataset for training and testing, and divide the dataset into feature matrix (input) and target vector (output). Make sure the data has been preprocessed and standardized as necessary. Step S1-2: Parameter setting. Set the relevant parameters of PSO and SVR, including the number of particles, number of iterations, inertia weight, acceleration factor, kernel function type, and regularization parameter, etc. The selection of these parameters can be determined by cross-validation or experience. Step S1-3: Initialize the particle swarm. Randomly initialize the position and velocity of each particle in the particle swarm. The position of the particle represents the hyperparameter combination of the SVR model, and the velocity represents the moving direction and amplitude of the particle in the search space. Step S1-4: Calculate fitness. For each particle, construct an SVR model based on its position and calculate the corresponding fitness using the training data. The fitness can be measured using the prediction error (such as mean square error) of the SVR model on the training set. Step S1-5: Update particle speed and position. Using the idea of particle swarm optimization, update the speed and position of particles according to fitness and global optimal solution. Through iterative updates, particles will gradually gather to a better position. Step S1-6: Repeat steps S1-4 and S1-5 until the preset number of iterations or convergence condition is reached. Each iteration updates the velocity and position of the particle and calculates a new fitness. Step S1-7: Select the best model. After the iteration is completed, the SVR model with the best performance is selected as the final prediction model according to the final fitness value and particle position.
3. The method for predicting adult height from the juvenile stage based on machine learning according to claim 1, It is characterized in that The specific steps of step S2 are as follows: Step S2-1: Dynamically update the weight coefficient by calculating the Euclidean distance between the particle position and the optimal particle at each stage. The formula for the adaptive weight coefficient is as follows: Among them, w 0 is the initialization weight coefficient, generally between [0.5, 1], w i The weight coefficient after each stage update, is the position of the ith particle at time t. Step S2-2: The sin and cos functions have the characteristics of periodic changes. By applying them to the update of the learning factor, the learning factor can show periodic changes in the iteration process. This periodic change can help the particle swarm to conduct more comprehensive and diverse exploration in the search space, thereby increasing the chance of finding the global optimal solution. The formula for updating the learning factor is as follows: Where i is the current iteration number, is the cognitive learning factor of the i-th iteration, is the social learning factor of the i-th iteration, and max_iter is the maximum number of iterations. When the learning factor is large, the particle swarm focuses more on development, that is, moving towards the current best solution; when the learning factor is small, the particle swarm focuses more on exploration and searches the solution space in a wider way.
4. The method for predicting adult height from the juvenile stage based on machine learning according to claim 1, It is characterized in that The specific steps of step S3 are as follows: Step S3-1: In order to obtain the optimal solution of the penalty parameter C, kernel parameter σ and boundary error ε in the SVR model, the parameters of the SVR are optimized by the improved ACPSO algorithm in step S2, and the optimal parameters are selected to improve the height prediction performance of the SVR model; the inverse mean square error of the adult height prediction result is used as the fitness function of the ACPSO algorithm, and the calculation formula is as follows: Where N is the total number of height samples, represents the predicted value of the i-th sample, y i represents the true value of the i-th sample, f MSE Represents the mean square error of the prediction results; Step S3-2: Use the ACPSO algorithm to select the optimal parameters of the hyperparameters of the SVR model, repeat step S1 to form an ACPSO-SVR prediction model, and perform height prediction.
5. A device for predicting adult height from juvenile stage based on machine learning, It is characterized in that It comprises a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, they are used to implement the method for predicting adult height in the juvenile stage based on machine learning as described in any one of claims 1-4.
6. A computer-readable storage medium, It is characterized in that A program is stored thereon, and when the program is executed by a processor, the method for predicting adult height in the juvenile stage based on machine learning as described in any one of claims 1-4 is implemented.
Citation Information
Cited By
State evaluation method for vehicle gearbox
CN120046024A
