Athlete physical state prediction method based on machine learning

Abstract

The invention relates to the field of specific calculation models in computer systems, in particular to an athlete physical state prediction method based on machine learning. The method comprises the following steps: acquiring and preprocessing historical training data and physiological state data of an athlete to obtain a physical state vector; based on the physical state vector, a hierarchical structure is introduced, a hierarchical state dependency matrix is constructed, and a hierarchical physical state estimated value is calculated; and based on the layered physical state estimation value, predicting the physical state of the athlete in the future, and obtaining a final physical state prediction result. The method solves the problems that a traditional athlete physical ability prediction method cannot accurately model the multi-level dependency relationship of the physical ability state, the prediction result is easily influenced by noise and error propagation, and the long-term prediction deviation is large; most technologies depend on fixed model parameters, are limited in generalization ability and usually depend on external sensing equipment, and have the problems of data loss and signal interference and influence on prediction reliability.

Description

Technical Field

[0001] The present invention relates to the field of specific computing models in computer systems, and particularly to a method for predicting the physical fitness status of athletes based on machine learning. Background Art

[0002] In the fields of modern competitive sports and sports science, the monitoring and prediction of athletes' physical fitness status are of great significance for optimizing training plans, improving competition performance, and preventing sports injuries. With the popularization of the concept of scientific training, the physical fitness assessment of athletes has gradually shifted from traditional empirical judgment to data-driven analysis methods. The change of physical fitness status is affected by various factors, including training load, recovery time, physiological indicators, environmental conditions, etc. Therefore, accurate prediction of physical fitness status can not only assist coaches in reasonably arranging training, but also help athletes adjust their own rhythm to maintain the best competitive state.

[0003] In recent years, the progress of data acquisition technology has enabled the storage and analysis of athletes' physiological data and training data on different time scales. From short-term heart rate fluctuations to long-term endurance change trends, various types of data provide a basis for physical fitness status modeling and prediction. However, how to extract effective information from complex time-series data and construct a reasonable computational model to achieve high-precision prediction remains an important direction in the cross-research of current sports science and intelligent computing.

[0004] Traditional methods for predicting athletes' physical fitness have the following technical problems: they cannot accurately model the multi-level dependence relationship of physical fitness status, and the prediction results are easily affected by noise and error propagation, resulting in large long-term prediction deviations; most technologies rely on fixed model parameters, making it difficult to optimize for different athletes, with limited generalization ability, and usually relying on external sensing devices, there are problems of data loss and signal interference, affecting the reliability of prediction. Summary of the Invention

[0005] The present invention provides a method for predicting the physical fitness status of athletes based on machine learning to solve the problems that traditional methods for predicting athletes' physical fitness cannot accurately model the multi-level dependence relationship of physical fitness status, the prediction results are easily affected by noise and error propagation, resulting in large long-term prediction deviations; most technologies rely on fixed model parameters, making it difficult to optimize for different athletes, with limited generalization ability, and usually relying on external sensing devices, there are problems of data loss and signal interference, affecting the reliability of prediction.

[0006] A method for predicting the physical fitness status of athletes based on machine learning according to the present invention specifically includes the following technical solutions: A method for predicting the physical fitness status of athletes based on machine learning includes the following steps: S1. After collecting and preprocessing the historical training data and physiological state data of athletes, a physical fitness state vector is obtained; based on the physical fitness state vector, a hierarchical structure is introduced to construct a hierarchical state dependence matrix, and the hierarchical physical fitness state estimation value is calculated; S2. Based on the hierarchical physical fitness state estimation value, a dynamic non-linear state evolution model is constructed to predict the future physical fitness state of athletes, and the final physical fitness state prediction result is obtained.

[0007] Preferably, the S1 specifically includes: The physical fitness state vector is decomposed into multiple layers of intrinsic mode components by the empirical mode decomposition technique, and the entropy value of the intrinsic mode components is calculated; based on the entropy value of the intrinsic mode components, a hierarchical weight factor is defined.

[0008] Preferably, the S1 specifically includes: Based on the hierarchical weight factor, combined with the physical fitness state vector, a hierarchical state dependence matrix is constructed.

[0009] Preferably, the S1 specifically includes: Based on the hierarchical state dependence matrix and the physical fitness state vector, a hierarchical smoothing factor and a local time adjustment term are introduced to evaluate the physical fitness state of athletes at different time scales, and the hierarchical physical fitness state estimation value is calculated; the calculation formula of the hierarchical physical fitness state estimation value is as follows: , where, is the physical fitness state estimation value at the th layer and the th time step; is the time step size; is the adjustment weight for adjusting the influence degree of the physical fitness state vector at the th time step on the physical fitness state vector at the th time step; is in the state dependence matrix of the th layer, the physical fitness state vector at the th time step and the physical fitness state vector at the th time step the state dependence degree between; is the hierarchical smoothing factor of the th layer; is the local time adjustment term at the th layer and the th time step.

[0010] Preferably, the S2 specifically includes: For the dynamic non-linear state evolution model, a state evolution operator is introduced to simulate the conversion law of the physical fitness state over time.

[0011] Preferably, the S2 specifically includes: The calculation formula of the state evolution operator is as follows: , where is the state evolution operator at the th layer and the th time step; is the state evolution rate parameter at the th time step; is the physical fitness state estimation value at the th layer and the th time step; represents the non-linear state conversion term at the th layer and the th time step, is the state conversion function, is the weight factor for each non-linear state conversion term; is the state perturbation coefficient at the th time step; is the physical fitness state estimation value at the th layer and the th time step.

[0012] Preferably, the S2 specifically includes: Based on the state evolution operator, the hierarchical physical fitness state estimation value is dynamically updated to obtain the updated physical fitness state estimation value.

[0013] Preferably, the S2 specifically includes: The updated physical fitness state estimation value is adjusted to obtain the final physical fitness prediction result.

[0014] The beneficial effects of the technical solution of the present invention are: 1. By constructing a hierarchical state dependence matrix and calculating the state dependence degree at different time scales, it ensures that the interaction between short-term physical fitness fluctuations and long-term physical fitness changes is fully modeled, enhancing the ability of the dynamic non-linear state evolution model to capture key physical fitness change trends; by using the empirical mode decomposition technique to extract the intrinsic mode components of the physical fitness state vector and combining the entropy value to calculate the hierarchical weight factor, the main characteristics of the physical fitness state can participate in the prediction more effectively, improving the accuracy of state estimation.

[0015] 2. By introducing a state evolution operator to describe the instantaneous change trend of physical fitness, the adaptability of the dynamic non-linear state evolution model to complex physical fitness states is enhanced. The state perturbation coefficient is used to simulate unpredictable external influences, making the dynamic non-linear state evolution model have better robustness in practical applications. The logarithmic correction term in the state evolution operator ensures the smoothness of the change in physical fitness state, avoids sharp fluctuations in predicted values in a short period of time, and improves the continuity and credibility of prediction results.

[0016] 3. Introduce a hierarchical structure, classify the physical fitness state vectors according to different time scales, so that the dynamic non-linear state evolution model can be optimized for the physical fitness state characteristics of different athletes; by introducing a global adjustment coefficient, improve the personalized adaptation ability of prediction results, so that the dynamic non-linear state evolution model can maintain a high generalization ability among different athletes. Brief Description of the Drawings

[0017] Figure 1 It is a flowchart of a method for predicting an athlete's physical fitness state based on machine learning according to the present invention. Detailed Embodiments

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0020] The following specifically describes the specific solution of a method for predicting an athlete's physical fitness state based on machine learning provided by the present invention with reference to the accompanying drawings.

[0021] Refer to the attached Figure 1 , which shows a flowchart of a method for predicting an athlete's physical fitness state based on machine learning provided by an embodiment of the present invention. The method includes the following steps: S1. After collecting and preprocessing the historical training data and physiological state data of the athlete, a physical fitness state vector is obtained; based on the physical fitness state vector, a hierarchical structure is introduced, a hierarchical state dependence matrix is constructed, and a hierarchical physical fitness state estimate value is calculated; Before predicting the physical fitness state, first collect and preprocess the historical training data and physiological state data of the athlete, so as to construct a hierarchical state dependence matrix and input it into a dynamic non-linear state evolution model for prediction.

[0022] Since the historical training data and physiological state data of the collected athletes will change over time and may contain noise, in order to ensure the stability of the data, it is necessary to preprocess the collected data; the preprocessing process includes: standardizing the collected data to map it to the same scale; and using the mean filtering method to correct outliers and smooth short-term fluctuations to improve the calculation accuracy of the hierarchical state dependence matrix, making the subsequent state evolution calculation more accurate.

[0023] The data set obtained after standardization and outlier correction is defined as , where each represents the physical fitness state vector of the athlete at the th time step, including key indicators affecting the evolution of physical fitness state, such as heart rate, blood oxygen saturation, stride frequency, speed, lactic acid concentration, etc., is the time step size.

[0024] The physical fitness state of the athlete is affected by multiple time scales. For example: short-term effects (seconds, minutes): such as the fluctuation of heart rate and the sharp change of lactic acid level during exercise; medium-term effects (hours, days): such as the recovery process after training and the gradual accumulation of physical fitness consumption; long-term effects (weeks, months): such as the endurance adaptation of the athlete and the long-term accumulation of the training cycle. In order to accurately model the dynamic changes of physical fitness state, a hierarchical structure is introduced, that is, data at different time scales will be divided into different levels, and each level captures the corresponding time scale information.

[0025] Decompose the physical fitness state vector into multiple layers of intrinsic mode components through empirical mode decomposition technology, and calculate the entropy value of each layer of intrinsic mode components. Define the hierarchical weight factor , represents the weight factor of the th layer, is the entropy value of the th layer of intrinsic mode components, is the number of levels, which is set according to the expert experience method; the levels with larger amount of information will be given higher weights to ensure that the main change characteristics of the data are fully utilized. Classify the levels according to the entropy value of the intrinsic mode components and the time scale: the high-frequency level (short-term effect) is used to capture the rapidly fluctuating physical fitness changes, the medium-frequency level (medium-term effect) is used to model the recovery process after exercise, and the low-frequency level (long-term effect) is used to model the long-term effect of the training cycle.

[0026] Due to the fact that the change in the physical fitness state of athletes is affected by complex factors and there is a dependence relationship at different time scales, it is necessary to construct a hierarchical state dependence matrix to describe the state correlation between different time steps; the hierarchical state dependence matrix is used to measure the state similarity between different time steps and is stored and calculated in a computer system to ensure that the dynamic non-linear state evolution model can accurately capture the dynamic evolution characteristics of the physical fitness state. The calculation method of any element in the hierarchical state dependence matrix is as follows: , where, is the degree of state dependence between the physical fitness state vector at the -th time step and the physical fitness state vector at the -th time step in the -th layer of the hierarchical state dependence matrix; is the weight factor of the -th layer, which is used to control the relative importance between different time steps. By calculating the Euclidean distance between the physical fitness state vector at the -th time step and the physical fitness state vector at the -th time step, and performing weight allocation through an exponential decay function, it is ensured that time steps with similar states have a high degree of correlation.

[0027] Based on the hierarchical state dependence matrix, the estimated value of the physical fitness state at each level is calculated, which is used to describe the physical fitness state of athletes at different levels. It is a comprehensive evaluation result of the physical fitness state of athletes at different time scales and can reflect the coupling relationship between short-term fluctuations and long-term trends; the calculation process of the estimated value of the physical fitness state incorporates the influence of the hierarchical state dependence matrix and introduces a local time adjustment term to ensure the smoothness of state changes. The calculation formula for the hierarchical estimated value of the physical fitness state is as follows: , , , where, is the estimated value of the physical fitness state at the -th time step in the -th layer, representing the estimation of the multi-index physical fitness state of the athlete (such as heart rate, lactate concentration, stride frequency, etc.) in the -th layer; is the adjustment weight, which is used to adjust the influence degree of the physical fitness state vector at the -th time step on the physical fitness state vector at the -th time step and is obtained through experiments; is the The hierarchical smoothing factor of the layer, which is used to control the smoothness of the change in physical fitness state; is the local time adjustment term at the time step of the layer, which is used to compensate for the short-term fluctuations of the physical fitness state and make the estimated value of the physical fitness state more stable; is the amplitude of the 𝑚-th sine component, which is used to control the contribution of this sine component to the overall periodic change and is obtained through experiments; is the angular frequency of the 𝑚-th sine component, which is used to describe the time change rate of this sine component and is obtained through experiments; is the phase offset of the 𝑚-th sine component, which is used to adjust the time alignment of the sine function and is obtained through experiments;

[0028] S2. Based on the hierarchical physical fitness state estimation value, construct a dynamic non-linear state evolution model to predict the future physical fitness state of the athlete and obtain the final physical fitness state prediction result.

[0029] Based on the hierarchical physical fitness state estimation value, construct a dynamic non-linear state evolution model to predict the future physical fitness state of the athlete; since the physical fitness state is affected by various non-linear factors, the change process of the physical fitness state is usually not a simple linear change. Therefore, in the dynamic non-linear state evolution model, a state evolution operator is introduced to describe the time evolution characteristics of the physical fitness state and is used to simulate the conversion law of the physical fitness state over time to ensure that the future physical fitness state of the athlete can be accurately predicted. The calculation formula of the state evolution operator is as follows: , where, is the state evolution operator at the time step of the layer, which represents the instantaneous change trend of the physical fitness state; is the state evolution rate parameter at the time step, which is used to control the change amplitude of the current physical fitness state over time and is obtained through experiments; represents the non-linear state conversion term at the time step of the layer, is the state conversion function, and the Bessel function in the existing technology can be selected to describe different non-linear state change modes, is the weight factor for each non-linear state conversion term, which is used to control the relative contribution of different state conversion functions and is obtained through experiments; is the The state perturbation coefficient for each time step is used to control the random perturbation part of the state update to simulate unpredictable physical state changes, such as environmental factors and temporary fatigue of athletes, and is obtained through experiments; is at the layer and the estimated physical state value at the time step; the state perturbation term

[0030] is used to ensure the stability of the state change, prevent the predicted value of the physical state from fluctuating violently, and balance the dynamic change of the physical state. , where is the updated estimated physical state value, that is, the physical state after being calculated by the state evolution operator; is the step factor of the state update, which is used to control the convergence and divergence of the state evolution process and is obtained through experiments; is at the layer and is the fusion weight, which is used to control the contribution of the state evolution operators at different levels and is obtained through experiments; is the regularization coefficient, which ensures that the state change remains stable and is obtained through experiments.

[0031] Based on the updated estimated physical state value, the predicted value is further calculated to obtain the final physical state prediction result; in the embodiments of the present application, the updated estimated physical state value can be adjusted through existing technologies (such as automatic differentiation, regularization, attention mechanism, and adaptive optimization algorithms), or can be adjusted by introducing the following formula, so that the final physical state prediction result not only needs to conform to the change trend of the historical physical state data, but also can effectively reduce the accumulation of long-term prediction errors. The specific calculation formula is: , where is the final physical state prediction result at the time step; represents the global adjustment coefficient, which is used to control the amplitude of the error adjustment; is the gradient information calculated according to the loss function and is used to adaptively adjust the prediction error. The loss function can be constructed by using the mean square error; is a state change smoothing term, which is used to ensure the smoothness of the final physical fitness state prediction result. is a state change smoothing factor, which is used to prevent the prediction value from changing violently between consecutive time steps and make the final physical fitness state prediction result more stable.

[0032] In summary, a method for predicting an athlete's physical fitness state based on machine learning is completed.

[0033] The sequence of the invention embodiments is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0034] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0035] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for predicting the physical fitness status of athletes based on machine learning, characterized in that, It includes the following steps: S1. After collecting and preprocessing the historical training data and physiological state data of the athlete, a physical fitness state vector is obtained; based on the physical fitness state vector, a hierarchical structure is introduced to construct a hierarchical state dependence matrix, and the hierarchical physical fitness state estimation value is calculated; S2. Based on the hierarchical physical fitness state estimation value, a dynamic non-linear state evolution model is constructed to predict the future physical fitness state of the athlete, and the final physical fitness state prediction result is obtained.

2. The method for predicting the physical fitness state of athletes based on machine learning according to claim 1, wherein The S1 specifically includes: The physical fitness state vector is decomposed into multiple layers of intrinsic mode components by the empirical mode decomposition technique, and the entropy value of the intrinsic mode components is calculated; based on the entropy value of the intrinsic mode components, the hierarchical weight factor is defined.

3. The method for predicting the physical fitness state of athletes based on machine learning according to claim 2, wherein The S1 specifically includes: Based on the hierarchical weight factor and combined with the physical fitness state vector, a hierarchical state dependence matrix is constructed.

4. A method for predicting the physical fitness state of athletes based on machine learning according to claim 3, characterized in that, The S1 specifically includes: Based on the hierarchical state dependence matrix and the physical fitness state vector, a hierarchical smoothing factor and a local time adjustment term are introduced to evaluate the physical fitness state of the athlete at different time scales, and the hierarchical physical fitness state estimation value is calculated; the calculation formula of the hierarchical physical fitness state estimation value is as follows: ， Among them, is the estimated physical fitness state value at the th layer and the th time step; is the time step size; is the adjustment weight used to adjust the influence degree of the physical fitness state vector at the th time step on the physical fitness state vector at the th time step; is the state dependence degree between the physical fitness state vector at the th layer, the th time step, and the physical fitness state vector at the th time step and the th time step ; is the layer-level smoothing factor of the th layer; is the local time adjustment term at the th layer and the th time step.​ 5. A method for predicting the physical fitness status of athletes based on machine learning according to claim 1, characterized in that The S2 specifically includes: For the dynamic non-linear state evolution model, a state evolution operator is introduced to simulate the conversion law of the physical fitness state over time.

6. The method for predicting the physical fitness state of athletes based on machine learning according to claim 5, wherein, The S2 specifically includes: The calculation formula of the state evolution operator is as follows: ， Among them, is the state evolution operator at the th layer and the th time step; is the state evolution rate parameter at the th time step; is the physical state estimation value at the th layer and the th time step; represents the non-linear state conversion term at the th layer and the th time step, is the state conversion function, is the weight factor for each non-linear state conversion term; is the state perturbation coefficient at the th time step; is the physical state estimation value at the th layer and the th time step.

7. A method for predicting the physical fitness status of athletes based on machine learning according to claim 6, characterized in that, The S2 specifically includes: Based on the state evolution operator, the hierarchical physical fitness state estimation value is dynamically updated to obtain the updated physical fitness state estimation value.

8. A method for predicting the physical fitness state of athletes based on machine learning according to claim 7, characterized in that, The S2 specifically includes: The updated physical fitness state estimation value is adjusted to obtain the final physical fitness state prediction result.

Images