An artificial intelligence-based athletic injury care assistance system

Abstract

The application discloses a kind of based on artificial intelligence's sports injury nursing auxiliary system, system includes: data acquisition module, data preprocessing module, constructs sports injury recovery state evaluation model module, model parameter search module and sports injury nursing auxiliary module.The application belongs to the field of data processing, specifically refers to a kind of based on artificial intelligence's sports injury nursing auxiliary system, this scheme introduces feature comprehensive weight division feature subset, is selected to complete the construction of decision tree with best split feature, designs the weighted voting strategy based on Kappa coefficient, and comprehensively considers Kappa coefficient, cross-entropy loss and vote count design loss function;Random chaos initialization strategy is introduced to initialize individual position, design population convergence degree based on similarity, judge population local optimum situation, design two different position updating methods, improve the performance and accuracy of model, provide reliable basis for the formulation of rehabilitation scheme.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically referring to an artificial intelligence-based sports injury care assistance system. Background Technology

[0002] Sports injury care assistance systems utilize artificial intelligence algorithms to analyze and process patient sports injury data, understand the patient's recovery status, and provide assistance in developing care plans. However, existing sports injury care assistance systems suffer from problems such as complex patient sports injury data, covering multiple dimensions, and imbalanced injury data; they also suffer from significant individual patient differences, leading to models that cannot adapt to various sports injuries and have poor generalization ability. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an artificial intelligence-based sports injury nursing assistance system. Addressing the problems of complex patient sports injury data, encompassing multiple dimensions, and data imbalance in existing systems, this solution introduces a feature comprehensive weighting system. This system considers the relationship between a data point and its nearest neighbor, the importance differences of features across different data categories, and the imbalance in data quantity. Based on the feature comprehensive weighting, a feature subset is partitioned, and the optimal splitting feature is selected to construct the decision tree, enhancing model adaptability. A weighted voting strategy based on the Kappa coefficient is designed to assess the sports injury recovery status, comprehensively considering the Kappa coefficient and cross-entropy loss. By incorporating a voting-count loss function, the risk of overfitting is reduced, improving the accuracy and reliability of the sports injury recovery status assessment model. Addressing the issue that existing sports injury nursing assistance systems suffer from poor generalization ability due to significant individual patient variability, this solution introduces a random chaotic initialization strategy. This increases the diversity of the initial population, improving search speed and accuracy. A similarity-based population convergence mechanism is designed to promptly identify and address local optima. Two different position update methods are employed, flexibly selected based on the population state, enhancing the model's adaptability and generalization ability to different situations. This allows for more efficient finding of suitable sports injury recovery status assessment model parameters, improving model performance and accuracy.

[0004] This invention provides an artificial intelligence-based sports injury nursing assistance system, comprising a data acquisition module, a data preprocessing module, a sports injury recovery status assessment model construction module, a model parameter search module, and a sports injury nursing assistance module;

[0005] The data acquisition module collects historical patient sports injury data;

[0006] The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data.

[0007] The module for constructing a sports injury recovery status assessment model introduces feature comprehensive weighting to divide feature subsets, performs optimal splitting feature selection to complete the construction of the decision tree, designs a weighted voting strategy based on the Kappa coefficient, and comprehensively considers the Kappa coefficient, cross-entropy loss, and vote count to design a loss function;

[0008] The model parameter search module introduces a random chaotic initialization strategy to initialize individual positions, designs a similarity-based population convergence, judges the local optimum of the population, designs two different position update methods, and completes the search for the optimal model parameters.

[0009] The sports injury nursing assistance module collects real-time patient sports injury data, understands the patient's recovery status based on the model's output labels, and formulates nursing plans.

[0010] Furthermore, the data acquisition module collects historical patient sports injury data; this data includes basic patient information, physiological data, specific details of the sports injury, recovery process data, and recovery status; physiological data includes heart rate, blood pressure, body temperature, muscle activity, blood oxygen saturation, respiratory rate, and joint range of motion; specific details of the sports injury include the exact time of the injury, type of injury, severity, location of injury, and post-injury symptoms; recovery process data includes daily exercise duration, specific type of exercise, exercise intensity, pain level, specific content and completion of rehabilitation training, degree of activity limitation, and sleep quality and duration; and recovery status is used as a data label.

[0011] Furthermore, the data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data; data cleaning includes handling missing values, outliers, and duplicate values; data transformation converts the data into vector form; data normalization unifies the data range based on the min-max normalization method; and dataset construction builds an evaluation dataset based on the processed data.

[0012] Furthermore, the module for constructing a sports injury recovery status assessment model specifically includes the following:

[0013] To construct F decision trees, the Bootstrap method is used to sample the evaluation dataset F times, obtaining F training subsets. A decision tree is then constructed based on each training subset, for a total of F decision trees. Constructing a decision tree based on a training subset includes the following:

[0014] To calculate the comprehensive weight of features, for each sample in the training subset, randomly select the k nearest neighbors that belong to the same data label as the sample and the k nearest neighbors that do not belong to the same data label, and construct a nearest neighbor dataset B of the same class and a nearest neighbor dataset C of different classes, respectively. Then, calculate the comprehensive weight of each feature based on B and C, using the following formula:

[0015] ;

[0016] ;

[0017] In the formula, It is data E u And data E v Regarding feature Z m distance, It is the u-th data E in the training subset. u The v-th data C(E) in the dataset C of different classes and its nearest neighbors v Regarding feature Z m distance, It is the u-th data E in the training subset. u The v-th data B(E) in its nearest neighbor dataset B. v Regarding feature Z m distance, and They are data E u And data E v The m-th feature Z m ,ω(Z) m ) is feature Z m The overall weight, G is the total number of data in the training subset, y is the data label index, and class (E) is the weight of the training subset. u ) is data E u The data labels, H(y) and These are the members of y and class (E) in the training subset, respectively. u The ratio of the number of data points to the total number of data points G;

[0018] Construct feature subsets, remove features with a comprehensive weight of 0 from the training subset, sort the remaining features in descending order of comprehensive weight, and divide the data into high-weight feature subsets, medium-weight feature subsets, and low-weight feature subsets on an average basis;

[0019] To construct a decision tree, the training subset is input into the root node, and the tree is built starting from the root node. If the dataset in the current node is not empty and the data in the dataset belongs to multiple classes, then the current node is an internal node, and splitting is performed based on the best splitting feature. Otherwise, the current node is a leaf node that will not split further, and other nodes are selected for further splitting. If all remaining nodes are leaf nodes, the decision tree is considered complete. Splitting based on the best splitting feature includes the following:

[0020] Construct a candidate feature set. The dataset in the current node contains L data points, which are randomly selected from the high-weight feature subset, the medium-weight feature subset, and the low-weight feature subset. Based on these features, construct a candidate feature set for the current node. It is a floor function;

[0021] Split: Calculate the Gini coefficient of each feature in the candidate feature set, select the feature with the smallest Gini coefficient as the best splitting feature for the current node, and split the current node into two child nodes based on the best splitting feature.

[0022] Weighted voting classification is used, based on an assessment model of sports injury recovery status composed of F decision trees. A weighted voting strategy is designed based on the Kappa coefficient, and the data label with the most votes is selected as the model evaluation result. The formula used is as follows:

[0023] ;

[0024] ;

[0025] In the formula, Kappa f Y is the Kappa coefficient of the f-th decision tree, Y is the total number of data labels, and M is the total number of data labels. fy O is the number of data points correctly classified by the f-th decision tree for the data label y. fy and R fy Let P(X) represent the number of real data belonging to data label y in the f-th decision tree and the number of data classified as data label y by the decision tree. Let s be the evaluation result of data X in the sports injury recovery status assessment model. f (X) is the classification result of data X in the f-th decision tree, and U is the set of data labels; This means that when the classification result of the decision tree is a data label in the data label set, the vote for that data label is incremented by 1; otherwise, the vote for that data label remains unchanged.

[0026] The loss function is designed by comprehensively considering the Kappa coefficient, cross-entropy loss, and vote count. The formula used is as follows:

[0027] ;

[0028] In the formula, Loss is the loss function, Q is the number of data points in the evaluation dataset, and α and β are the first and second balancing factors, respectively. and These are the cross-entropy loss of the nth data point in the f-th decision tree and the sum of all vote counts, respectively.

[0029] Furthermore, the model parameter search module specifically includes the following:

[0030] Initialize individual positions, and normalize the parameters of the sports injury recovery status assessment model, including the number of samplings F, the number of nearest neighbor data k, the first balance factor α, and the second balance factor β. Establish a search space based on the normalized model parameters, using individual positions as representatives of the sports injury recovery status assessment model parameters. Use the loss function of the sports injury recovery status assessment model established based on the model parameters as the fitness value of the individual. Design a stochastic chaotic initialization strategy to initialize N individual positions within the search space. These N individuals constitute a population. The formula used is as follows:

[0031] ;

[0032] ;

[0033] In the formula, h i and h i+1 Let be the i-th and (i+1)-th random chaotic values, respectively. `rand(0,1)` generates a random value within the range (0,1). `UP` and `LP` are the upper and lower bounds of the search space, respectively. i It is the initial position of the i-th individual;

[0034] Calculate the population convergence, which includes the following:

[0035] Construct an N×2D matrix A based on the current positions of all individuals and the globally optimal position during the iteration process. Normalize A so that all elements are within the interval [0, 1] to obtain the normalized matrix. Matrix A is represented as follows:

[0036] ;

[0037] In the formula, A(t) is an N×2D matrix at the t-th iteration, t is the current iteration number, and D is the individual dimension. , and These represent the positions of the first individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the positions of the second individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the positions of the Nth individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the optimal positions of the first individual since iteration t in the 1st, 2nd, and Dth dimensions, respectively. , and These represent the optimal positions of the second individual since iteration t, in the 1st, 2nd, and Dth dimensions, respectively. , and These are the optimal positions of the Nth individual since iteration t in the 1st, 2nd, and Dth dimensions, respectively. The optimal position is the position of the individual with the smallest fitness value.

[0038] Calculate the normalized matrix The similarity between any two rows is calculated using the following formula:

[0039] ;

[0040] In the formula, x p，q (t) is the normalized matrix at the t-th iteration. The similarity between row p and row q, where t is the current iteration number. and They are The values ​​in the p-th row and q-th row of the d-th column;

[0041] The formula used to calculate population convergence based on similarity is as follows:

[0042] ;

[0043] In the formula, S(t) is the population convergence degree at the t-th iteration;

[0044] For position updating, two different position updating methods are designed, with a pre-set convergence threshold and a maximum number of iterations T. If the population convergence S(t) is greater than the convergence threshold and the current iteration number is much less than the maximum number of iterations, it indicates that the population is trapped in a local optimum, and the first position updating method based on an adaptive lens imaging learning strategy is used for population position updating; otherwise, the second position updating method is used for population position updating. The two different position updating methods include the following:

[0045] The first position update method uses the following formula:

[0046] ;

[0047] ;

[0048] In the formula, It is w i，j (t) The position after updating using the first update method, w i，j (t) represents the position of the i-th individual in the j-th dimension at the t-th iteration, where a j (t) and b j z(t) represents the upper and lower bounds of the positions of all individuals in the j-th dimension at the t-th iteration, and z(t) is the adaptive adjustment factor at the t-th iteration.

[0049] The second position update method uses the following formula:

[0050] ;

[0051] ;

[0052] In the formula, It is w i，j (t) The position after updating using the second update method, where r is a random value, o and c are the first control factor randomly selected from [0, 2] and the second control factor randomly selected from (0°, 180°), respectively, and w best，j (t) is the optimal position of the population in the j-th dimension at the t-th iteration;

[0053] Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the optimal position of the population are updated. When the fitness value corresponding to the optimal position of the population is lower than the fitness threshold, the corresponding model parameters are the optimal model parameters. A sports injury recovery status assessment model is constructed based on the optimal model parameters. Otherwise, if the maximum number of iterations T is reached, the individual position is re-initialized. Otherwise, the number of iterations is incremented by 1 and the population convergence is recalculated for position updates.

[0054] Furthermore, the sports injury nursing assistance module collects real-time patient sports injury data. The real-time patient sports injury data includes basic patient information, physiological data, specific sports injury data, and recovery process data. After preprocessing the real-time patient sports injury data, it is input into a sports injury recovery status assessment model constructed based on optimal model parameters. Based on the model's output labels, the patient's recovery status is understood, and a nursing plan is formulated.

[0055] The beneficial effects achieved by the present invention using the above solution are as follows:

[0056] (1) In view of the problems of complex patient sports injury data, multiple dimensions and imbalance of injury data in the existing sports injury nursing assistance system, this scheme introduces feature comprehensive weight. Feature comprehensive weight considers the relationship between data and its nearest neighbor data points, the difference in importance of features between different categories of data and the imbalance of data quantity, accurately highlighting key features. Based on feature comprehensive weight, feature subsets are divided, and the best split feature selection is performed to complete the construction of decision tree, which better handles the complexity and diversity of data and enhances the adaptability of model. A weighted voting strategy based on Kappa coefficient is designed to complete the assessment of sports injury recovery status. The loss function is designed by comprehensively considering Kappa coefficient, cross-entropy loss and vote count, reducing the risk of overfitting and improving the accuracy and reliability of sports injury recovery status assessment model, providing a more reliable basis for the formulation of rehabilitation plan.

[0057] (2) In view of the problem that the existing sports injury nursing assistance system has large individual differences among patients, which makes the model unable to adapt to various sports injuries and has poor generalization ability, this solution introduces a random chaotic initialization strategy to increase the diversity of the initial population, improve the speed and accuracy of the search, and designs a population convergence based on similarity in order to discover and deal with local optima in a timely manner. Two different position update methods are adopted and flexibly selected according to the population state, which improves the adaptability and effectiveness of the algorithm, enhances the model's adaptability and generalization ability to different situations, and can find suitable sports injury recovery state assessment model parameters more efficiently, thereby improving the model's performance and accuracy. Attached Figure Description

[0058] Figure 1 A schematic diagram of an artificial intelligence-based sports injury care assistance system provided by the present invention;

[0059] Figure 2 A schematic diagram of a module for constructing a sports injury recovery status assessment model;

[0060] Figure 3 This is a schematic diagram of the model parameter search module.

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0062] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0063] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0064] Example 1, see Figure 1 The present invention provides an artificial intelligence-based sports injury nursing assistance system, including a data acquisition module, a data preprocessing module, a sports injury recovery status assessment model construction module, a model parameter search module, and a sports injury nursing assistance module;

[0065] The data acquisition module collects historical patient sports injury data and sends the data to the data preprocessing module;

[0066] The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data, and sends the data to the module for constructing a sports injury recovery status assessment model.

[0067] The module for constructing a sports injury recovery status assessment model introduces feature comprehensive weighting to divide feature subsets, performs optimal splitting feature selection to complete the construction of a decision tree, designs a weighted voting strategy based on the Kappa coefficient, and comprehensively considers the Kappa coefficient, cross-entropy loss, and vote count to design a loss function, and sends the data to the model parameter search module.

[0068] The model parameter search module introduces a random chaotic initialization strategy to initialize individual positions, designs a population convergence based on similarity, judges the local optimum of the population, designs two different position update methods, completes the search for the best model parameters, and sends the data to the sports injury care auxiliary module.

[0069] The sports injury nursing assistance module collects real-time patient sports injury data, understands the patient's recovery status based on the model's output labels, and formulates nursing plans.

[0070] Example 2, see Figure 1This embodiment is based on the above embodiment. The data acquisition module collects historical patient sports injury data. The historical patient sports injury data includes basic patient information, physiological data, specific details of the sports injury, recovery process data, and recovery status. The physiological data includes heart rate, blood pressure, body temperature, muscle activity, blood oxygen saturation, respiratory rate, and joint range of motion data. The specific details of the sports injury include the exact time of the sports injury, the type of sports injury, its severity, the location of the injury, and the symptoms after the injury. The recovery process data includes daily exercise duration, specific type of exercise, exercise intensity, pain level, specific content and completion of rehabilitation training, degree of activity limitation, and sleep quality and duration data. The recovery status is used as a data label.

[0071] Example 3, see Figure 1 This embodiment is based on the above embodiment. The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data. Data cleaning includes handling missing values, outliers, and duplicate values. Data transformation converts the data into vector form. Data normalization unifies the data range based on the min-max normalization method. Dataset construction builds an evaluation dataset based on the processed data.

[0072] Example 4, see Figure 1 and Figure 2 This embodiment, based on the above embodiments, specifically includes the following components in constructing the sports injury recovery status assessment model module:

[0073] To construct F decision trees, the Bootstrap method is used to sample the evaluation dataset F times, obtaining F training subsets. A decision tree is then constructed based on each training subset, for a total of F decision trees. Constructing a decision tree based on a training subset includes the following:

[0074] To calculate the feature weights, for each sample in the training subset, randomly select the k nearest neighbors that belong to the same data label as the sample and the k nearest neighbors that do not belong to the same data label, constructing a nearest neighbor dataset B of the same class and a nearest neighbor dataset C of different classes, respectively. Then, calculate the weight of each feature based on B and C. The feature weights take into account the relationship between the data and its nearest neighbors, the difference in importance of the feature between different classes of data, and the imbalance of the number of data. The formula used is as follows:

[0075] ;

[0076] ;

[0077] In the formula, It is data E u And data E v Regarding feature Zm distance, It is the u-th data E in the training subset. u The v-th data C(E) in the dataset C of different classes and its nearest neighbors v Regarding feature Z m distance, It is the u-th data E in the training subset. u The v-th data B(E) in its nearest neighbor dataset B. v Regarding feature Z m distance, and They are data E u And data E v The m-th feature Z m ,ω(Z) m ) is feature Z m The overall weight, G is the total number of data in the training subset, y is the data label index, and class (E) is the weight of the training subset. u ) is data E u The data labels, H(y) and These are the members of y and class (E) in the training subset, respectively. u The ratio of the number of data points to the total number of data points G;

[0078] Construct feature subsets, remove features with a comprehensive weight of 0 from the training subset, sort the remaining features in descending order of comprehensive weight, and divide the data into high-weight feature subsets, medium-weight feature subsets, and low-weight feature subsets on an average basis;

[0079] To construct a decision tree, the training subset is input into the root node, and the tree is built starting from the root node. If the dataset in the current node is not empty and the data in the dataset belongs to multiple classes, then the current node is an internal node, and splitting is performed based on the best splitting feature. Otherwise, the current node is a leaf node that will not split further, and other nodes are selected for further splitting. If all remaining nodes are leaf nodes, the decision tree is considered complete. Splitting based on the best splitting feature includes the following:

[0080] Construct a candidate feature set. The dataset in the current node contains L data points, which are randomly selected from the high-weight feature subset, the medium-weight feature subset, and the low-weight feature subset. Based on these features, construct a candidate feature set for the current node. It is a floor function;

[0081] Split: Calculate the Gini coefficient of each feature in the candidate feature set, select the feature with the smallest Gini coefficient as the best splitting feature for the current node, and split the current node into two child nodes based on the best splitting feature.

[0082] Weighted voting classification is used, based on an assessment model of sports injury recovery status composed of F decision trees. A weighted voting strategy is designed based on the Kappa coefficient, and the data label with the most votes is selected as the model evaluation result. The formula used is as follows:

[0083] ;

[0084] ;

[0085] In the formula, Kappa f Y is the Kappa coefficient of the f-th decision tree, Y is the total number of data labels, and M is the total number of data labels. fy O is the number of data points correctly classified by the f-th decision tree for the data label y. fy and R fy Let P(X) represent the number of real data belonging to data label y in the f-th decision tree and the number of data classified as data label y by the decision tree. Let s be the evaluation result of data X in the sports injury recovery status assessment model. f (X) is the classification result of data X in the f-th decision tree, and U is the set of data labels; This means that when the classification result of the decision tree is a data label in the data label set, the vote for that data label is incremented by 1; otherwise, the vote for that data label remains unchanged.

[0086] The loss function is designed by comprehensively considering the Kappa coefficient, cross-entropy loss, and vote count. The formula used is as follows:

[0087] ;

[0088] In the formula, Loss is the loss function, Q is the number of data points in the evaluation dataset, and α and β are the first and second balancing factors, respectively. and These are the cross-entropy loss of the nth data point in the f-th decision tree and the sum of all vote counts, respectively.

[0089] By performing the above operations, this solution addresses the problems of complex patient sports injury data, multiple dimensions, and data imbalance in existing sports injury care assistance systems. It introduces a feature comprehensive weighting system, which considers the relationship between data and its nearest neighbor, the importance differences of features across different data categories, and the imbalance in data quantity. This accurately highlights key features. Based on the feature comprehensive weighting, feature subsets are divided, and the optimal splitting feature is selected to construct the decision tree. This better handles the complexity and diversity of the data, enhances model adaptability, and designs a weighted voting strategy based on the Kappa coefficient to assess the sports injury recovery status. Furthermore, it comprehensively considers the Kappa coefficient, cross-entropy loss, and vote count to design a loss function, reducing the risk of overfitting and improving the accuracy and reliability of the sports injury recovery status assessment model. This provides a more reliable basis for the development of rehabilitation plans.

[0090] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the model parameter search module specifically includes the following:

[0091] Initialize individual positions, and normalize the parameters of the sports injury recovery status assessment model, including the number of samplings F, the number of nearest neighbor data k, the first balance factor α, and the second balance factor β. Establish a search space based on the normalized model parameters, using individual positions as representatives of the sports injury recovery status assessment model parameters. Use the loss function of the sports injury recovery status assessment model established based on the model parameters as the fitness value of the individual. Design a stochastic chaotic initialization strategy to initialize N individual positions within the search space. These N individuals constitute a population. The formula used is as follows:

[0092] ;

[0093] ;

[0094] In the formula, h i and h i+1 Let be the i-th and (i+1)-th random chaotic values, respectively. `rand(0,1)` generates a random value within the range (0,1). `UP` and `LP` are the upper and lower bounds of the search space, respectively. i It is the initial position of the i-th individual;

[0095] Calculating population convergence, which represents the similarity between individuals in a population, results in higher population convergence. The calculation of population convergence includes the following:

[0096] Construct an N×2D matrix A based on the current positions of all individuals and the globally optimal position during the iteration process. Normalize A so that all elements are within the interval [0, 1] to obtain the normalized matrix. Matrix A is represented as follows:

[0097] ;

[0098] In the formula, A(t) is an N×2D matrix at the t-th iteration, t is the current iteration number, and D is the individual dimension. , and These represent the positions of the first individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the positions of the second individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the positions of the Nth individual in the 1st, 2nd, and Dth dimensions, respectively, during the t-th iteration. , and These represent the optimal positions of the first individual since iteration t in the 1st, 2nd, and Dth dimensions, respectively. , and These represent the optimal positions of the second individual since iteration t, in the 1st, 2nd, and Dth dimensions, respectively. , and These are the optimal positions of the Nth individual since iteration t in the 1st, 2nd, and Dth dimensions, respectively. The optimal position is the position of the individual with the smallest fitness value.

[0099] Calculate the normalized matrix The similarity between any two rows is calculated using the following formula:

[0100] ;

[0101] In the formula, x p，q (t) is the normalized matrix at the t-th iteration. The similarity between row p and row q, where t is the current iteration number. and They are The values ​​in the p-th row and q-th row of the d-th column;

[0102] The formula used to calculate population convergence based on similarity is as follows:

[0103] ;

[0104] In the formula, S(t) is the population convergence degree at the t-th iteration;

[0105] For position updating, two different position updating methods are designed, with a pre-set convergence threshold and a maximum number of iterations T. If the population convergence S(t) is greater than the convergence threshold and the current iteration number is much less than the maximum number of iterations, it indicates that the population is trapped in a local optimum, and the first position updating method based on an adaptive lens imaging learning strategy is used for population position updating; otherwise, the second position updating method is used for population position updating. The two different position updating methods include the following:

[0106] The first position update method uses the following formula:

[0107] ;

[0108] ;

[0109] In the formula, It is w i，j (t) The position after updating using the first update method, w i，j (t) represents the position of the i-th individual in the j-th dimension at the t-th iteration, where a j (t) and b j z(t) represents the upper and lower bounds of the positions of all individuals in the j-th dimension at the t-th iteration, and z(t) is the adaptive adjustment factor at the t-th iteration.

[0110] The second position update method uses the following formula:

[0111] ;

[0112] ;

[0113] In the formula, It is w i，j (t) The position after updating using the second update method, where r is a random value, o and c are the first control factor randomly selected from [0, 2] and the second control factor randomly selected from (0°, 180°), respectively, and w best，j (t) is the optimal position of the population in the j-th dimension at the t-th iteration;

[0114] Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the optimal position of the population are updated. When the fitness value corresponding to the optimal position of the population is lower than the fitness threshold, the corresponding model parameters are the optimal model parameters. A sports injury recovery status assessment model is constructed based on the optimal model parameters. Otherwise, if the maximum number of iterations T is reached, the individual position is re-initialized. Otherwise, the number of iterations is incremented by 1 and the population convergence is recalculated for position updates.

[0115] By performing the above operations, this solution addresses the problem that existing sports injury nursing assistance systems suffer from significant individual patient differences, leading to the model's inability to adapt to various sports injuries and poor generalization ability. It introduces a random chaotic initialization strategy to increase the diversity of the initial population, improving search speed and accuracy. A similarity-based population convergence is designed to promptly identify and address local optima. Two different position update methods are employed, flexibly selected based on the population state, improving the algorithm's adaptability and effectiveness. This enhances the model's adaptability and generalization ability to different situations, enabling it to more efficiently find suitable parameters for assessing sports injury recovery status, thereby improving model performance and accuracy.

[0116] Example 6, see Figure 1 This embodiment is based on the above embodiment. The sports injury nursing assistance module collects real-time patient sports injury data. The real-time patient sports injury data includes basic patient information data, physiological data, specific sports injury data, and recovery process data. After preprocessing the real-time patient sports injury data, it is input into the sports injury recovery status assessment model constructed based on the optimal model parameters. Based on the output labels of the model, the patient's recovery status is understood and a nursing plan is formulated.

[0117] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0118] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0119] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based sports injury care assistance system, characterized in that: It includes a data acquisition module, a data preprocessing module, a module for constructing a sports injury recovery status assessment model, a model parameter search module, and a sports injury nursing assistance module; The data acquisition module collects historical patient sports injury data; The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data. The module for constructing a sports injury recovery status assessment model introduces feature comprehensive weighting to divide feature subsets, performs optimal splitting feature selection to complete the construction of the decision tree, designs a weighted voting strategy based on the Kappa coefficient, and comprehensively considers the Kappa coefficient, cross-entropy loss, and vote count to design a loss function; The model parameter search module introduces a random chaotic initialization strategy to initialize individual positions, designs a similarity-based population convergence, judges the local optimum of the population, designs two different position update methods, and completes the search for the optimal model parameters. The sports injury nursing assistance module collects real-time patient sports injury data, understands the patient's recovery status based on the model's output labels, and formulates nursing plans.

2. The artificial intelligence-based sports injury care assistance system according to claim 1, characterized in that: The module for constructing a sports injury recovery status assessment model specifically includes the following: To construct F decision trees, the Bootstrap method is used to sample the evaluation dataset F times to obtain F training subsets. A decision tree is constructed based on each training subset, for a total of F decision trees. Weighted voting classification is used, based on an assessment model of sports injury recovery status composed of F decision trees. A weighted voting strategy is designed based on the Kappa coefficient, and the data label with the most votes is selected as the model evaluation result. The formula used is as follows: ； ； In the formula, Kappa f Y is the Kappa coefficient of the f-th decision tree, Y is the total number of data labels, and M is the total number of data labels. fy O is the number of data points correctly classified by the f-th decision tree for the data label y. fy and R fy Let P(X) represent the number of real data belonging to data label y in the f-th decision tree and the number of data classified as data label y by the decision tree. Let s be the evaluation result of data X in the sports injury recovery status assessment model. f (X) is the classification result of data X in the f-th decision tree, and U is the set of data labels; This means that when the classification result of the decision tree is a data label in the data label set, the vote for that data label is incremented by 1; otherwise, the vote for that data label remains unchanged. The loss function is designed by comprehensively considering the Kappa coefficient, cross-entropy loss, and vote count. The formula used is as follows: ； In the formula, Loss is the loss function, Q is the number of data points in the evaluation dataset, and α and β are the first and second balancing factors, respectively. and These are the cross-entropy loss of the nth data point in the f-th decision tree and the sum of all vote counts, respectively.

3. The artificial intelligence-based sports injury care assistance system according to claim 2, characterized in that: The construction of F decision trees based on a training subset specifically includes the following: To calculate the comprehensive weight of features, for each sample in the training subset, randomly select the k nearest neighbors that belong to the same data label as the sample and the k nearest neighbors that do not belong to the same data label, and construct a nearest neighbor dataset B of the same class and a nearest neighbor dataset C of different classes, respectively. Then, calculate the comprehensive weight of each feature based on B and C, using the following formula: ； ； In the formula, It is data E u And data E v Regarding feature Z m distance, It is the u-th data E in the training subset. u The v-th data C(E) in the dataset C of different classes and its nearest neighbors v Regarding feature Z m distance, It is the u-th data E in the training subset. u The v-th data B(E) in its nearest neighbor dataset B. v Regarding feature Z m distance, and They are data E u And data E v The m-th feature Z m ,ω(Z) m ) is feature Z m The overall weight, G is the total number of data in the training subset, y is the data label index, and class (E) is the weight of the training subset. u ) is data E u The data labels, H(y) and These are the members of y and class (E) in the training subset, respectively. u The ratio of the number of data points to the total number of data points G; Construct feature subsets, remove features with a comprehensive weight of 0 from the training subset, sort the remaining features in descending order of comprehensive weight, and divide the data into high-weight feature subsets, medium-weight feature subsets, and low-weight feature subsets on an average basis; To construct a decision tree, input the training subset into the root node of the decision tree and start building the decision tree from the root node. If the dataset in the current node is not empty and the data in the dataset belongs to multiple classes, it means that the current node is an internal node, and then splitting is performed based on the best splitting feature. Otherwise, it means the current node is a leaf node that will not split further, and other nodes are selected for further splitting. If all remaining nodes are leaf nodes, the decision tree is complete. Splitting based on the best splitting feature includes the following: Construct a candidate feature set. The dataset in the current node contains L data points, which are randomly selected from the high-weight feature subset, the medium-weight feature subset, and the low-weight feature subset. Based on these features, construct a candidate feature set for the current node. It is a floor function; Split: Calculate the Gini coefficient of each feature in the candidate feature set, select the feature with the smallest Gini coefficient as the best splitting feature for the current node, and split the current node into two child nodes based on the best splitting feature.

4. The artificial intelligence-based sports injury care assistance system according to claim 1, characterized in that: The model parameter search module specifically includes the following: Initialize individual positions, and normalize the parameters of the sports injury recovery status assessment model, including the number of samplings F, the number of nearest neighbor data k, the first balance factor α, and the second balance factor β. Establish a search space based on the normalized model parameters, using individual positions as representatives of the sports injury recovery status assessment model parameters. Use the loss function of the sports injury recovery status assessment model established based on the model parameters as the fitness value of the individual. Design a stochastic chaotic initialization strategy to initialize N individual positions within the search space. These N individuals constitute a population. The formula used is as follows: ； ； In the formula, h i and h i+1 Let be the i-th and (i+1)-th random chaotic values, respectively. `rand(0,1)` generates a random value within the range (0,1). `UP` and `LP` are the upper and lower bounds of the search space, respectively. i It is the initial position of the i-th individual; Calculate the population convergence, which includes the following: Construct an N×2D matrix A based on the current positions of all individuals and the globally optimal position during the iteration process. Normalize A so that all elements are within the interval [0, 1] to obtain the normalized matrix. D is the individual dimension; Calculate the normalized matrix The similarity between any two rows is calculated using the following formula: ； In the formula, x p，q (t) is the normalized matrix at the t-th iteration. The similarity between row p and row q, where t is the current iteration number. and They are The values ​​in the p-th row and q-th row of the d-th column; The formula used to calculate population convergence based on similarity is as follows: ； In the formula, S(t) is the population convergence degree at the t-th iteration; For position update, two different position update methods are designed. A convergence threshold and a maximum number of iterations T are preset. If the population convergence S(t) is greater than the convergence threshold and the current number of iterations is much less than the maximum number of iterations, it indicates that the population is trapped in a local optimum. The first position update method based on the adaptive lens imaging learning strategy is used to update the population position. Otherwise, the second position update method is used to update the population position. Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the optimal position of the population are updated. When the fitness value corresponding to the optimal position of the population is lower than the fitness threshold, the corresponding model parameters are the optimal model parameters. A sports injury recovery status assessment model is constructed based on the optimal model parameters. Otherwise, if the maximum number of iterations T is reached, the individual position is re-initialized. Otherwise, the number of iterations is incremented by 1 and the population convergence is recalculated for position updates.

5. The artificial intelligence-based sports injury care assistance system according to claim 4, characterized in that: The location update involves two different location update methods, specifically including the following: The first position update method uses the following formula: ； ； In the formula, It is w i，j (t) The position after updating using the first update method, w i，j (t) represents the position of the i-th individual in the j-th dimension at the t-th iteration, where a j (t) and b j z(t) represents the upper and lower bounds of the positions of all individuals in the j-th dimension at the t-th iteration, and z(t) is the adaptive adjustment factor at the t-th iteration. The second position update method uses the following formula: ； ； In the formula, It is w i，j (t) The position after updating using the second update method, where r is a random value, o and c are the first control factor randomly selected from [0, 2] and the second control factor randomly selected from (0°, 180°), respectively, and w best，j (t) is the optimal position of the population in the j-th dimension at the t-th iteration.

6. The artificial intelligence-based sports injury care assistance system according to claim 1, characterized in that: The sports injury nursing assistance module collects real-time patient sports injury data. The real-time patient sports injury data includes basic patient information, physiological data, specific sports injury data, and recovery process data. After preprocessing the real-time patient sports injury data, it is input into a sports injury recovery status assessment model constructed based on optimal model parameters. Based on the model's output labels, the patient's recovery status is understood, and a nursing plan is formulated.

7. The artificial intelligence-based sports injury care assistance system according to claim 1, characterized in that: The data acquisition module collects historical patient sports injury data; the historical patient sports injury data includes basic patient information, physiological data, specific details of the sports injury, recovery process data, and recovery status; the recovery status is used as a data label.

8. The artificial intelligence-based sports injury care assistance system according to claim 1, characterized in that: The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data; data cleaning includes handling missing values, outliers, and duplicate values. Data transformation converts data into vector form; data normalization unifies the data range based on the min-max normalization method; and dataset construction builds an evaluation dataset based on the processed data.

Images