Student sports performance assessment method based on weighted GG clustering

Abstract

The invention discloses a student sports performance assessment method based on weighted GG clustering. The method comprises the following steps: 1, acquiring all body test item data and corresponding evaluation grades of students of all grades, wherein the body test items comprise body weight index, vital capacity, 50-meter running, sitting-position body forward flexion, standing long jump, pull-up (male) / 1-minute sit-up (female) and 1000-meter running (male) / 800-meter running (female); 2, normalizing the acquired body measurement data of each item to [0, 1]; 3, training a weighted GG clustering model by using data obtained by normalization and corresponding scores, and accurately determining a clustering center of each physical test level; 4, classifying actually measured physical measurement parameters by using the weighted GG clustering model trained in the step 3, and obtaining an accurate physical measurement result; and 5, taking an unreasonable or inaccurate result of physical test score evaluation as a new sample to correct the model so as to improve the precision of the model. The physical test score of the student can be rapidly and accurately judged, and the method has good practical application value.

Description

Technical Field

[0001] The invention relates to the evaluation of students' physical test scores, and in particular to a method for evaluating students' physical test scores based on weighted GG clustering. Background Art

[0002] Cultivating college students' interest in physical exercise, making them consciously participate in sports and move towards lifelong exercise is an important task of physical education in colleges and universities. At present, exercise has become an important measure for many countries to improve physical health. Maintaining a moderate amount of exercise every day is beneficial to the physical health of people of all ages. Exercise can effectively enhance physical fitness and improve psychological quality. At the same time, the muscles of the human body will be more developed and stronger. For a long time, colleges and universities have paid too much attention to the intellectual development factors of college students and ignored their physical condition.

[0003] Physical fitness test is one of the important means to detect students' physical fitness level in physical education in colleges and universities. The physical fitness test for college students mainly evaluates the physical condition and training effect through the test results of various items of students. It is a very sound and effective strategy to urge college students to actively participate in physical training. College physical education teachers can provide students with the most scientific training plan based on the test results, so as to continuously enhance their physical fitness.

[0004] At present, students' physical fitness scores are mostly calculated manually, and each physical fitness item is tested and scored, and finally all the scores are accumulated to obtain the final score. This method takes a long time to calculate, and the quantification of students' physical fitness scores is relatively simple. In a sense, it cannot objectively evaluate students' physical fitness. For physical fitness items, objective evaluation should be a fusion of multiple aspects, rather than a simple linear superposition of various physical fitness items. Summary of the invention

[0005] To solve the above problems, the present invention proposes a student sports performance evaluation method based on weighted GG clustering on the basis of the GG (Gath-Geva, GG) clustering algorithm. Taking into account the differences in the parameter standards of each physical test item, an improved data normalization method is proposed to reduce the convergence difficulty of the model and avoid the overfitting of the model to a certain dimension of data; at the same time, in the cluster center iteration process of GG clustering, the weighted idea is introduced, and the convergence of the model can be accelerated faster by adjusting the weight coefficient; finally, the weighted GG clustering model proposed in this patent supports reverse correction of the model, which further improves the generalization of the model and the accuracy of the evaluation. To achieve this purpose, the present invention provides a student sports performance evaluation method based on weighted GG clustering, the specific steps are as follows, and it is characterized in that:

[0006] Step 1: According to the National Student Physical Fitness Standards, all physical fitness test data and corresponding evaluation levels of students of all grades were collected. The physical fitness test items included body mass index (BMI), vital capacity, 50-meter run, sit-up, standing long jump, pull-ups (male) / 1-minute sit-ups (female), and 1000-meter run (male) / 800-meter run (female);

[0007] Step 2, normalize the collected physical test data of each item to [0,1];

[0008] Step 3, using the normalized data and the corresponding scores to train the weighted GG clustering model, accurately determine the cluster center of each physical test level;

[0009] Step 4, using the weighted GG clustering model trained in step 3 to classify the measured physical test parameters and obtain accurate physical test scores;

[0010] Step 5: For unreasonable or inaccurate results of physical test scores, use them as new samples to modify the model to improve the accuracy of the model.

[0011] Furthermore, the full marks for each physical test item in step 1 are 15, 15, 20, 10, 10, 10 and 20 respectively, with a total score of 100. The evaluation levels are divided into failing, passing, good and excellent, and the corresponding score ranges are [0,59), [60,79), [80,89) and [90,100) respectively.

[0012] Furthermore, in step 2, the algorithm criterion for normalizing the collected physical test data of each item to [0,1] is:

[0013]

[0014] Among them, x represents the collected physical measurement data, x max and x min They represent the reference values ​​corresponding to 100 points and 0 points for each physical test item. In this patent, the physical test value corresponding to 0 points is 0, that is, x min is 0.

[0015] Furthermore, the specific steps of training the weighted GG clustering model in step 3 are:

[0016] Step 3.1. Use the collected training data to train the model, where the data is described as follows: the total number of data sets is P, represented by X p (1≤p≤P), for the set X p For example, it has N elements and can be expressed as: X p The kth element in For example, it has Q-dimensional characteristics, that is, express The qth feature of , in this patent, the values ​​of P and Q are 4 and 7 respectively. In addition, the vector V of P cluster centers can be expressed as: V = {v1, v2, ..., v P}, membership matrix U = [u pk ] P×N , where u pk is the kth element The membership degree of the pth class ranges from u pk ∈[0,1].

[0017] Step 3.2. Use the improved weighted method to update the cluster center. The updating criteria are:

[0018]

[0019] in, represents the cluster center of the p-th class sample determined in the l-th iteration, represents the membership degree u obtained for the l-1th time pk , w k is the weighting coefficient, and the specific expression is:

[0020]

[0021] in Represents the arithmetic mean feature point of each eigenvalue of the pth set.

[0022] Step 3.3. Calculate the elements To the corresponding cluster center v p The maximum likelihood distance

[0023]

[0024] Among them A p represents the covariance matrix of the p-th cluster, α p represents the prior probability of cluster p.

[0025] Step 3.4. Update of membership matrix:

[0026]

[0027] m is a weighted index, which is selected as 2 in this patent.

[0028] Step 3.5. Repeat the above steps 3.2 to 3.4, and iteratively update U and V to minimize the objective function J(U,V), where the expression of J(U,V) is:

[0029]

[0030] The condition for the iteration to terminate is: the absolute value of the membership degree of the lth iteration and the membership degree corresponding to the l-1th iteration is less than or equal to the termination tolerance ε, that is, |U (l) -U (l-1) |≤ε, in this patent, ε is selected as 1e-2.

[0031] Furthermore, the specific steps for correcting the trained model in step 5 are the same as step 3. The number of samples for a single training is recommended to be no less than twice the number of categories, which is more than 8 in this patent.

[0032] The present invention provides a method for evaluating student sports performance based on weighted GG clustering, and the beneficial effects are as follows:

[0033] 1. This patent takes into account the differences in parameter standards of various physical test items and proposes an improved data normalization method to reduce the difficulty of model convergence and avoid overfitting of the model to data of a certain dimension;

[0034] 2. This patent introduces the weighted idea in the cluster center iteration process of the GG clustering model. By adjusting the weight coefficient, the convergence of the model can be accelerated more quickly;

[0035] 3. The weighted GG clustering model proposed in this patent supports reverse correction of the model, which further improves the generalization of the model and the accuracy of evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of the present invention;

[0037] Figure 2 This is a training flowchart of the weighted GG clustering model described in this patent. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0039] The present invention proposes a student sports performance evaluation method based on weighted GG clustering, aiming to achieve accurate evaluation of student sports performance. Figure 1 This is a flow chart of the present invention. The method described in the patent is introduced in detail below in conjunction with the flow chart.

[0040] Step 1: According to the National Student Physical Fitness Standards, all physical fitness test data and corresponding evaluation levels of students of all grades were collected. The physical fitness test items included body mass index (BMI), vital capacity, 50-meter run, sit-up, standing long jump, pull-ups (male) / 1-minute sit-ups (female), and 1000-meter run (male) / 800-meter run (female);

[0041] Specifically, the full marks for each physical test item in step 1 are 15, 15, 20, 10, 10, 10 and 20 respectively, with a total score of 100. The evaluation levels are divided into failing, passing, good and excellent, and the corresponding score ranges are [0,59), [60,79), [80,89) and [90,100) respectively.

[0042] Step 2, normalize the collected physical test data of each item to [0,1];

[0043] Specifically, the algorithm criterion for normalizing the collected physical test data of each item to [0,1] in step 2 is:

[0044]

[0045] Among them, x represents the collected physical measurement data, x max and x min They represent the reference values ​​corresponding to 100 points and 0 points for each physical test item. In this patent, the physical test value corresponding to 0 points is 0, that is, x min is 0.

[0046] Step 3, using the normalized data and the corresponding scores to train the weighted GG clustering model, accurately determine the cluster center of each physical test level;

[0047] Step 4, using the weighted GG clustering model trained in step 3 to classify the measured physical test parameters and obtain accurate physical test scores;

[0048] Step 5: For unreasonable or inaccurate results of physical test scores, use them as new samples to modify the model to improve the accuracy of the model.

[0049] Specifically, the specific steps for correcting the trained model in step 5 are the same as step 3. The number of samples for a single training is recommended to be no less than twice the number of categories, which is more than 8 in this patent.

[0050] Figure 2 This is the training flow chart of the weighted GG clustering model described in this patent. As can be seen from the figure, the training mainly consists of two parts: data preprocessing and iterative training. The iterative training is to update the cluster center through the objective function and the membership matrix. When it is judged that the convergence condition is met, the iteration of the model is terminated. If it is not met, the iteration continues until the model converges. The specific steps are as follows:

[0051] Step 1. Use the collected training data to train the model, where the data is described as follows: The total number of data sets is P, represented by X p (1≤p≤P), for the set X p For example, it has N elements and can be expressed as: Xp The kth element in For example, it has Q-dimensional characteristics, that is, express The qth feature of , in this patent, the values ​​of P and Q are 4 and 7 respectively. In addition, the vector V of P cluster centers can be expressed as: V = {v1, v2, ..., v P}, membership matrix U = [u pk ] P×N , where u pk is the kth element The membership degree of the pth class ranges from u pk ∈[0,1].

[0052] Step 2. Use the improved weighted method to update the cluster center. The updating criteria are:

[0053]

[0054] in, represents the cluster center of the p-th class sample determined in the l-th iteration, represents the membership degree u obtained for the l-1th time pk , w k is the weighting coefficient, and the specific expression is:

[0055]

[0056] in Represents the arithmetic mean feature point of each eigenvalue of the pth set.

[0057] Step 3. Calculate the elements To the corresponding cluster center v p The maximum likelihood distance

[0058]

[0059] Among them A p represents the covariance matrix of the p-th cluster, α p represents the prior probability of cluster p.

[0060] Step 4. Update of membership matrix:

[0061]

[0062] m is a weighted index, which is selected as 2 in this patent.

[0063] Step 5. Repeat steps 3 to 4 above, and iteratively update U and V to minimize the objective function J(U, V), where the expression of J(U, V) is:

[0064]

[0065] The condition for the iteration to terminate is: the absolute value of the membership degree of the lth iteration and the membership degree corresponding to the l-1th iteration is less than or equal to the termination tolerance ε, that is, |U (l) -U (l-1) |≤ε, in this patent, ε is selected as 1e-2.

[0066] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A method for evaluating student sports performance based on weighted GG clustering, the specific steps are as follows, characterized in that: Step 1: Collect all physical test data and corresponding evaluation levels of students in each grade, including body mass index (BMI), vital capacity, 50-meter run, sit-up, standing long jump, pull-ups for men / 1-minute sit-ups for women, and 1000-meter run for men / 800-meter run for women; Step 2, normalize the collected physical test data of each item to [0,1]; Step 3, using the normalized data and the corresponding scores to train the weighted GG clustering model, accurately determine the cluster center of each physical test level; Step 4, using the weighted GG clustering model trained in step 3 to classify the measured physical test parameters and obtain accurate physical test scores; Step 5: For unreasonable or inaccurate results of physical test scores, use them as new samples to modify the model to improve the accuracy of the model.

2. A method for evaluating student sports performance based on weighted GG clustering according to claim 1, characterized in that: The full marks for each physical test item in step 1 are 15, 15, 20, 10, 10, 10 and 20 respectively, with a total score of 100. The evaluation levels are divided into failing, passing, good and excellent, and the corresponding score ranges are [0,59), [60,79), [80,89) and [90,100) respectively.

3. A method for evaluating student sports performance based on weighted GG clustering according to claim 1, characterized in that: The algorithm criteria for normalizing the collected physical test data of each item to [0,1] in step 2 is: Among them, x represents the collected physical measurement data, x max and x min They represent the reference values ​​corresponding to 100 points and 0 points for each physical test item. The physical test value corresponding to 0 points is 0, that is, x min is 0.

4. A method for evaluating student sports performance based on weighted GG clustering according to claim 1, characterized in that: The specific steps of training the weighted GG clustering model in step 3 are: Step 3.

1. Use the collected training data to train the model, where the data is described as follows: the total number of data sets is P, represented by X p (1≤p≤P), for X p For example, it has N elements and can be expressed as: X p The kth element in For example, it has Q-dimensional characteristics, that is, express The qth feature of , the values ​​of P and Q in this patent are 4 and 7 respectively. In addition, the vector V of P cluster centers can be expressed as: V={v1,v2,...,v P }, membership matrix U = [u pk ] P×N , where u pk is the kth element The membership degree of the pth class ranges from u pk ∈[0,1]. Step 3.

2. Use the improved weighted method to update the cluster center. The updating criteria are: in, represents the cluster center of the p-th class sample determined in the l-th iteration, represents the membership degree u obtained for the l-1th time pk , w k is the weighting coefficient, and the specific expression is: in The arithmetic mean feature point of each eigenvalue of the pth set; Step 3.

3. Calculate the elements To the corresponding cluster center v p The maximum likelihood distance Among them A p represents the covariance matrix of the p-th cluster, α p represents the prior probability of cluster p; Step 3.

4. Update of membership matrix: m is the weighting index, which is selected as 2. Step 3.

5. Repeat the above steps 3.2 to 3.4, and iteratively update U and V to minimize the objective function J(U,V), where the expression of J(U,V) is: The condition for the iteration to terminate is: the absolute value of the membership degree of the lth iteration and the membership degree corresponding to the l-1th iteration is less than or equal to the termination tolerance ε, that is, |U (l) -U (l-1) |≤ε, in this patent, ε is selected as 1e-2.

5. A method for evaluating student sports performance based on weighted GG clustering according to claim 1, characterized in that: The specific steps for correcting the trained model in step 5 are the same as those in step 3. The number of samples for a single training is recommended to be no less than twice the number of categories.

Images