Personalized Training Path Design for Civil Aviation Flight Cadet Physical Education Course Based on Genetic Algorithm

Civil aviation flight trainees should have all-round talents, i.e. aviation knowledge, flight ability, adaptability, perseverance, etc. In order to cultivate flight trainees with the physical quality of excellent pilots, it is very necessary to strengthen the quality training of flight trainees [1-2]. It can not only improve the flight ability of the trainees, but also make the trainees resistant to bumps, incapacitation, hypoxia, fatigue, and the ability to correctly make judgments and dispositions under the complex and changeable conditions of the flight environment under the conditions of high-altitude airtight operating environment [3-5]. Promote the trainees gradually become an excellent pilot with commitment and ability, and genetic algorithm can play an important role in this training process [6].

Civil aviation flight cadet sports course training is a very meaningful flight ability training, which is not only the normal quality training for flight cadets, but also for the characteristics of the flight cadets and flight needs, and then develop a personalized quality training to help improve the flight ability of the flight cadets [7-10]. Such as sensitive coordination quality training, quick reaction quality training, balance and orientation sensory ability training, attention distribution ability training, high altitude tolerance ability training, aviation sports specialized equipment training and so on. It can be said that the significance of aviation sports quality training for the effective improvement of flight ability, the improvement of aviation sports quality can increase the improvement of flight ability of flight cadets [11-14]. Aviation sports quality training for the cadets means to cultivate the cadets’ flying ability, and to promote the improvement of the flying ability of the cadets. At the same time, the formation and enhancement of flight ability can effectively promote the improvement of the quality of flight technology training, and reduce the possibility of flight trainees being grounded in the process of flight, and indirectly reduce the cost of flight training [15-18].

In this paper, a sports scheduling method based on chaotic real number genetic algorithm is designed. Firstly, according to the flight trainee sports data with constraints and MEARA algorithm, the personalized training path of sports courses is transformed into the solution of optimization problem. Then for the problem of binary coding leading to the inability of cross-mutation operation to cross, considering that real number coding does not have any encoding and decoding process, real number coding is used to replace the traditional binary coding in order to accurately reach the lowest value that the computer can allow. The designed algorithm is used for personalized path recommendation for a new semester of physical education course in a flight school to verify the training effect of the algorithm.

2

Key algorithms for personalized training generation in physical education courses

2.1

Flight Cadet Modeling

2.1.1

Flight Cadet Modeling and Information Collection Methods

This paper analyzes the characteristics of flight trainees and organizes and refines them, which are exactly the characteristics needed for flight trainee modeling. They include basic information (gender, age, height, weight, etc.), physical information (body size, physique, cardiorespiratory index, etc.), exercise goals (weight loss, muscle building, shape building, enhanced sensitivity, enhanced flexibility, etc.), exercise ability (the basic five qualities), exercise preferences (running, playing ball, jumping rope, etc.), and exercise conditions (venues, equipment, etc.).

2.1.2

Flight participant information processing

1)

The body mass index (BMI), referred to as the body mass index, is commonly used to measure the degree of fatness and thinness of the human body, and is calculated as shown in formula (1). (1) $B M I = \frac{w e i g h t}{h e i g h t^{2}}$

Where weight represents weight (unit kg) and height represents height (unit m). 2)

Body Fat Percentage (BFP), also called body fat percentage, indicates the proportion of fat content in the human body. The body fat percentage of adults is calculated as shown in formula (2). (2) $B F P = \frac{0.74 \cdot w a i s t - 0.082 \cdot w e i g h t + g e n d e r}{w e i g h t}$

Among them, waist represents waist circumference (unit cm), weight represents weight (unit kg), and gender represents different values for different genders. 3)

Cardiorespiratory capacity assessment, the cardiorespiratory capacity assessment done in this paper adopts the step test method, and the assessment index E is calculated as shown in formula (3). (3) $E = \frac{100 \times t i m e}{2 \times (r_{1} + r_{2} + r_{3})}$

Where, time represents the duration of the step climbing exercise in seconds, and r₁, r₂, and r₃ represent the pulse rate during the 3 recovery periods.

2.2

Physical Education Program Selection

Based on the attribute information of the exercise prescription and the core four parameters (exercise effect, exercise intensity, exercise time, and exercise frequency), this paper summarizes some characteristics of the exercise courses in the analysis of the exercise prescription, and carries out the preliminary selection of the exercise courses based on these characteristics as well as the basic information of the flight participants. At the same time, since the courses in the course library do not have the specific value of course intensity, it is also necessary to quantify the intensity of the courses. The selection of exercise courses is shown in Figure 1.

2.2.1

Methods for quantifying the intensity of physical education programs

Exercise intensity in the existing exercise program is described by subjective exertion sensation and does not have specific quantitative values, in order to ensure the consistency between the attributes of the exercise program and the attributes of the exercise prescription, it is necessary to quantify the intensity of the exercise program. The parameters needed for quantification are the calorie consumption attribute and the exercise duration attribute of the exercise program, and their quantification calculation is shown in Equation (4). (4) $Q_{c l a s s} = \frac{C_{c l a s s} \cdot K + T_{c l a s s} \cdot β}{T_{c l a s s} \cdot γ}$

Where, C_class is the calorie consumption value of the exercise session, T_class is the exercise duration of the exercise session, Q_class is the quantified intensity of the session, K represents the conversion factor, β represents the normal range of the exercise health index, and γ represents the deviation rate.

2.2.2

Physical education course selection program

The exercise courses summarized in this paper have the following characteristics in the analysis in conjunction with exercise prescription: they contain functional attributes (exercise effect, object of action) and conditional attributes (calorie consumption, intensity of exercise, exercise time, frequency of exercise, exercise equipment, exercise venue, contraindicated diseases, contraindicated people).

2.3

Optimization of Physical Education Course Portfolio Parameters

This section focuses on the specific parameters that need to be optimized during the generation of the sports course portfolio and the determination of the objective function during the optimization process.

2.3.1

Motion Load Calculation and Constraint Determination

1)

Exercise load

Exercise load can be transformed into an integral of exercise intensity over exercise time, the exercise load for a day is expressed in the form of equation (5), then the total exercise load for a cycle is expressed in equation (6). (5) $W_{d a y} = \int_{0}^{T} Q d t$ (6) $W_{w e e k} = \sum_{0}^{P} \int_{0}^{T} Q d t$

Where, Q is the motion intensity, T is the motion time and P is the motion frequency. 2)

Constraints

In the case of a certain amount of exercise in the exercise prescription, in order to ensure that the values of the exercise course combination parameters optimized by MEARA are closer to the values of the attributes of the exercise prescription, this paper needs to impose constraints on the range of values of exercise intensity, exercise time, and exercise frequency.

Exercise intensity constraints: the exercise intensity of each exercise program needs to be within a certain range, and the value of the total exercise program intensity must not be greater than (n + 1) times of the intensity value Q of the current flight trainee’s exercise prescription, and must not be less than the intensity value Q of the exercise prescription, and n is the number of additional exercise programs. The exercise intensity constraint is shown in Equation (7). (7) ${\begin{matrix} q_{m_\min} < q_{m} < q_{m_\max} \\ q_{i_\min} < q_{i} < q_{i_\max} \\ Q \leq \sum_{i = 1}^{n} q_{i} + q_{m} \leq (n + 1) Q \end{matrix}$

Exercise time constraint: for the exercise time only the exercise time range of the total exercise course needs to be controlled, and the constraint formula for the exercise time is Equation (8). Where the sum of the exercise time t_m of the essential exercise course and the exercise time t_i of the n additional exercise courses must not be greater than the time T of the current exercise prescription. (8) $0 < \sum_{i = 1}^{n} t_{i} + t_{m} \leq T$

Motion frequency constraints: for motion frequency p the constraints are that the motion frequency within a week is greater than or equal to 1 and must not be greater than 7, so the motion frequency range is $[1, 7]$ .

2.3.2

Determination of the objective function

Based on the exercise intensity, exercise time and exercise frequency in the flight trainee’s exercise prescription, the cycle exercise quantity is calculated, and from the perspective of exercise prescription application-oriented, the cycle exercise quantity of the exercise course combination generated by the final combination of exercise prescription needs to be maximally close to the cycle exercise quantity of this exercise prescription, and only in this way can we guarantee the effectiveness of the application of this exercise prescription. In the case of a certain amount of periodical exercise, the three quantities of exercise intensity, exercise time and exercise frequency have mutual constraints, which is a multi-objective optimization problem.

In this paper, the combination of an exercise program is set up as an essential exercise program plus an additional exercise program. In other words, it is necessary to first determine a necessary exercise program that has the same effect as an exercise prescription, and then add an additional exercise program that has an additional effect. Accordingly, the expected amount of exercise desired attainment function established with exercise intensity, exercise time, and exercise frequency is f_ea(x), the exercise intensity desired attainment function established with exercise intensity, the maximum expected difference of exercise intensity and the minimum expected difference of exercise intensity is f_ci(x), and the exercise time desired attainment function established with exercise time, the maximum expected difference of exercise time and the minimum expected difference of exercise time is f_et(x), where f_ei(x) and f_et(x) are constraints on each other. The three objective functions are constructed as shown in Equation (9) on the following page. (9) ${\begin{matrix} f_{e a} (x) = 1 - (\frac{1}{α * | W - p * (\int_{0}^{t_{n}} q_{m} d t + \sum_{i = 1}^{n} \int_{0}^{t_{i}} q_{i} d t) | + 1}) \\ f_{e i} (x) = 1 - {(\frac{\sum_{i = 1}^{n} q_{i} + q_{m} - q_{\min}}{q_{\max} - q_{\min}})}^{α} \\ f_{e i} (x) = 1 - {(\frac{\sum_{i = 1}^{n} t_{i} + t_{m} - t_{\min}}{t_{\max} - t_{\min}})}^{α} \end{matrix}$

The simulation of the effect of different α-values on the difference of the motion is carried out according to Eq. (10), and when the sensitivity coefficient is smaller, the more sensitive to the proximity of the motion is, the higher the accuracy the parameter needs to achieve [19]. (10) $D (d) = 1 - {(\frac{d - d_{\min}}{d_{\max} - d_{\min}})}^{α}$

Where, D is the proximity, d is the difference of the motions, d_min is the minimum expected difference of the motions, d_max is the maximum expected difference of the motions, and α is the sensitivity coefficient.

In summary, the multi-objective optimization model can be obtained as shown in Eq. (11). (11) $\begin{matrix} \min F (x) = (f_{e a} (x), f_{e i} (x), f_{e t} (x)) \\ s . t . q_{m_\min} < q_{m} < q_{m_\max} \\ q_{i_\min} < q_{i} < q_{i_\max} \\ Q \leq \sum_{i = 1}^{n} q_{i} + q_{m} \leq (n + 1) Q \\ 0 < \sum_{i = 1}^{n} t_{i} + t_{m} \leq T \\ 0 < p \leq 7 \end{matrix}$

Once the objective function as well as the constraints have been determined, parameter optimization experiments are conducted using specific flight cadet case data in conjunction with MEARA to optimize the parameters of the exercise course combinations that are more in line with the flight cadets themselves.

3

Training path generation based on chaotic real number genetic algorithm

3.1

Basic elements of genetic algorithms

3.1.1

Chromosome coding and decoding

1)

Binary encoding

If the value range of a parameter is set to $[U_{1}, U_{2}]$ and the length of the binary code is 1, the corresponding relationship at the time of encoding is shown below: (12) $\begin{array}{l} 000000 \dots 0000 = 0 \to U_{1} \\ 000000 \dots 0001 = 1 \to U_{1} + δ \\ 000000 \dots 0010 = 2 \to U_{1} + 2 δ \\ \dots \\ 111111 \dots 1111 = 2^{l} - 1 \to U_{2} \end{array}$

where $δ = \frac{U_{2} - U_{1}}{2^{I} - 1}$ .

Decoding: assuming that an individual is coded as a_ka_k−1a_k−2…..a₂a₁, the corresponding decoding equation is: (13) $X = U_{1} + (\sum_{i = 1}^{k} a_{i} 2^{i - 1}) \times \frac{U_{2} - U_{1}}{2^{k} - 1}$

2)

Real encoding

A variable for a problem can be directly transformed from the set of all solutions to the search space by using a real number encoding [20]. Its chromosome is shaped like Eq: (14) $X = (\begin{array}{l} x_{1}, & x_{2}, x_{3}, \dots, x_{n} \end{array}), x_{i} \in R, i = 1, 2, \dots, n$

3)

Gray code encoding

Compared to the code corresponding to two uninterrupted integers between which only one coding point is different, all other coding points are the same.

The conversion formula for binary code to Gray code is: (15) ${\begin{array}{l} x_{i} = y_{i} \\ x_{j} = y_{j} \oplus y_{j + 1}, j = i - 1, i - 2 \dots, 2, 1 \end{array}$

The conversion formula for converting Gray code to binary is: (16) ${\begin{array}{l} x_{i} = y_{i} \\ x_{j} = x_{j} \oplus y_{j + 1}, j = i - 1, i - 2 \dots, 2, 1 \end{array}$

From Eqs. (15) and (16), it can be seen that the binary code and Gray code can be transformed to each other. For the same original code, there is a difference between representation in binary and representation in Gray code.

3.1.2

Adaptation function design

The fitness function can reflect the quality of individuals in the population, which is the main basis for selecting the parent individuals. In the process of problem solving, the fitness function is firstly established according to the requirements of the problem, and then it is evaluated and the possibility of selecting a certain individual for the next operation is derived. Since the value of the objective function can be positive or negative, the relationship between it and the fitness function is not unique. Therefore, in order to maximize the objective function, the direction of the objective function should be the same as the direction of the change in the adaptation value.

3.1.3

Genetic operational design

Genetic algorithms are more or less the same as genetic operations of biological genes, which are randomly selected in a certain way. The task of genetic operation is to realize the evolutionary process of survival of the fittest and elimination of the unfit by evaluating the fitness function and then performing specific operations. It generally includes operations such as selection, crossover, and mutation.

3.1.4

Genetic algorithm parameterization

1)

Termination conditions, generally taken as 100 to 500;

2)

Population size, generally taken as 20~100;

3)

Crossover probability, generally taken as 0.4~0.99;

4)

Probability of variation, generally taken as 0.001~0.1.

3.2

Steps of the genetic algorithm

Firstly, the initial population is randomly generated, secondly, the individual fitness of the population is calculated F, then the good individuals X and Y are selected according to F, then X and Y are subjected to crossover and mutation operations, and finally, it is judged whether it is the same as the termination condition, if it is the same, then it is ended, otherwise, it continues to execute the second step. The specific realization process diagram is shown in Figure 2.

3.3

Application of Chaotic Genetic Algorithm to Scheduling Problems

The improved genetic algorithm is used in the study of scheduling problem, firstly, the sports course information is arranged in real number coding way to form a chromosome, followed by the initial population, then the individual fitness F is determined and the excellent parent individual is selected, in the chaotic crossover and chaotic mutation operation, and finally, in the judgment of the end conditions [21]. The specific realization process is shown in Figure 3.

3.3.1

Coding

According to the lesson plan, the teacher, the course taught by the teacher and the class to which he/she belongs are determined and are treated as an element, denoted by A. Placing A in one of the cells of the two-dimensional table, this creates our PE schedule for the week. The courses include the course name, the gender of the class (male, female, mixed), and the classrooms include the campus and specific classroom number. The value definitions can be represented by the following equation. (17) $X_{i} Y_{j} = {\begin{array}{l} X Y, X_{i} Classroom Y_{j} Classes available at time \\ 0, X_{i} Classroom Y_{j} Time for no class \end{array}$

3.3.2

Population initialization

In this paper, we incorporate the chaotic algorithm in the global search process to increase the computational speed, reduce the computational time and improve the quality of the initialized population. The initialized population is the n points generated by chaotic mapping denoted as: (18) $X_{n + 1} = μ X_{n} (1 - X_{n}), n = 1, 2, 3, \dots, n$

Where: μ represents the control parameter.

3.3.3

Selection of the adaptation function

Because the fitness function not only has a direct connection to the convergence speed of the algorithm, but also has a direct impact on whether the best solution can be found in the end. Therefore we can select the fitness function from these two aspects.

According to the objective constraint analysis described above, this paper adopts the following fitness function: (19) $g (x) = \sum_{i = 1}^{3} w_{i} \times f_{i} \times \frac{θ}{3}$

3.3.4

Selection

This paper uses the roulette selection method. A brief description of this method is given below.

Let the population size be N and the fitness of individual i be g(i). The probability that individual i is selected is P_i: (20) $P_{i} = g (i) / \sum_{i = 1}^{N} g (i), i = 1, 2, \dots, N$

From equation (20), the probability of being selected is proportional to P_i. The larger it is, the more likely it is to be selected; conversely, the probability of a selected individual is almost none, and may even be eliminated. By calculating the probability of being selected through Eq. (20), it is then possible to decide which individuals to select as the initial individuals of the parent generation.

3.3.5

Chaotic crossover

Crossover operation is one of the most critical operations of genetic algorithms and is the main way of generating new individuals, which involves replacing some of the genes of two selected parent individuals in some way to form a new individual. 1)

Determine the initial conditions. That is, assume a crossover probability p_c, so that its value range is 0.4 ≤ p_c ≤ 0.99.

2)

Determine the chaotic sequence x_n. i.e., randomly select a number x₀ as the initial value, generate a chaotic sequence x_n according to Eq. (21), and if x_n is smaller than the set value of p_c, map the sequence x_n into a two-dimensional matrix, otherwise, continue to select a random number x₀ to generate a chaotic sequence and then make a judgment.

3)

Determine which row to perform the crossover operation. That is, when x_n ∈ [0, 1], the row labeled for the crossover operation is: (21) $i = [D \times x_{n}]$

Where: D denotes a class period, there are five weekdays in a week and five class periods in a weekday, i.e., the value of D is 25. $[*]$ denotes rounding.

3.3.6

Chaotic mutations

The mutation operation is the interchanging of a particular element of a two-dimensional matrix with another particular element to form a new individual. The specific operation process is as follows: 1)

Determine the mutation probability p_n.

Let its value range be 0.001 ≤ p_n ≤ 0.1.

2)

Determine the chaotic sequence.

To perform the mutation operation it is necessary to know the specific element location of the mutation, i.e., the row and column coordinates are known. Therefore, random numbers x₀, x₁, x₂, and x₃ are chosen as initial values to generate four chaotic sequences x_n, x_1n, x_2n, and x_3n. If x_n, x_1n, x_2n, and x_3n are smaller than the set value of p_n, the sequences x_n and x_1n are mapped to a two-dimensional matrix to find the row coordinates of the mutation, and the sequences x_2n and x_3n are mapped to a two-dimensional matrix to find the column coordinates of the mutation; otherwise, continue to choose the random numbers x₀, x₁, and x₂, x₃ to generate a chaotic sequence and then make a judgment.

3)

Determine the specific location where the mutation is performed, i.e., including the row labeling and column labeling.

When x_n ∈ [0, 1], the first row labeled: (22) $i_{1} = [D \times x_{n}]$

When x_2n ∈ [0, 1], the second line is labeled: (23) $i_{2} = [D \times x_{2 n}]$

When x_3n ∈ [0, 1], the first column is labeled: (24) $c_{1} = [R \times x_{3 n}]$

When x_4n ∈ [0, 1], the second column is labeled: (25) $c_{2} = [R \times x_{4 n}]$

Where: D denotes the number of class periods, there are five weekdays in a week and five class periods in a weekday, i.e., the value of D is 25. R denotes the total number of classrooms. $[*]$ denotes rounding. 4)

Perform the mutation operation. Comparing two elements to be exchanged, if all of the two elements are 0 or one element is 1, no operation is performed, then the mutation operation is not performed; on the contrary, if one element is not 0, the two elements are swapped in position, which, probabilistically speaking, achieves the purpose of mutation.

3.3.7

Judgment

Set the number of iterations, if the number of iterations is reached, it ends and outputs the final solution as the final obtained solution; otherwise, the individual fitness calculation loop continues to be executed until the number of iterations is reached.

4

Simulation experiments and practical applications

4.1

Historical data simulation experiment

In order to verify that the algorithm in this paper is prioritized over the genetic algorithm in the scheduling problem. This experiment uses Matlab R2010a to program this paper’s algorithm and the traditional GA respectively in the paper, running on a PC with 4.6GHz CPU and 16GB RAM. The scheduling task of the Graduate School of Mathematics and Statistics for the first semester of the academic year 2022-2023 was chosen as the test example, where the population size was 100. The metrics examined for the performance of the algorithm include the average running time, the number of times a better solution occurs.

4.1.1

Experimental results

Based on two algorithms for college scheduling solutions (where the method of determining the fitness function is the same for both algorithms is made), the number of iterations is set to 50,100,150,200,250,270,290 in order, and each of them is performed 50 times for testing, running and recording the results. The average number of better solutions is shown in Fig. 4 and the average time of running is shown in Fig. 5.

The effectiveness of the algorithm of this paper is considered in two directions. Firstly, according to the number of better solutions, the number of better solutions of this algorithm and the traditional GA tends to stabilize with the increase of the number of iterations, but when the number of iterations is 100, the number of better solutions of this algorithm is obviously two times more than the number of better solutions of the traditional GA. Finally, the number of better solutions of this algorithm is stabilized at about 11, while that of GA is 5.

Secondly, in terms of convergence speed, the experiment shows that when the number of better solutions of traditional GA stabilizes, it takes 115.8 seconds, while the algorithm of this paper stabilizes in only 28.6 seconds, which shows that the convergence speed of the hybrid genetic algorithm is three times faster than that of the genetic algorithm.

4.1.2

Analysis of experimental results

The above results show that the algorithm adopted in this paper is characterized by fast convergence speed and easy convergence to the global optimal solution. This comes from: first, the algorithm in this paper adopts floating-point coding which is better than the traditional GA; second, the initial population of this paper’s algorithm is of better quality than that of the traditional GA, and the proportion of feasible chromosomes is higher, which effectively avoids more unsatisfactory chromosomes produced at the initial stage of the algorithm; third, the chaotic optimization operation of this paper’s algorithm effectively accelerates the convergence of the progeny individuals.

4.2

Analysis of practical applications

In order to test the effectiveness of this paper’s algorithm for flight cadets’ physical education course path recommendation, the record data of 700 students and 79 physical education course instructors of a flight school were chosen to personalize the path recommendation of physical education courses for the upcoming new semester. Table 1 shows the standard table of some course parameters, Table 2 shows the difficulty parameter evaluation table, and Table 3 shows the course evaluation level table.

Table 1.

Part of Course parameter standard table

Course ID	Rs_i	Rd_i	w_i	Pr
9104716	0.6	0.7	0.9	0.22
9098606	0.7	0.6	0.4	0.651
9076607	1	0.7	0.2	0.614
9124687	0.7	0.7	0.2	0.159
9121201	0.6	0.3	0.6	0.738
9113545	1	1	0.5	0.716
9103936	0.6	0.4	0.8	0.81
9062993	0.9	0.6	0.7	0.991
9129910	0.9	0.6	0.5	0.999
9145046	0.8	0.7	0.7	0.728
9082595	0.9	0.8	0.5	0.157
9061804	0.9	0.8	0.3	0.389
9087248	0.2	0.6	0.7	0.243
9152541	0.2	0.8	0.9	0.448
9097639	0.4	0.3	0.3	0.167
9115983	0.4	0.4	0.3	0.189
9111235	0.9	0.4	0.7	0.865
9116459	0.9	0.3	0.5	0.249
9101055	0.7	0.2	0.3	0.177
9100591	0.8	0.8	0.5	0.591
9120519	0.4	0.4	0.6	0.327
9119303	0.6	0.3	0.8	0.273
9085241	0.8	0.3	0.3	0.542
9158493	0.9	0.4	0.9	0.724
9098077	0.9	0.6	0.4	0.407
9118030	0.5	0.2	0.7	0.244
9061517	0.7	0.8	0.8	0.357
9157155	0.9	1	0.4	0.397
9093878	0.7	0.7	0.9	0.597
9061248	0.9	0.6	0.7	0.446

Table 2.

Part of Difficulty parameter evaluation

Course ID	Ty_n	Te_n	Sy_n	Se_n	v	C_max	C_min	P_e	C_e
9156794	164	78	519	90	0.7	0.86	0.26	0.479	0.251
9155019	95	89	491	84	0.4	0.59	0.83	0.647	0.456
9152365	174	90	267	87	0.7	0.72	0.76	0.418	0.589
9081898	112	81	500	92	0.7	0.48	0.24	0.268	0.365
9079052	219	80	394	88	0.2	0.27	0.7	0.407	0.24
9144924	113	79	366	84	0.7	0.25	0.52	0.719	0.768
9149461	229	78	472	93	0.6	0.48	0.88	0.532	0.868
9140811	200	80	503	90	0.9	0.32	0.82	0.345	0.223
9121933	257	87	426	76	0.3	0.36	0.67	0.579	0.31
9122508	213	90	395	95	0.7	0.29	0.61	0.844	0.583
9106360	226	82	522	76	0.2	0.67	0.41	0.404	0.802
9077001	214	89	487	81	0.9	0.55	0.82	0.746	0.862
9142288	200	92	449	90	0.8	0.89	0.87	0.208	0.467
9084531	96	91	512	88	0.6	0.72	0.46	0.22	0.279
9095009	206	90	597	92	0.8	0.28	0.49	0.603	0.806

Table 3.

Grade of course evaluation table

Course ID	Ad_i	w	C_max	C_min	Q	Cd
9071912	0.517	0.3	0.83	0.81	0.64	0.866
9114057	0.606	0.7	0.25	0.48	0.74	0.521
9106017	0.449	0.4	0.73	0.69	0.41	0.673
9149532	0.338	0.4	0.22	0.49	0.63	0.518
9119202	0.549	0.7	0.48	0.31	0.78	0.588
9088068	0.621	0.5	0.71	0.35	0.49	0.222
9095329	0.672	0.4	0.27	0.63	0.87	0.393
9145476	0.229	0.3	0.59	0.33	0.62	0.803
9065403	0.847	0.4	0.24	0.78	0.54	0.824
9094533	0.327	0.5	0.63	0.8	0.84	0.468
9117828	0.304	0.4	0.68	0.29	0.58	0.736
9142447	0.551	0.6	0.23	0.62	0.47	0.611
9157060	0.428	0.7	0.22	0.68	0.48	0.598
9089846	0.75	0.9	0.89	0.57	0.23	0.491
9061237	0.244	0.8	0.81	0.75	0.41	0.524

In Table 1 of the course parameter criteria, Course ID indicates the subdiscipline course code under the first level, with 7 digits, the first digit represents the school district, 2 to 4 digits represent the school site, the 5th digit represents the section of the physical education course, and 6 and 7 digits are the specific physical education training programs.

Rd_i and Rs_i indicate the relevance of the course to the physical training syllabus, respectively, and w_i is the course major weight value. Due to space limitations, only representative course information is listed in Table 1. The difficulty parameter evaluation table 2: C_e indicates the overall rating of the course; P_e is the average annual score of the course participation evaluation, the size of the value is determined by Ty_n, Te_n, Sy_n, Se_n, expert teachers and students on the course of the overall evaluation of the weight sum of one.

In table 3 of the course evaluation scale: Ad_i is the difficulty value given by the questionnaire; Q indicates the difficulty value of the course, and the indicator of the difficulty coefficient of the course Cd results given the difficulty coefficient interval of the course and the web voting questionnaire comes. The network voting questionnaire contains the questionnaire of the expert teachers of the system and the questionnaire of the students who have completed the course, the more the questionnaire volume can reflect the real situation. The weight of the difficulty value given to the course by the expert teachers and the students is 1.

In Table 4 of the course instructor information: the value of Rt_i is defined by ri, ri indicates the relevance of the course instructor’s ird major in the course he/she teaches, and Rt_i takes the value of the initial value set within the system by the system expert during the assessment of the course instructor.

Table 4.

Course teacher information list

Teacher ID	Course ID	Tf	Df	Rt_i	Tr
101728	9110234	0.5	0.8	0.4	0.595
102680	9098389	0.3	0.8	0.4	0.228
102704	9142310	0.5	0.4	0.5	0.477
101849	9144462	0.6	0.2	0.3	0.324
100797	9133783	0.8	0.2	0.2	0.628
100646	9137106	0.4	0.3	0.6	0.33
102173	9074481	0.8	0.2	0.2	0.569
102473	9123551	0.6	0.7	0.3	0.264
102819	9154533	0.5	0.8	0.5	0.888
100655	9072666	0.8	0.3	0.5	0.748

Table 5 shows the personalized training path solution results of the model for the sports courses. The multi-objective optimization problem ultimately needs to obtain the global optimal solution set rather than a single optimal solution, thus 10 mutually non-dominated optimal solutions are selected in the result set to represent the optimal 10 solutions for setting the course selection parameters in the multi-objective course-guided teaching system. The uniformly distributed weight Λ1, Λ2, Λ3, Λ4 serves as the preference information of the course selector 4 objectives.

Table 5.

Results of course selection

Scheme	Course ID	λ1	λ2	λ3	λ4	Tr	Pr	Cd	C_e
1	9152778	0.384	0.832	0.398	0.427	0.322	0.324	0.519	0.383
2	9072347	0.514	0.513	0.509	0.582	0.601	0.546	0.269	0.638
3	9091817	0.848	0.576	0.706	0.21	0.549	0.828	0.494	0.716
4	9146788	0.546	0.898	0.378	0.413	0.806	0.66	0.713	0.395
5	9133564	0.255	0.54	0.235	0.226	0.856	0.68	0.464	0.847
6	9151047	0.658	0.588	0.779	0.844	0.361	0.871	0.346	0.821
7	9142711	0.79	0.888	0.571	0.641	0.825	0.628	0.202	0.586
8	9134857	0.475	0.242	0.291	0.618	0.435	0.586	0.207	0.639
9	9132675	0.57	0.576	0.854	0.781	0.618	0.615	0.826	0.52
10	9117481	0.555	0.592	0.62	0.66	0.484	0.392	0.632	0.403

The algorithm in this paper improves the convergence accuracy and the distributability of the solution set, and successfully realizes the accurate recommendation of personalized training of sports courses by targeting the characteristics and willingness of individual flight trainees.

5

Conclusion

In this paper, the chaotic genetic algorithm is applied to the sports course combination training problem to solve the optimization problem of personalized training path for civil aviation flight trainees’ sports courses. Simulation experiments are carried out on historical data, and the results show that the number of better solutions of this paper’s algorithm is two times more than that of the traditional GA. Finally, the number of better solutions of this algorithm is stabilized at about 11, while the GA is 5. This algorithm has the characteristics of fast convergence speed and easy to converge to the global optimal solution. The use of floating-point coding is better than traditional GA, and the initial population is of excellent quality than traditional GA, which accelerates the convergence of the progeny individuals. In practical applications, it successfully achieves accurate recommendation of personalized training for sports courses for the characteristics and wishes of individual flight trainees.

Język:: Angielski

Częstotliwość wydawania:: 1 razy w roku
Dziedziny czasopisma:: Nauki biologiczne, Nauki biologiczne, inne, Matematyka, Matematyka stosowana, Matematyka ogólna, Fizyka, Fizyka, inne

Kanał RSS czasopisma

Personalized Training Path Design for Civil Aviation Flight Cadet Physical Education Course Based on Genetic Algorithm

Wei Song

Lina Wang

Data publikacji: 26 wrz 2025

Otrzymano: 14 sty 2025

Przyjęty: 16 kwi 2025

DOI: https://doi.org/10.2478/amns-2025-1044

Słowa kluczoweObjective function, Chaotic system, Genetic algorithm, Personalized training, Sports course

© 2025 Wei Song and Lina Wang, published by Sciendo.

This work is licensed under the Creative Commons Attribution 4.0 International License.

Słowa kluczowe
Objective function, Chaotic system, Genetic algorithm, Personalized training, Sports course