Hybrid kernel extreme learning machine for evaluation of athletes' competitive ability based on particle swarm optimization

https://doi.org/10.1016/j.compeleceng.2018.10.017Get rights and content

Abstract

A physical education strategy benefit analysis method based on minimal flux parallel random tree algorithm is proposed. The physical education problem model with the optimization objective functions is put forward. The optimization mathematical model is established and its multi-objective weight adaptive form is presented. Then the parallel random tree algorithm is introduced to solve the physical education strategy benefit model. In order to further improve the performance of the parallel random tree algorithm their parameters are adaptively learned by using the minimal flux. Thus the convergence of the algorithm is improved. Finally the example analysis is performed to verify the effectiveness of the proposed physical education strategy benefit analysis algorithm.

Introduction

Physical education is deemed as an extremely heavy work. This is due to many uncertainties involved in lot of curriculums, students and venues, etc. [1], [2]. Especially in recent years, the expansion plan of enrollment in colleges and universities about how to achieve a more reasonable and efficient distribution of physical education curriculums is an important part. The colleges and universities pay more attention to it. In particular, with the concept of multi-school curriculum resource sharing, physical education has become increasingly important [3]. The essence of physical education problem is a multi-objective and multi-constraint NP difficult combination optimization problem [4]. There are many mature algorithms for solving such problems, such as branch-and-bound [5], grouping optimization algorithm [6], and association rules algorithm [7] etc. Such algorithms has good performance in solving NP difficult of combination optimization problem, but there are following unsolved problems: (1) in the solution process, the algorithm only solves certain problems, and cannot form a general physical curriculum arrangement method; (2) there are few discrimination judgment standards for pros and cons of the curriculum arrangement, and the algorithm overemphasizes the optimization in a certain direction, and cannot achieve global optimization; (3) The association rule method has difficulties in obtaining the universality association rule during the solution process, the solution results did not meet expectations.

How to obtain more accurate weights in the asset calculation process is the key to the problem solving in order to calculate the effective boundary in the physical education composition problem model. In addition to some classical algorithms, evolutionary computation algorithm can also be used in such work. Such as James, an American scholar, who designed the evolutionary computation algorithm based on the behavior of the bird flock in the process of food search, and obtained the parallel random tree algorithm (parallel random tree) [3]. The parallel random tree algorithm models the foraging bird as a "particle", and the algorithm is solved by optimizing in the search space set. The particle in the algorithm contains two main attributes: the adaptive value and the flight direction. The former can be determined with the set optimization function, and the latter can be determined with the direction parameter. In the optimization of the algorithm, the tendency of the particle running in the algorithm is toward the best position particle in the current population. The computation speed of the parallel random tree algorithm is related to the development level of the computer hardware, and with the development of hardware, the performance of the algorithm has also been continuously improved. In recent years, with the progress of the research on the combination of physical education, the application of evolutionary computation has become increasingly widespread. For example, the literature [4] uses the ant colony algorithm to carry on the principle analysis, builds the education problems model, and calculates the model weight. The literature [5] regards the education benefits as the main research objectives. It takes the expected benefits as the main constraints, and minimizes the weight of the expected benefits as the main objectives to obtain the best education benefits. The Literature [6] studies education problems using genetic algorithm, and improves the coding process of the algorithm in the process of model solving. The experimental results show that the proposed method has more education optimization performance. The literature [7] studies the rate of return problem in the education, and compares with the quadratic programming problem, so as to analyze the performance advantages of the proposed education problem. The literature [8] focuses on the quality of education at the same level of education. The main purpose is to realize the minimization of resource consumption. The literature [9] encodes the combination weights of education in a binary way, so as to obtain an optimized function, and takes risks and benefits as two objectives that can be balanced to form a multi-objective education optimization problem. Finally, a differential evolution algorithm is used to achieve the solution of the model. The literature [10] uses integer coding instead of binary coding to improve the genetic algorithm, and then uses the improved algorithm to solve the education model. Literature [11] uses transactional cost constraints derived from preferences for education.. The optimization algorithm is mainly genetic algorithm, and the constraint set is the maximum of the upper limit of the education. In the above mentioned algorithms, there are two problems in the research: the first is the optimization performance of the algorithm, the guarantee of the convergence of the algorithm, and the second is the suitable and efficient evaluation of the effect, and there is no effective evaluation index for the performance of the algorithm.

This paper mainly uses the parallel random tree algorithm to study the problem of physical education. It mainly involves two aspects, one is the improvement of the performance of the algorithm; the second is the efficient evaluation of the algorithm. This paper studies the problem model of physical education, proposes the objective functions and constraint conditions of the physical education problem, establishes its optimization mathematical model, and designs its multi-objective weight adaptive form. At the same time, it introduces the parallel random tree algorithm to solve the physical education strategy benefit model problem. In addition, in order to further improve the performance of the parallel random tree algorithm in solving process, the parameters related to the parallel random tree algorithm are adaptively learned with the minimal flux, and the convergence of the algorithm is improved.

The remaining part of the manuscript is organized as follows. Section 2 discuss about the representation of sports movement characteristics. Compression algorithm of parallel random tree test of minimal flux objective is presented in Section 3. The experimental analysis is explained in the Section 4 and Section 5 concludes the manuscript.

Section snippets

Representation of sports movement characteristics

Sports movement is a continuous dynamic process. With the help of the animation frame concept, it can be decomposed into multiple static body shapes. In this way, the characteristics of sports movement can be represented by a series of body features.

Random tree algorithm

The random tree is a tree-based simulation algorithm that generates and maintains the data structure of the tree during simulation. It can trace back to different states that were previously visited, which makes it more versatile than random walk based simulation algorithms (such as Monte Carlo). It realizes the gradual increase by adding margins between the existing states and the new states. Each node of tree simulates a point in the state space of physical education. Each edge is a short

Experimental analysis

The experimental subjects were selected from a certain university in China for the experiment verification of the physical education curriculum. The elements of physical education scheduling are shown in Table 1. The developed physical education scheduling system is implemented based on Visual C++.

The standard parallel random tree algorithm was selected as a comparison algorithm in the experiment, and the target function value and evolution time were selected as the evaluation indexes. The

Conclusion

A novel physical education training hybrid model was proposed. The hybrid combination involves parallel random tree algorithm with optimized mathematical model on minimum flux. A multi-objective weight adaptive form is designed for the same. This introduces the parallel random tree algorithm. Later this solves the physical education strategy benefit model designed. The parameters related to the parallel random tree algorithm are adaptively learned using the minimal flux. From here, the

Zhao Yanpeng is a lecturer in Shandong University of Science and Technology. He received his Master degree from the Capital University of Physical Education and Sports in 2011. He has been working in Shandong University of Science and Technology since his graduation. His research interests mainly focus on the physical education training.

References (11)

  • JenniferW. Chan et al.

    Amphiphilic macromolecule self-assembled monolayers suppress smooth muscle cell proliferation

    Bioconjugate Chem

    (2015)
  • N. Arunkumar et al.

    Automatic detection of epileptic seizures using permutation entropy, Tsallis entropy and Kolmogorov complexity

    J Med Imaging Health Inform

    (2016)
  • S.L. Fernandes et al.

    A novel nonintrusive decision support approach for heart rate measurement

    Pattern Recognit Lett

    (2017)
  • N. Arunkumar et al.

    Approximate entropy based ayurvedic pulse diagnosis for diabetics—a case study

  • M.P. Malarkodi et al.

    Gabor wavelet based approach for face recognition

    Int J Appl Eng Res

    (2013)
There are more references available in the full text version of this article.

Cited by (7)

  • Evolving kernel extreme learning machine for medical diagnosis via a disperse foraging sine cosine algorithm

    2022, Computers in Biology and Medicine
    Citation Excerpt :

    In the original KELM, grid search (GS) is employed to evolve the parameters, but in the actual operation process it is not easy to determine the parameter interval accurately because GS tends to fall into local optima. In order to solve the problem, a wide range of meta-heuristic algorithms have been used to optimize the model parameters [55–57], including particle swarm optimization (PSO) [32,41,58,59], genetic algorithm (GA) [60,61], simulated annealing (SA) [61], moth-flame optimization (MFO) [46], grey wolf optimization (GWO) [34,45,62–65], grasshopper optimization algorithm (GOA) [36,43,66], salp swarm algorithm (SSA) [67], and sine cosine algorithm (SCA) [68]. Although the sine cosine algorithm (SCA) was originally developed for numerical optimization problems and airfoil design problems [69], SCA has been used to solve numerous real optimization problems due to its simple structure and better optimization performance, including but not limited to, feature selection (FS) [70,71], optimal power flow (OPF) [72], hydrothermal scheduling [73], fault diagnosis [74], pairwise local sequence alignment [75], support vector regression (SVR) parameter optimization [76], object tracking [77,78], engineering design problems [79,80], wind speed forecasting [81], parameters optimization [82], image segmentation [83,84], train multilayer perceptions [85], simultaneous network reconfiguration [86], hyperspectral image [87], capacitated vehicle routing problem (CVRP) [88], and construction duration and schedule robustness optimization [89].

  • Prediction of effluent quality in papermaking wastewater treatment processes using dynamic kernel-based extreme learning machine

    2020, Process Biochemistry
    Citation Excerpt :

    As a kind of improved ELM, now it has been used in many areas. For example, Zhao used KELM as a base learner of the minimal flux parallel random tree algorithm, proving that this algorithm is effective for the evaluation of athletes’ competitive ability [37]. Besides, there are many other applications of KELM such as human activity recognition [38], carbon pricing prediction [39], medical diagnosis [40].

View all citing articles on Scopus

Zhao Yanpeng is a lecturer in Shandong University of Science and Technology. He received his Master degree from the Capital University of Physical Education and Sports in 2011. He has been working in Shandong University of Science and Technology since his graduation. His research interests mainly focus on the physical education training.

This paper is for CAEE special section SI-hai. Reviews processed and recommended for publication to the Editor-in-Chief by Associate Editor Dr. Gustavo Ramirez Gonzalez.

View full text