Earticle

본 연구는 저출생 여파로 학령인구가 감소하는 추세에 금오공과대학교의 입시 지원자수 감소로 인한 피해를 완 화 또는 방지하고자 수시 입시생들에게 모집 수, 지원자수, 성적현황을 예측해 줌으로써 지원자 증대에 목적을 두고 있 다. 시계열 데이터를 활용하기 위해 LSTM 순환신경망을 사용하고, matlab의 딥러닝 기능을 이용하여 학습을 한 후 향 후 2년간의 예측치를 도출한다. 딥러닝 실행 후 훈련진행상황을 통해 학습상태를 확인하고 2015년부터 2023년까지 예측 데이터를 도출한 뒤 실제 데이터와 수치적 유사성과 흐름의 유사성을 파악하고, 유의성 검정을 거쳐 훈련 모델의 수치 적인 신뢰도를 확보하였다. 이 수치적인 신뢰를 바탕으로 2024, 2025년의 수시 모집 수, 지원자수, 성적현황의 예측치를 공개하였다. 그러나 LSTM 모델을 사용하기엔 데이터양이 많이 부족하여 과적합이 일어날 수 있는 문제점과 상호연관 도가 깊은 데이터 활용도가 낮아 향후연구에서는 이부분을 보완한다면 더 나은 결과를 도출할 수 있을 것으로 보인다.

The purpose of this study is to increase the number of applicants by predicting the number of applicants, the number of applicants, and their grades to alleviate or prevent the damage caused by the decrease in the number of applicants for the entrance examination at Kumoh Institute of Technology in the trend of decreasing school age population in the aftermath of low birth rates. are leaving In order to utilize time series data, an LSTM recurrent neural network is used, and after learning using matlab's deep learning function, a forecast for the next two years is derived. After running deep learning, check the learning status through training progress, derive predicted data from 2015 to 2023, identify numerical similarity and flow similarity with actual data, and verify the numerical reliability of the training 모델 through significance test has been secured. Based on this numerical trust, the predictions of the number of regular recruitment, the number of applicants, and the performance status for 2024 and 2025 were disclosed. However, due to the problem of overfitting due to the insufficient amount of data to use the LSTM model and the low utilization of highly correlated data, it is expected that future research will be able to derive better results if these areas are supplemented.

This study shows a variance regularization method in order to obtain the stable optimal solutions, when a numerical method accelerating design variables is used for material topology optimization algorithm. Since a moved and regularized Heaviside function used in the accelerated method is composed of nonlinear concave and convex functions in a given design domain between 0 and 1, design variables below 0.5 can move fast toward the value of 0 and those over 0.5 are rapidly located to the value of 1. However optimal solutions may be not stable due to singularity of element stiffness, while the accelerated design variables are too closed to value 0. In particular this instability may occur to the accelerated method-based material topology optimization algorithms much repeating the moved and regularized Heaviside function. In order to resolve the problem, in this study, a variance regularization is formulated within a linear governing equation for structural analyses of optimization procedures. Numerical examples for topologically optimally modeling a linear elastostatic MBB-beam verify that the accelerated method of design variables take numerical stability of topological optimal solutions by being associated with the variance regularization method.

The krill herd (KH) algorithm is a novel swarm intelligent algorithm which is inspired the herding behavior of the krill swarms. The various test results in the relevant literature show that the KH algorithm has better performance than the other swarm intelligent algorithm for optimization problem. In order to further improve the performance of the KH algorithm, an improved KH is proposed in this paper. The algorithm is performed on ten test functions and the results are compared with the basic KH algorithm, PSO, DE and GA algorithm. The experimental results indicate that the improved algorithm is a good method for numerical optimization problem.

Genetic algorithm (GA) is a population-based approach for heuristic search in optimi- zation problems based on the principle of biologic evolution and natural selection. In this paper, we present a hybrid adaptive genetic algorithm with chaos searching technique for numerical optimization. On the one hand, two sets of crossover and mutation rates are for- mulated to automatically maintain the balance between exploration and exploitation during the genetic search process. On the other hand, the chaos searching technique is introduced into the adaptive genetic algorithm based on the decision mechanism for premature conver- gence adopted in this paper, whose main goal is to avoid being trapped into the local opti- mum. In addition, half of the total evolutionary generation is utilized as one of the decision conditions so as to speed up the convergent process. To validate the effectiveness and efficiency of the proposed approach, we apply it to four benchmark functions obtained from the literature, and the experimental results show that the proposed algorithm can find global optimal or the closer-to-optimal solutions and have faster search speed as well as higher convergence rate.

The scalability performance of the traditional evolutionary algorithms (EAs) deteriorates rapidly as the dimensionality of the optimization problems increases. Therefore, cooperative coevolutionary (CC) framework is proposed to overcome the defect. Different from existing CC algorithms, a novel self-adaptive based cooperative coevolutionary (SaCC) algorithm is presented in this paper. The SaCC employs three algorithms which with self-adaptive mechanism as sub-algorithms. The focus of this paper is on investigating two different cooperative coevolutionary manners. In the first manner, SaCC executes its sub-algorithms in parallel during evolve process and the corresponding algorithm is termed as SaCC-M1. In the second manner, SaCC executes its sub-algorithms in serial and the corresponding algorithm is termed as SaCC-M2. 26 test functions with 1000 dimensionalities are employed to verify the validity of SaCC-M1 and SaCC-M2. Experiment results demonstrate that SaCC-M2 outperforms its sub-algorithms and SaCC-M1. Besides, the results indicate that serial manner is another simple yet efficient manner for CC algorithms to solve large-scale global optimization problems.

A two-bar plane truss is builtin MATLAB based on mathematical model. Then the authors use genetic algorithm toolbox to solve this problem. The parametric truss model is set up in the finite element analysis software ANSYS. It is analyzed by first-order algorithm. The comparison of two kinds results show the pure genetic algorithm doesn’t always have an advantage over other algorithms. In the end, a joint optimization method is put forward on the basis of genetic algorithm. It combines genetic algorithm based on MATLAB toolbox and numerical algorithm based on quasi-Newton method. This method is illustrated by the numerical example of the two-bar plane truss. The results show this joint optimization method can get the global optimal solution of this problem every time.

In order to obtain the minimum supply and demand scheduling under Cloud manufacturing platform, the fisher fishing algorithm is applied in it. Firstly, the optimization algorithm of fisher fishing is studied. Secondly, the supply and demand scheduling mode under Cloud manufacturing platform is constructed, and the corresponding optimization mathematical model is established. Finally, the simulation results of supply and demand scheduling time is carried out, results show that the fisher fishing algorithm is an effective tool.

Heuristic optimization using hybrid algorithms have provided a robust and efficient approach for solving many optimization problems. In this paper, a new hybrid algorithm using adaptive genetic algorithm (aGA) and particle swarm optimization (PSO) is proposed. The proposed hybrid algorithm is applied to solve numerical optimization functions. The results are compared with those of GA and other conventional PSOs. Finally, the proposed hybrid algorithm outperforms others.