CAJ | 학술논문

针对传统人工神经网络中的BP(back propagation)神经网络自身局限以及其迭代次数多、收敛精度不高和泛化性差等缺点,提出了一种基于粒子群(particle swarm optimizer,PSO)算法的BP神经网络优化证券投资组合方法.在BP神经网络优化方法中,采用PSO算法替代了BP神经网络的梯度下降法,得到最优解,从而对BP神经网络模型进行优化.将该方法应用于证券投资组合的优化中,实验结果证明:该优化方法优于传统的BP神经网络优化方法.
침대전통인공신경망락중적BP(back propagation)신경망락자신국한이급기질대차수다、수렴정도불고화범화성차등결점,제출료일충기우입자군(particle swarm optimizer,PSO)산법적BP신경망락우화증권투자조합방법.재BP신경망락우화방법중,채용PSO산법체대료BP신경망락적제도하강법,득도최우해,종이대BP신경망락모형진행우화.장해방법응용우증권투자조합적우화중,실험결과증명:해우화방법우우전통적BP신경망락우화방법.
In response to the shortcomings in traditional artificial neural network,such as excessive iterations, low convergence accuracy and poor generalization of BP(Back Propagation)neural network,an optimized portfolio approach of BP neural network based on particle swarm algorithm is proposed. According to the BP neural network optimization method,the gradient descent method is replaced by the PSO algorithm,which ob-tains the optimal solution so as to optimize the BP neural network model. Through the application of the pro-posed method into the optimization of the portfolio,this study shows that the method is obviously superior to the traditional BP neural network optimization.