摘要
针对股价系统内部结构的复杂性、外部因素的多变性,分析了基于BP网络进行股市预测的原理,利用三层前馈神经网络对股市建立预测模型,探讨了网络的拓扑结构、隐节点个数确定的原则、样本数据的选取和预处理、初始参数的确定等问题。为了避免网络陷入局部最小点和提高网络的收敛速度,算法采用改进后的LM-BP,并与其他BP算法进行比较。以最具代表性的上证指数为例,仿真实验表明了经过对筛选后的样本学习,并对所建的预测模型进行训练后,该LM-BP算法能够对有短期上证指数走势进行有效稳定预测。
Aiming at the complexity of inside structure and levity of exterior complication in the system of stock market which make stock market prediction a complex problem, a method of modeling stock marketusing BP neural network with three-layer feed forward neural network that is based on thorough study of the difficult problems facing stock predication is proposed. The problems including the structure of net-work, the number of hidden nodes, the choose and pretreatment of swatch datum and the determination of preliminary parameters have been discussed. In order to avoid local extremum and promote convergencespeed, Levenberg-Marquardt BP algorithm has been adopted. The performance of standard BP algorithm and other ameliorated BP algorithm has been compared in experiments. At last Shanghai Stock Index havebeen applied to train the established network model, then stock datum have been predicted using the trained network and good effect has been gained.
出处
《系统管理学报》
北大核心
2009年第6期667-671,共5页
Journal of Systems & Management
关键词
股票指数预测
BP神经网络
隐层节点数确定
样本选取策略
stock index prediction
BP neural network
number of hidden nodes
choose and pretreatment of swatch datum