摘要
艾略特波浪理论作为金融市场的研究工具,描述了股价的结构规律。针对艾略特波浪理论,结合人工智能方法,以时间序列为基础,提出并比较了两种基于人工神经网络的分类器。第一种技术是结合了后向传播学习算法的多层人工神经网络,1 600次迭代后均方误差小于0. 87。根据传统后向传播网络的缺陷与金融市场的特性,提出第二种改进网络,即与模糊理论相结合的基于缩放共轭梯度算法的人工神经网络。经120次迭代后均方误差小于0. 22,相比于第一种方法,准确率提高74. 7%,收敛速度提高92. 5%。
Elliott wave principle is one of the research tools in financial market. The principle describes the structure of stock price with Elliott wave. Aiming at Elliott wave principle, we combined artificial intelligent method, proposed and compared two classifiers which were based on artificial neural network on the basis of time sequence. The first one was the multi-layer artificial neural network combined with back propagation algorithm. After 1 600 iterations, the mean square error was less than 0.87. According to disadvantages of traditional back propagation network and the characters of financial market, we proposed an improvement network. The second technique combined with fuzzy theory was an artificial neural network based on scaled conjugate gradient algorithm. After 120 iterations, the mean square error was less than 0.22. Compared with the first one, the second one increased by 74.7% in the accuracy rate and improved by 92.5% in the convergence speed.
作者
李音润
欧鸥
Li Yinrun;Ou Ou(College of Information Science and Technology,Chengdu University of Technology,Chengdu 610059,Sichuan,China)
出处
《计算机应用与软件》
北大核心
2018年第12期285-292,共8页
Computer Applications and Software
关键词
人工神经网络
模糊神经网络
艾略特波浪理论
后向传播算法
模糊理论
缩放共轭梯度算法
Artificial neural network
Fuzzy neural network
Elliott wave principle
Back propagation algorithm
Fuzzy theory
Scaled conjugate gradient algorithm