您的位置:首页  > 论文页面

结合粒子群算法的神经网络预测模型

发表时间:2019-08-30  浏览量:2019  下载量:247
全部作者: 张煌,梁朋,肖琨武,甘俊哲,施三支
作者单位: 长春理工大学理学院;长春理工大学光电工程学院;长春理工大学电子信息工程学院
摘 要: 股票数据量大,走势不稳定,峰值尖锐且通常噪声较多,这些因素都给预测带来了一定的难度,而目前现有的单一方法预测模型往往准确率较低。故针对现有预测模型的现状,本文意图将多种算法相结合从而寻求出一种最优的组合模型。沪深300指数样本的选择覆盖了大部分证券市场,具有很强的代表性,因此本文针对沪深300指数数据首先比较了同样为前馈式神经网络的径向基函数 (radial basis function,RBF)与BP(back propagation )两种神经网络模型,通过两种模型的比较得出在沪深300指数的预测上BP神经网络更具有优越性。因此,基于BP神经网络模型,本文加入了粒子群算法与之相结合,先针对沪深300指数数据进行预处理,寻求最优权值,再利用BP神经网络进行学习分析,有效减少了模型的误差,最终的准确率能够达到99.13%. 因此,结合粒子群算法的BP神经网络模型能够明显减小误差,提高预测的精确度。
关 键 词: 计算数学;沪深300指数预测;粒子群算法;径向基函数神经网络;BP神经网络
Title: Neural network prediction model combined with particle swarm optimization
Author: ZHANG Huang, LIANG Peng, XIAO Kunwu, GAN Junzhe, SHI Sanzhi
Organization: School of Science, Changchun University of Science and Technology; School of Optoelectronic Engineering, Changchun University of Science and Technology; School of Electronic Information Engineering, Changchun University of Science and Technology
Abstract: Stock data is large, and its trend is unstable. The peak of it is sharp and it has a lot of noise. These factors all bring some difficulties to the prediction. However, the existing single method prediction model tends to have lower accuracy. Therefore, in view of the current status of existing prediction models, this paper intends to combine various algorithms to find an optimal combination model. The selection of the CSI 300 index samples covers a strong representation of most securities markets. Therefore, this paper first compares the radial basis function (RBF) and back propagation (BP) of the feedforward neural network for the CSI 300 index data. Compared with two neural network models, the comparison results obtained by the two models show that the BP neural network is superior in the prediction of the CSI 300 index. Therefore, based on the BP neural network model, this paper combines the particle swarm optimization algorithm and combines it with the pre-processing of the CSI 300 index data to find the optimal weight. Then BP neural network is used for learning analysis, which effectively reduces the error of the model. The final accuracy rate can reach 99.13%. Therefore, the BP neural network model combined with the particle swarm optimization algorithm can significantly reduce the error and improve the accuracy of the prediction.
Key words: computational mathematics; CSI 300 index prediction; particle swarm optimization; radial basis function neural network; BP neural network
发表期数: 2019年8月第4期
引用格式: 张煌,梁朋,肖琨武,等. 结合粒子群算法的神经网络预测模型[J]. 中国科技论文在线精品论文,2019,12(4):537-544.
 
9 评论数 1

用户duantao0601:全文论理清晰严密,语言还需精炼准确

2020-03-28 21:22:43
友情链接