摘 要:股票数据是具有高度复杂性的时间序列数据。与传统的机器学习和一般的神经网络相比,长短期记忆神经网络处理股票数据具有更好的效果。但是该模型的性能表现对预置参数的设置具有极高的要求,缺乏经验的预置参数会降低其泛化能力和预测性能。针对以上问题,文章提出基于长短期记忆神经网络和改进的粒子群算法的深度学习复合模型 MPSO-LSTM。实证表明,该模型的预测性能相比于其他模型具有显著优势,验证了该方法的可行性。
关键词:股票指数预测;时间序列;改进粒子群算法;长短期记忆神经网络
DOI:10.19850/j.cnki.2096-4706.2021.04.001
基金项目:国家自然科学基金项目(61672337)
中图分类号:TP301.6 文献标识码:A 文章编号:2096-4706(2021)04-0001-04
CHEN Zhiquan,LEI Jingsheng
(Shanghai University of Electric Power,Shanghai 200090,China)
Abstract:Stock data is time series data with high degree of complexity. Compared with traditional machine learning and general neural networks,long short-term memory neural network has better effects in processing stock data. However,the performance of this model has extremely high requirements for the setting of preset parameters. Inexperienced preset parameters will reduce its generalization ability and predictive performance. Aiming at the above problems,a deep learning composite model MPSO-LSTM based on long short-term memory neural network and improved particle swarm algorithm is proposed. The empirical evidence shows that the prediction performance of this model has significant advantages compared with other models,which verifies the feasibility of this method.
Keywords:prediction of stock index;time series;improved particle swarm algorithm;long short-term memory neural network
参考文献:
[1] 吴威廷 . 股票基本面分析的主要内容 [J]. 财经界,2020(20):49-50.
[2] 郭平刚 . 技术分析法的逻辑基础 [J]. 中国证券期货,2012(11):283.
[3] 李悦 . 基于支持向量机的股市预测 [D]. 上海:复旦大学,2014.
[4] 彭燕,刘宇红,张荣芬 . 基于 LSTM 的股票价格预测建模与分析 [J]. 计算机工程与应用,2019,55(11):209-212.
[5] CHUNG J,GULCEHRE C,CHOK H,et al. Empirical evaluation of gated recurrent neural networks on sequence modeling [J/OL].arXiv:1412.3555 [cs.NE].(2014-12-11).https://arxiv.org/abs/1412.3555.
[6] HOCHREITER S,SCHMIDHUBER J. Long Short-Term Memory [J].Neural Computation,1997,9(8):1735-1780.
[7] KENNEDY J,EBERHART R. Particle Swarm Optimization [C]//Icnn95-international Conference on Neural Networks.Perth:IEEE,2002:1942-1948.
[8] BANSAL J C,SINGH P K,SARASWAT M,et al. Inertia weight strategies in particle swarm optimization [C]//2011 Third World Congress on Nature and Biologically Inspired Computing.Salamanca:IEEE,2011:633-640.
[9] 胡建秀,曾建潮 . 微粒群算法中惯性权重的调整策略 [J].计算机工程,2007(11):193-195.
作者简介:陈志全(1993—),男,汉族,湖北武汉人,硕士研究生,研究方向:股票预测;雷景生(1966—),男,汉族,陕西韩城人,教授,博士,研究方向:智能电网、电力大数据、机器学习、数据挖掘。