基于人工神经网络模型的金融板块指数走势研究  

Research on the Trend of Financial Sector Index Based on Artificial Neural Network Model

作　　者：杨义 章剑林[1] 刘闯[1] YANG Yi;ZHANG Jian-lin;LIU Chuang(Alibaba Research Center for Complexity Sciences,Hangzhou Normal University,Hangzhou 310036,China)

机构地区：[1]杭州师范大学阿里巴巴复杂科学研究中心,浙江杭州310036

出　　处：《软件》2020年第8期140-146,185,共8页Software

摘　　要：分析并预测股票市场中板块指数的涨跌是自股票市场创立以来,受到持续关注的研究热点之一。但由于股票市场具有非线性的时序特征,使得这一研究方向进展得颇为坎坷。而神经网络恰好在一定程度上可以捕捉非线性特征,这给研究带来了一种可能的途径。本文基于长短期记忆网络(LSTM)和全连接神经网络(FCNN)设计模型,将大盘行情指数、关联板块指数和金融板块三个方面的历史价格和成交量以及十年期国债收益率的历史价格作为输入,对TDX金融行业指数涨跌的走势进行研究。实证结果表明使用39天的先验数据使得走势预测效果最优,达到了理想的预测效果,且没有出现过拟合。Analyzing and predicting the rise and fall of the sector index in the stock market has been one of the hot topics since the establishment of the stock market.However,due to the non-linear temporal characteristics of stock market,this research direction is rather bumpy.However,the neural network can capture the nonlinear features to some extent,which brings a possible approach to the research.Based on the design model of long and short term memory network(LSTM)and fully connected neural network(FCNN),this paper studies the trend of the rise and fall of the TDX financial index by taking the historical price and trading volume of the three aspects of the market index,the related sector index and the financial sector and the historical price of the 10-year Treasury bond yield as input.The empirical results show that using the 39-day prior data makes the trend prediction effect optimal,achieving the ideal prediction effect,and there is no overfitting.

关 键 词：神经网络 板块指数 国债 数据预处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]