基于深度卷积神经网络的股票交易模型研究  

Research on Stock Trading Model Based on Deep Convolutional Neural Network

作　　者：牛晓健[1] 吴宇轩 NIU Xiaojian;WU Yuxuan(School of Ecomomics,Fudan University,Shanghai 200433)

机构地区：[1]复旦大学经济学院,上海200433

出　　处：《贵州商学院学报》2024年第4期24-36,共13页Journal of Guizhou University Of Commerce

摘　　要：为探讨卷积神经网络模型(CNN模型)的有效性,构建基于卷积神经网络模型的沪深300选股策略。首先通过分层法和IC测试法对CNN模型预测得到的上涨因子进行有效性测试;其次基于年化收益率、最大回撤等风险与收益指标,判断上涨因子选股策略的具体表现,进一步验证CNN选股模型的有效性。随后构建了基于宽基指数的择时策略,结果表明CNN模型在上证50指数上预测表现最佳。卷积神经网络的量化选股和择时模型的研究结论证实,卷积神经网络不仅能在沪深300中选出表现更好的股票,而且在量化择时方面也同样有效。To explore the effectiveness of the Convolutional Neural Network(CNN)model,a stock selection strategy for the Shanghai and Shenzhen 300 based on the CNN model is constructed.Firstly,the effectiveness of the upward factor predicted by the CNN model is tested with the usage of the layering method and IC testing method;secondly,based on risk and return indicators such as annualized returns and maximum drawdown,the specific performance of the rising factor stock selection strategy is judged to further verify the effectiveness of the CNN stock selection model.Subsequently,a timing strategy based on broad-based index was constructed,and the results showed that the CNN model performed the best in predicting the Shanghai Stock Exchange 50 Index.The research conclusion of the quantitative stock selection and timing model of Convolutional Neural Network confirms that convolutional neural network can not only select stocks with better performance in the Shanghai and Shenzhen 300,but also be equally effective in quantitative timing.

关 键 词：深度学习 卷积神经网络 量化选股 量化择时 

分 类 号：F832.5[经济管理—金融学]