摘要
采用基于长短时记忆神经网络(LSTM)模型预测金融交易中投资产品的未来价格,由此预判涨跌情况,考虑交易(佣金)成本,分析做多、做空两种交易方式的最佳策略.通过实验得出:当预测准确度接近50%或大于50%时,交易模型才能更大程度获利;在测试集为25%时获得收益达到最大,且四品种收益率高低顺序为比特币、原油、美元指数、黄金.改变相对佣金,对组合交易的偏向起到一定影响,收益呈梯度式变化.模型按一日交易一次进行了简化,并不能用于单日高频交易的分析.
Based on the long short-term memory neural network(LSTM)model,the rise and fall in the future price of investment products in financial transactions are predicted,and the best strategies for long and short trading methods are analyzed considering the transaction cost.Through experiments,it is concluded that the trading model can make a greater profit when the prediction accuracy is close to 50%or greater than 50%,and the maximum profit can be obtained when the test set is 25%,and the order of return of the four varieties is Bitcoin,crude oil,US dollar index,and gold.Changing the relative commission has a certain impact on the bias of portfolio transactions,and the return changes in a gradient style.The model is simplified on a daily basis and cannot be used for single-day high-frequency trading analysis.
作者
董涵
陈佳丽
王浩然
叶晓辉
Dong Han;Chen Jiali;Wang Haoran;Ye Xiaohui(School of Information Science&Technology,Xiamen University Tan Kah Kee College,Zhangzhou 363105,China;School of Economics and Management,Fuzhou Institute of Technology,Fuzhou 350506,China;School of Computing and Information Science,Fuzhou Institute of Technology,Fuzhou 350506,China)
出处
《南京师范大学学报(工程技术版)》
CAS
2023年第4期19-28,共10页
Journal of Nanjing Normal University(Engineering and Technology Edition)
基金
福建省中青年教师教育科研项目(JAT210609)
厦门大学嘉庚学院校级科研孵化项目(PY2023L01)。
关键词
期货交易
LSTM模型
单类交易
组合交易
futures trading
LSTM model
single class trading
portfolio trading