深度学习的金融实证应用:动态、贡献与展望

发布时间：2018-04-01 15:17

本文选题：深度学习　切入点：金融市场预测　出处：《金融研究》2017年05期

【摘要】：随着智能时代来临以及金融数据分析需求提升,深度学习已经成为金融领域中的应用前沿,特别是在预测金融市场运动、处理文本信息、改进交易策略方面。深度学习包含深度神经网络、深度信念网络等多种结构,通过分层结构提取深层特征,强化重要因素、过滤噪音,对提升预测准确率具有重要意义;其应用及由此衍生的优化技术改进了金融领域预测分析方法,促使实证研究范式从线性向非线性转变、从关注参数显著性向关注模型结构和动态特征转变,同时为丰富金融经济理论做出贡献。构建结构合适、效果稳健的模型以捕捉金融数据有效特征并进行经济含义阐释是应用深度学习方法的难点与重点;未来研究可以从挖掘深层经济意义、提炼一般性预测分析框架、探索其对异质信息的适用性等角度展开。
[Abstract]:With the advent of the intelligent age and the increasing demand for financial data analysis, in-depth learning has become the application frontier in the field of finance, especially in forecasting financial market movements, processing text information, and improving trading strategies.Depth learning includes deep neural network, deep belief network and so on. It is very important to extract the deep features, strengthen the important factors and filter the noise through the hierarchical structure, which is of great significance to improve the accuracy of prediction.Its application and the optimization techniques derived therefrom improve the forecasting and analysis methods in the financial field, and promote the paradigm of empirical research to change from linear to nonlinear, from the attention to the significance of parameters to the attention to the structure and dynamic characteristics of the model.At the same time, to enrich the financial and economic theory to make a contribution.It is difficult and important to construct a well-structured and robust model to capture the effective characteristics of financial data and explain the economic meaning.The general prediction analysis framework is refined to explore its applicability to heterogeneous information.
【作者单位】：中央财经大学;中国人民大学国际货币研究所;中国传媒大学;
【基金】：国家哲学社会科学基金重大项目(15ZDC024) 国家自然科学基金面上项目(71473279)资助
【分类号】：F830

【相似文献】