中文摘要：

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。

英文摘要：

With the coming intelligent era and increasing demand for financial data analysis, deep learning has become the foreland in financial filed, especially in the prediction of financial market movements, text information processing and trading strategies improvement. Deep learning contains deep neural networks, deep belief networks and other structures. They extract deep features through the layered structure, strengthen important factors and filter the noise, which is of great significance to improve forecast accuracy. The applications and the variable optimization technologies of deep learning improve the forecast and analysis method in financial filed, make empirical research paradigm shift from linear to nonlinear and the concern from significance of parameters to the structure and dynamic characteristics of models, and also contribute to enriching financial and economic theories. It is important and difficult to build models with suitable structure and steady effect to capture the effective characteristics of financial data and interpret economic meaning with deep learning. The future research about deep learning in financial filed can be from these aspects, such as digging deep economic significance, refining the general analysis framework and exploring the applicability for heterogeneous information.