基本面量化投资融合了基本面分析与量化投资，是金融科技和量化投资研究中的新热点。现有研究已经提出了数以百计从基本面出发并能提供超额收益的市场异象因子，但因子过多也导致传统的资产定价研究方法难以深入挖掘其模式。深度学习专门研究从数据中自动地学习出有效模式以进行预测，能从不同的视角发现异象因子与超额收益之间的预测关系。由此，本项目拟通过深度学习深入研究基本面量化投资。首先，基于现有的海量异象因子，借鉴深度学习方法提出适用于中国市场的、更为有效的超额收益预测模型，并筛选和提取重要因子；其次，建立考虑重要因子的量化投资组合模型，最终设计出深度学习驱动的基本面量化投资算法，并通过实证研究检验其在中国市场的绩效；最后，依托于所设计的量化投资算法，设计计算实验定量地研究量化投资对资产定价和金融市场的影响。本课题将有助于加深对中国市场股票截面收益影响因素的理解，为基本面量化投资的发展和监管提供理论基础。

Quantamental investing, which combines fundamental analysis and quantitative investing, has become a hot topic in the research of financial technologies and quantitative investments. Existing studies have proposed hundreds of market anomaly factors that can provide excess return. Unfortunately, too much factor leads to the fact that traditional asset pricing methods cannot fully exploit efficient patterns. Deep learning focuses on automatically exploiting efficient patterns from data and aims for better prediction and provides a different aspect to exploit the prediction relationship between factors and asset’s excess return. Therefore, this proposal aims to study the quantamental investing problem via deep learning technology. Firstly, we will propose new alpha prediction methods based on the huge amount of anomaly factors. The new methods will leverage the deep learning techniques and are deemed to be more suitable for the Chinese financial markets and more efficient than existing methods. By evaluating the factors’ marginal importance in the prediction model, we can identify and extract important factors. Secondly, we will construct quantitative portfolio models that consider the important factors and finally design the deep learning driven quantamental investing algorithms. We will then empirically evaluate their performance using Chinese historical data. Finally, based on the designed algorithms, we will further study the impact of new quantamental investing methods on the financial markets and asset pricing via computational experiments method. This project will deepen our understanding of the factors that affect Chinese stocks’ cross-sectional return. It will also provide theoretical foundation for the development and supervision of quantamental investing.