作为经济金融领域中最热门的话题之一，“金融市场波动率预测”除了能为投资者带来丰厚的收益，更对维护金融秩序、保持金融市场稳定有着重要的意义。本项目以提高金融市场波动率的预测精度为目的，提出一种新的模型平均法–模型平均异质自回归法（MAHAR）。在理论研究中，我们将论证MAHAR的大样本性质，特别是论证MAHAR的渐进最优性，为本项目的实证研究建立坚实的理论基础。实证研究分为两部分：（1）以预测美股波动指数VIX为例，验证MAHAR在金融大数据环境下的表现。我们会在MAHAR计算过程中引入由Janys Analytics公司创建的美国金融情绪指数USSI以提高预测精度，并研究和解决模型平均在大数据环境下应用会遇到的难题。（2）我们将重点研究MAHAR在中国金融市场上的实证应用。通过对中国金融市场以及个股波动率更精确的预测，为建立更加完善的金融预警机制提供重要参考。

As one of the most popular topics in economics and finance studies, a precise forecast of financial market volatility can bring huge profits to investors, and it is of great significance to maintaining financial market stability. In order to improve the forecast precision of financial market volatility, this project proposes a new model averaging method -- model averaging heterogeneous autoregression or simply MAHAR. We will first prove that MAHAR is asymptotically optimal in the sense of achieving the lowest possible mean squared forecast error, therefore, establish a solid theoretical foundation for the empirical studies of this project. We then apply the MAHAR method to forecast the CBOE VIX. Specifically, we will verify the empirical performance of MAHAR in the big data environment. We will introduce the U.S. sentiment index created by Janys Analytics in the MAHAR calculation process in order to improve forecast precision. We will study and solve the problems that model averaging encounters in the big data environment. Finally, we will concentrate on the empirical application of MAHAR in the Chinese financial market. By providing accurate forecasts of financial market volatility, our MAHAR method will contribute significantly on building up a comprehensive early-warning mechanism for the Chinese financial market.