Applications of certain non-Gaussian processes in financial mathematics

本研究项目将探讨在更现实的资产价格非高斯利维模型下金融工程产生的一些问题以及人工智能和大数据在金融中的应用：*****1。开发 Levy 模型下金融工程的有效数值方法。例如，当资产价格遵循一些更现实的非高斯过程模型时，尝试探索期权价格和最优投资组合可能的解析或近似解析公式。如果这样的公式不存在或者很难找到，那么我将尝试设计高效的蒙特卡罗/准蒙特卡罗（QMC）模拟方法。尽管我在具有单一从属者的更现实的 Levy 模型下的期权定价方差减少 (VR) 方法方面取得了一些显着的结果，但这些 VR 方法非常依赖于问题。我计划继续研究有效方差 VR 与 QMC 方法相结合，用于奇异的多资产期权，以及时变布朗运动 (TCBM) 模型下的多附属机构资产价格的金融工程问题。 ******2.进一步利用 Malliavin 微积分在金融中的应用。在过去的二十年里，Malliavin 微积分通过以下方式成功地应用于金融领域：期权敏感性或希腊字母的估计；模拟美式期权；然而，大部分工作都是针对资产价格的几何布朗运动模型完成的。我做了一些关于多资产期权希腊字母的模拟和多资产美式期权价格的模拟以及在资产价格单一从属的 TCBM 模型下模拟希腊字母的工作。我计划将我之前的工作扩展到多资产价格遵循具有多附属机构的 TCBM 模型的情况。******3。继续研究更现实的资产价格 Levy 模型下的投资组合优化（PO）问题。我在 PO 问题上取得了一些不错的成果。我计划继续研究更现实的资产价格模型下的 PO 和养老基金投资问题。我会尝试通过传统方式和Malliavin演算方法来寻找最优投资组合和有效边界等。 ******4。大多数关于衍生品定价和投资组合优化的论文都是在资产价格遵循某些随机过程的假设下进行讨论的。这种方法有一些限制：(1)．测试随机过程模型是否能够很好地模拟资产价格并不容易。 （2）。估计给定随机模型的参数也很困难，特别是对于多资产情况。因此，我们计划利用人工智能和大数据分析的优势，尝试“无模型​​”的衍生品定价和投资组合优化方法。这个方向还有很多问题值得探讨。*****

This research program will explore some problems arising from financial engineering under more realistic non-Gaussian Levy models for asset prices and applications of artificial intelligence and big data in finance:******1. To develop efficient numerical methods for financial engineering under Levy models. For example, try to explore possible analytic or approximate analytic formulas for option prices and optimal portfolios when asset prices follow some more realistic non-Gaussian process models. If such formulas do not exist or are hard to find, then I will try to design efficient Monte Carlo / quasi-Monte Carlo (QMC) simulation methods. Although I have obtained some remarkable results on variance reduction (VR) methods for option pricing under more realistic Levy models with single subordinator, these VR methods are quite problem dependent. I plan to keep working on efficient variance VRs combined with QMC methods for exotic multi-asset options and for financial engineering problems under time-changed Brownian motion (TCBM) models for asset prices with multi-subordinator. ******2. To make further use of Malliavin calculus in finance. In the past twenty years, Malliavin calculus was successfully applied in finance in the following ways: in the estimation of option sensitivities or Greek letters; in the simulation of American style options; in the estimation of optimal portfolios, etc. However, most of the works were done for Geometric Brownian motion models for asset prices. I have done some work on simulations of Greek letters for multi-asset options and simulations of multi-asset American option prices as well as their Greek letters under TCBM models with single subordinator for asset prices. I plan to extend my previous work to the cases where the multi-asset prices follow the TCBM models with multi-subordinators.******3. To continue to work on portfolio optimization (PO) problems under more realistic Levy models for asset prices. I have achieved some good results for PO problems. I plan to continue to work on PO and pension fund investment problems under more realistic asset price models. I will try to find optimal portfolios and efficient frontiers, etc. by both the traditional way and the Malliavin calculus method. ******4. Most papers on derivative pricing and portfolio optimizations are discussed under the assumption that asset prices follow certain stochastic processes. There are some restrictions on this approach: (1). It is not easy to test whether a stochastic process model can model the asset prices very well. (2). It is also hard to estimate parameters for a given stochastic model, especially for the multi-asset case. Thus, we plan to try a "model-free" approach for derivative pricing and portfolio optimizations by using the advantage of artificial intelligence and big data analytics. There are lots of problems in this direction worth to be explored.*****