基于粒子滤波的期货价格随机分析与参数估计

发布时间：2018-04-19 12:36

本文选题：粒子滤波 + 卡尔曼滤波　；参考：《上海交通大学》2014年硕士论文

【摘要】：本文主要对股指期货价格进行分析，以双因子随机模型对期货价格进行建模，并使用粒子滤波方法对状态和参数进行联合估计。在对期货价格进行建模时，考虑到股指期货的特性，依然使用短期波动和长期均衡作为建模的因子，来反映短期内期货价格有一定的均值回复现象，而长期会因为国家经济环境，产业政策等因素呈现随机不确定性。针对已建立的状态空间模型，本文使用粒子滤波，而非通常使用的卡尔曼滤波。卡尔曼滤波主要适用于线性高斯动态系统，针对非线性的情况，扩展卡尔曼滤波和无味卡尔曼滤波相应作了近似调整，，而针对非高斯噪声的情况，卡尔曼滤波的估计效果将降低。然而粒子滤波在对状态空间建模上没有任何限制，面对非线性非高斯的模型，粒子滤波将显得游刃有余。最后本文通过以沪深300为标的的股指期货主力合约作为实证对象，来对期货价格进行状态和参数的联合估计。
[Abstract]:In this paper, the price of stock index futures is analyzed, the futures price is modeled by double factor stochastic model, and the state and parameters are estimated by particle filter.In the modeling of futures prices, considering the characteristics of stock index futures, we still use short-term volatility and long-term equilibrium as modeling factors to reflect the phenomenon of average return of futures prices in the short term.And long-term because of the national economic environment, industrial policy and other factors presented random uncertainty.For the established state space model, particle filter is used instead of Kalman filter.The Kalman filter is mainly suitable for the linear Gao Si dynamic system. For the nonlinear case, the extended Kalman filter and the tasteless Kalman filter are approximately adjusted accordingly, while for the non- noise, the extended Kalman filter and the tasteless Kalman filter are adjusted approximately.The estimated effect of Kalman filter will be reduced.However, particle filter has no limitation on the modeling of state space. In the face of nonlinear and non- model, particle filter will be able to work well.Finally, this paper uses the main stock index futures contract of Shanghai and Shenzhen 300 as the empirical object to estimate the state and parameters of the futures price.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F724.5;F224

【参考文献】