非线性框架下指数期货套利策略研究

发布时间：2018-01-01 03:14

本文关键词：非线性框架下指数期货套利策略研究　出处：《浙江大学》2014年硕士论文　论文类型：学位论文

【摘要】：2010年沪深300股指期货和融资融券交易的相继推出完善了证券市场双向套利机制和衍生品的发展,股指期货成为投机、套利和套期保值的重要工具,在这样的背景下有许多研究者对股指期货市场的特征、套利技术等进行研究,本文基于以前学者的成果,使用非线性的方法对股指期货期现套利的两个重要问题进行探讨。指数复制是期现套利和指数型基金管理的重要环节,本文采用抽样复制法和优化方法,构建股票组合来跟踪沪深300股指,基于支持向量机(SVM)结构风险最小化的观念,建立了目标函数,确定约束条件,保持在整个跟踪过程中成份股权重不变。使用遗传算法确定股票组合并对支持回归机的参数进行优化,建立SVM模型,记录样本内外的跟踪误差并与二次规划方法所建立的模型进行比较。发现SVM模型在测试集上的表现要优于二次规划的模型,说明SVM方法基于结构风险最小化的优良特质使得模型的泛化能力得到提高。基差预测是实证的第二个部分,准确地判断基差的走势有利于投资者进行正向基差套利和反向基差套利。本文使用前21个交易目的股指期货交易数据建立了基差的预测模型,使用前一天的交易数据对第二天收盘价的基差进行预测。为了方便对各种模型进行比较,本文分别建立了线性回归模型、高斯径向核函数SVM模型、拉普拉斯径向核函数SVM模型、ANOVA径向核函数SVM模型、贝塞尔核函数SVM模型,从实证结果中可以看出非线性的SVM模型相对于线性回归模型的预测能力更为优异。
[Abstract]:In 2010, the Shanghai and Shenzhen 300 stock index futures and margin trading have launched the perfect development of the securities market two-way arbitrage mechanism and derivatives, stock index futures has become an important tool for speculation, arbitrage and hedging, there are many researchers on the stock index futures market characteristics in such a background, this paper studied the arbitrage technology, the achievements of previous scholars based on the two important problems of nonlinear method using stock index futures arbitrage are discussed.
Index replication is an important link of arbitrage and index fund management, this paper adopts sampling replication method and optimization method, construct the stock portfolio to track the CSI 300 index, the support vector machine (SVM) based on structural risk minimization concept, establishing the objective function to determine the constraints, keep in the whole process of tracking components of equity invariant. Using genetic algorithm to determine the parameters of the stock portfolio and support regression optimization, establish the SVM model, the tracking error and the sample records and compared the two methods of planning model. The performance of the SVM model on the test set is better than two times planning model, SVM method is excellent based on the characteristics of structural risk minimization model generalization ability is improved.
The second part is the empirical basis to predict, accurately determine the basis trend in favor of investors positive basis arbitrage and reverse arbitrage basis. This paper uses the first 21 trading stock index futures trading data to establish a prediction model of the basis, the use of transaction data, the day before the second day of the closing price of the basis in order to facilitate the prediction. Comparison of various models, this paper establishes a linear regression model, Gauss radial kernel function of the SVM model, Laplasse radial kernel function of the SVM model, ANOVA radial kernel function of the SVM model, the Bessel kernel SVM model, from the empirical results can be seen in the nonlinear model of SVM with respect to the predictive ability of linear regression model is more excellent.

【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F832.51

【参考文献】